external/ton
1
0
Fork 0
mirror of https://github.com/ton-blockchain/ton synced 2025-02-12 11:12:16 +00:00
Code Issues Releases Wiki Activity
ton/tolk/ast.h

1056 lines
41 KiB
C
Raw Permalink Normal View History

[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
/*
This file is part of TON Blockchain Library.
TON Blockchain Library is free software: you can redistribute it and/or modify
it under the terms of the GNU Lesser General Public License as published by
the Free Software Foundation, either version 2 of the License, or
(at your option) any later version.
TON Blockchain Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public License
along with TON Blockchain Library. If not, see <http://www.gnu.org/licenses/>.
*/
#pragma once
#include <string>
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
#include "fwd-declarations.h"
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
#include "platform-utils.h"
#include "src-file.h"
#include "lexer.h"
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
#include "symtable.h"
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
/*
* Here we introduce AST representation of Tolk source code.
* Historically, in FunC, there was no AST: while lexing, symbols were registered, types were inferred, and so on.
* There was no way to perform any more or less semantic analysis.
* In Tolk, I've implemented parsing .tolk files into AST at first, and then converting this AST
[Tolk] Compilation pipeline, register global symbols in advance Since I've implemented AST, now I can drop forward declarations. Instead, I traverse AST of all files and register global symbols (functions, constants, global vars) as a separate step, in advance. That's why, while converting AST to Expr/Op, all available symbols are already registered. This greatly simplifies "intermediate state" of yet unknown functions and checking them afterward. Redeclaration of local variables (inside the same scope) is now also prohibited.
2024-10-31 07:04:58 +00:00
* into legacy representation (see pipe-ast-to-legacy.cpp).
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
* In the future, more and more code analysis will be moved out of legacy to AST-level.
*
* From the user's point of view, all AST vertices are constant. All API is based on constancy.
* Even though fields of vertex structs are public, they can't be modified, since vertices are accepted by const ref.
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
* Generally, there are three ways of accepting a vertex:
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
* * AnyV (= const ASTNodeBase*)
* the only you can do with this vertex is to see v->type (ASTNodeType) and to cast via v->as<node_type>()
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
* * AnyExprV (= const ASTNodeExpressionBase*)
* in contains expression-specific properties (lvalue/rvalue, inferred type)
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
* * V<node_type> (= const Vertex<node_type>*)
* a specific type of vertex, you can use its fields and methods
* There is one way of creating a vertex:
* * createV<node_type>(...constructor_args) (= new Vertex<node_type>(...))
* vertices are currently created on a heap, without any custom memory arena, just allocated and never deleted
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
* The only way to modify a field is to use "mutate()" method (drops constancy, the only point of mutation)
* and then to call "assign_*" method, like "assign_sym", "assign_src_file", etc.
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
*
* Having AnyV and knowing its node_type, a call
* v->as<node_type>()
* will return a typed vertex.
* There is also a shorthand v->try_as<node_type>() which returns V<node_type> or nullptr if types don't match:
* if (auto v_int = v->try_as<ast_int_const>())
* Note, that there casts are NOT DYNAMIC. ASTNode is not a virtual base, it has no vtable.
* So, as<...>() is just a compile-time casting, without any runtime overhead.
*
* Note, that ASTNodeBase doesn't store any vector of children. That's why there is no way to loop over
* a random (unknown) vertex. Only a concrete Vertex<node_type> stores its children (if any).
* Hence, to iterate over a custom vertex (e.g., a function body), one should inherit some kind of ASTVisitor.
* Besides read-only visiting, there is a "visit and replace" pattern.
* See ast-visitor.h and ast-replacer.h.
*/
namespace tolk {
enum ASTNodeType {
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
ast_identifier,
// expressions
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
ast_empty_expression,
ast_parenthesized_expression,
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
ast_tensor,
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
ast_typed_tuple,
ast_reference,
ast_local_var_lhs,
ast_local_vars_declaration,
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
ast_int_const,
ast_string_const,
ast_bool_const,
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
ast_null_keyword,
[Tolk] Get rid of ~tilda with `mutate` and `self` methods This is a very big change. If FunC has `.methods()` and `~methods()`, Tolk has only dot, one and only way to call a `.method()`. A method may mutate an object, or may not. It's a behavioral and semantic difference from FunC. - `cs.loadInt(32)` modifies a slice and returns an integer - `b.storeInt(x, 32)` modifies a builder - `b = b.storeInt()` also works, since it not only modifies, but returns - chained methods also work, they return `self` - everything works exactly as expected, similar to JS - no runtime overhead, exactly same Fift instructions - custom methods are created with ease - tilda `~` does not exist in Tolk at all
2024-10-31 07:18:54 +00:00
ast_argument,
ast_argument_list,
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
ast_dot_access,
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
ast_function_call,
ast_underscore,
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
ast_assign,
ast_set_assign,
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
ast_unary_operator,
ast_binary_operator,
ast_ternary_operator,
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
ast_cast_as_operator,
// statements
ast_empty_statement,
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
ast_sequence,
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
ast_return_statement,
ast_if_statement,
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
ast_repeat_statement,
ast_while_statement,
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
ast_do_while_statement,
ast_throw_statement,
ast_assert_statement,
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
ast_try_catch_statement,
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
ast_asm_body,
// other
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
ast_genericsT_item,
ast_genericsT_list,
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
ast_instantiationT_item,
ast_instantiationT_list,
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
ast_parameter,
ast_parameter_list,
ast_annotation,
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
ast_function_declaration,
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
ast_global_var_declaration,
ast_constant_declaration,
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
ast_tolk_required_version,
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
ast_import_directive,
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
ast_tolk_file,
};
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
enum class AnnotationKind {
inline_simple,
inline_ref,
method_id,
pure,
deprecated,
unknown,
};
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
template<ASTNodeType node_type>
struct Vertex;
template<ASTNodeType node_type>
using V = const Vertex<node_type>*;
#define createV new Vertex
struct UnexpectedASTNodeType final : std::exception {
AnyV v_unexpected;
std::string message;
explicit UnexpectedASTNodeType(AnyV v_unexpected, const char* place_where);
const char* what() const noexcept override {
return message.c_str();
}
};
// ---------------------------------------------------------
struct ASTNodeBase {
const ASTNodeType type;
const SrcLocation loc;
ASTNodeBase(ASTNodeType type, SrcLocation loc) : type(type), loc(loc) {}
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
ASTNodeBase(const ASTNodeBase&) = delete;
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
template<ASTNodeType node_type>
V<node_type> as() const {
#ifdef TOLK_DEBUG
if (type != node_type) {
throw Fatal("v->as<...> to wrong node_type");
}
#endif
return static_cast<V<node_type>>(this);
}
template<ASTNodeType node_type>
V<node_type> try_as() const {
return type == node_type ? static_cast<V<node_type>>(this) : nullptr;
}
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
#ifdef TOLK_DEBUG
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
std::string to_debug_string() const { return to_debug_string(false); }
std::string to_debug_string(bool colored) const;
void debug_print() const;
#endif
GNU_ATTRIBUTE_NORETURN GNU_ATTRIBUTE_COLD
void error(const std::string& err_msg) const;
};
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct ASTNodeExpressionBase : ASTNodeBase {
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
friend class ASTDuplicatorFunction;
TypePtr inferred_type = nullptr;
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
bool is_rvalue: 1 = false;
bool is_lvalue: 1 = false;
ASTNodeExpressionBase* mutate() const { return const_cast<ASTNodeExpressionBase*>(this); }
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
void assign_inferred_type(TypePtr type);
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
void assign_rvalue_true();
void assign_lvalue_true();
ASTNodeExpressionBase(ASTNodeType type, SrcLocation loc) : ASTNodeBase(type, loc) {}
};
struct ASTNodeStatementBase : ASTNodeBase {
ASTNodeStatementBase(ASTNodeType type, SrcLocation loc) : ASTNodeBase(type, loc) {}
};
struct ASTExprLeaf : ASTNodeExpressionBase {
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
friend class ASTVisitor;
friend class ASTReplacer;
protected:
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
ASTExprLeaf(ASTNodeType type, SrcLocation loc)
: ASTNodeExpressionBase(type, loc) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct ASTExprUnary : ASTNodeExpressionBase {
friend class ASTVisitor;
friend class ASTReplacer;
protected:
AnyExprV child;
ASTExprUnary(ASTNodeType type, SrcLocation loc, AnyExprV child)
: ASTNodeExpressionBase(type, loc), child(child) {}
};
struct ASTExprBinary : ASTNodeExpressionBase {
friend class ASTVisitor;
friend class ASTReplacer;
protected:
AnyExprV lhs;
AnyExprV rhs;
ASTExprBinary(ASTNodeType type, SrcLocation loc, AnyExprV lhs, AnyExprV rhs)
: ASTNodeExpressionBase(type, loc), lhs(lhs), rhs(rhs) {}
};
struct ASTExprVararg : ASTNodeExpressionBase {
friend class ASTVisitor;
friend class ASTReplacer;
protected:
std::vector<AnyExprV> children;
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
AnyExprV child(int i) const { return children.at(i); }
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
ASTExprVararg(ASTNodeType type, SrcLocation loc, std::vector<AnyExprV> children)
: ASTNodeExpressionBase(type, loc), children(std::move(children)) {}
public:
int size() const { return static_cast<int>(children.size()); }
bool empty() const { return children.empty(); }
};
struct ASTStatementUnary : ASTNodeStatementBase {
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
friend class ASTVisitor;
friend class ASTReplacer;
protected:
AnyV child;
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
AnyExprV child_as_expr() const { return reinterpret_cast<AnyExprV>(child); }
ASTStatementUnary(ASTNodeType type, SrcLocation loc, AnyV child)
: ASTNodeStatementBase(type, loc), child(child) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct ASTStatementVararg : ASTNodeStatementBase {
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
friend class ASTVisitor;
friend class ASTReplacer;
protected:
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
std::vector<AnyV> children;
AnyExprV child_as_expr(int i) const { return reinterpret_cast<AnyExprV>(children.at(i)); }
ASTStatementVararg(ASTNodeType type, SrcLocation loc, std::vector<AnyV> children)
: ASTNodeStatementBase(type, loc), children(std::move(children)) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
public:
int size() const { return static_cast<int>(children.size()); }
bool empty() const { return children.empty(); }
};
struct ASTOtherLeaf : ASTNodeBase {
friend class ASTVisitor;
friend class ASTReplacer;
protected:
ASTOtherLeaf(ASTNodeType type, SrcLocation loc)
: ASTNodeBase(type, loc) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct ASTOtherVararg : ASTNodeBase {
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
friend class ASTVisitor;
friend class ASTReplacer;
protected:
std::vector<AnyV> children;
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
AnyExprV child_as_expr(int i) const { return reinterpret_cast<AnyExprV>(children.at(i)); }
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
ASTOtherVararg(ASTNodeType type, SrcLocation loc, std::vector<AnyV> children)
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
: ASTNodeBase(type, loc), children(std::move(children)) {}
public:
int size() const { return static_cast<int>(children.size()); }
bool empty() const { return children.empty(); }
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_identifier is "a name" in AST structure
// it's NOT a standalone expression, it's "implementation details" of other AST vertices
// example: `var x = 5` then "x" is identifier (inside local var declaration)
// example: `global g: int` then "g" is identifier
// example: `someF` is a reference, which contains identifier
// example: `someF<int>` is a reference which contains identifier and generics instantiation
// example: `fun f<T>()` then "f" is identifier, "<T>" is a generics declaration
struct Vertex<ast_identifier> final : ASTOtherLeaf {
std::string_view name; // empty for underscore
Vertex(SrcLocation loc, std::string_view name)
: ASTOtherLeaf(ast_identifier, loc)
, name(name) {}
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
};
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
//
// ---------------------------------------------------------
// expressions
//
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_empty_expression is "nothing" in context of expression, it has "unknown" type
// example: `throw 123;` then "throw arg" is empty expression (opposed to `throw (123, arg)`)
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_empty_expression> final : ASTExprLeaf {
[Tolk] Compilation pipeline, register global symbols in advance Since I've implemented AST, now I can drop forward declarations. Instead, I traverse AST of all files and register global symbols (functions, constants, global vars) as a separate step, in advance. That's why, while converting AST to Expr/Op, all available symbols are already registered. This greatly simplifies "intermediate state" of yet unknown functions and checking them afterward. Redeclaration of local variables (inside the same scope) is now also prohibited.
2024-10-31 07:04:58 +00:00
explicit Vertex(SrcLocation loc)
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
: ASTExprLeaf(ast_empty_expression, loc) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_parenthesized_expression is something surrounded embraced by (parenthesis)
// example: `(1)`, `((f()))` (two nested)
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_parenthesized_expression> final : ASTExprUnary {
AnyExprV get_expr() const { return child; }
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex(SrcLocation loc, AnyExprV expr)
: ASTExprUnary(ast_parenthesized_expression, loc, expr) {}
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_tensor is a set of expressions embraced by (parenthesis)
// in most languages, it's called "tuple", but in TVM, "tuple" is a TVM primitive, that's why "tensor"
// example: `(1, 2)`, `(1, (2, 3))` (nested), `()` (empty tensor)
// note, that `(1)` is not a tensor, it's a parenthesized expression
// a tensor of N elements occupies N slots on a stack (opposed to TVM tuple primitive, 1 slot)
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_tensor> final : ASTExprVararg {
const std::vector<AnyExprV>& get_items() const { return children; }
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
AnyExprV get_item(int i) const { return child(i); }
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex(SrcLocation loc, std::vector<AnyExprV> items)
: ASTExprVararg(ast_tensor, loc, std::move(items)) {}
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_typed_tuple is a set of expressions in [square brackets]
// in TVM, it's a TVM tuple, that occupies 1 slot, but the compiler knows its "typed structure"
// example: `[1, x]`, `[[0]]` (nested)
// typed tuples can be assigned to N variables, like `[one, _, three] = [1,2,3]`
struct Vertex<ast_typed_tuple> final : ASTExprVararg {
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
const std::vector<AnyExprV>& get_items() const { return children; }
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
AnyExprV get_item(int i) const { return child(i); }
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex(SrcLocation loc, std::vector<AnyExprV> items)
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
: ASTExprVararg(ast_typed_tuple, loc, std::move(items)) {}
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
};
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_reference is "something that references a symbol"
// examples: `x` / `someF` / `someF<int>`
// it's a leaf expression from traversing point of view, but actually, has children (not expressions)
// note, that both `someF()` and `someF<int>()` are function calls, where a callee is just a reference
struct Vertex<ast_reference> final : ASTExprLeaf {
private:
V<ast_identifier> identifier; // its name, `x` / `someF`
V<ast_instantiationT_list> instantiationTs; // not null if `<int>`, otherwise nullptr
public:
const Symbol* sym = nullptr; // filled on resolve or type inferring; points to local / global / function / constant
auto get_identifier() const { return identifier; }
bool has_instantiationTs() const { return instantiationTs != nullptr; }
auto get_instantiationTs() const { return instantiationTs; }
std::string_view get_name() const { return identifier->name; }
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex* mutate() const { return const_cast<Vertex*>(this); }
void assign_sym(const Symbol* sym);
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
Vertex(SrcLocation loc, V<ast_identifier> name_identifier, V<ast_instantiationT_list> instantiationTs)
: ASTExprLeaf(ast_reference, loc)
, identifier(name_identifier), instantiationTs(instantiationTs) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_local_var_lhs is one variable inside `var` declaration
// example: `var x = 0;` then "x" is local var lhs
// example: `val (x: int, [y redef], _) = rhs` then "x" and "y" and "_" are
// it's a leaf from expression's point of view, though technically has an "identifier" child
struct Vertex<ast_local_var_lhs> final : ASTExprLeaf {
private:
V<ast_identifier> identifier;
public:
const LocalVarData* var_ref = nullptr; // filled on resolve identifiers; for `redef` points to declared above; for underscore, name is empty
TypePtr declared_type; // not null for `var x: int = rhs`, otherwise nullptr
bool is_immutable; // declared via 'val', not 'var'
bool marked_as_redef; // var (existing_var redef, new_var: int) = ...
V<ast_identifier> get_identifier() const { return identifier; }
std::string_view get_name() const { return identifier->name; } // empty for underscore
Vertex* mutate() const { return const_cast<Vertex*>(this); }
void assign_var_ref(const LocalVarData* var_ref);
void assign_resolved_type(TypePtr declared_type);
Vertex(SrcLocation loc, V<ast_identifier> identifier, TypePtr declared_type, bool is_immutable, bool marked_as_redef)
: ASTExprLeaf(ast_local_var_lhs, loc)
, identifier(identifier), declared_type(declared_type), is_immutable(is_immutable), marked_as_redef(marked_as_redef) {}
};
template<>
// ast_local_vars_declaration is an expression declaring local variables on the left side of assignment
// examples: see above
// for `var (x, [y])` its expr is "tensor (local var, typed tuple (local var))"
// for assignment `var x = 5`, this node is `var x`, lhs of assignment
struct Vertex<ast_local_vars_declaration> final : ASTExprUnary {
AnyExprV get_expr() const { return child; } // ast_local_var_lhs / ast_tensor / ast_typed_tuple
Vertex(SrcLocation loc, AnyExprV expr)
: ASTExprUnary(ast_local_vars_declaration, loc, expr) {}
};
template<>
// ast_int_const is an integer literal
// examples: `0` / `0xFF`
// note, that `-1` is unary minus of `1` int const
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_int_const> final : ASTExprLeaf {
td::RefInt256 intval; // parsed value, 255 for "0xFF"
std::string_view orig_str; // original "0xFF"; empty for nodes generated by compiler (e.g. in constant folding)
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex(SrcLocation loc, td::RefInt256 intval, std::string_view orig_str)
: ASTExprLeaf(ast_int_const, loc)
, intval(std::move(intval))
, orig_str(orig_str) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_string_const is a string literal in double quotes or """ when multiline
// examples: "asdf" / "Ef8zMz..."a / "to_calc_crc32_from"c
// an optional modifier specifies how a string is parsed (probably, like an integer)
// note, that TVM doesn't have strings, it has only slices, so "hello" has type slice
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_string_const> final : ASTExprLeaf {
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
std::string_view str_val;
char modifier;
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
bool is_bitslice() const {
char m = modifier;
return m == 0 || m == 's' || m == 'a';
}
bool is_intval() const {
char m = modifier;
return m == 'u' || m == 'h' || m == 'H' || m == 'c';
}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
Vertex(SrcLocation loc, std::string_view str_val, char modifier)
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
: ASTExprLeaf(ast_string_const, loc)
, str_val(str_val), modifier(modifier) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_bool_const is either `true` or `false`
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_bool_const> final : ASTExprLeaf {
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
bool bool_val;
Vertex(SrcLocation loc, bool bool_val)
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
: ASTExprLeaf(ast_bool_const, loc)
, bool_val(bool_val) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_null_keyword is the `null` literal
// it should be handled with care; for instance, `null` takes special place in the type system
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_null_keyword> final : ASTExprLeaf {
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
explicit Vertex(SrcLocation loc)
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
: ASTExprLeaf(ast_null_keyword, loc) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
[Tolk] Get rid of ~tilda with `mutate` and `self` methods This is a very big change. If FunC has `.methods()` and `~methods()`, Tolk has only dot, one and only way to call a `.method()`. A method may mutate an object, or may not. It's a behavioral and semantic difference from FunC. - `cs.loadInt(32)` modifies a slice and returns an integer - `b.storeInt(x, 32)` modifies a builder - `b = b.storeInt()` also works, since it not only modifies, but returns - chained methods also work, they return `self` - everything works exactly as expected, similar to JS - no runtime overhead, exactly same Fift instructions - custom methods are created with ease - tilda `~` does not exist in Tolk at all
2024-10-31 07:18:54 +00:00
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_argument is an element of an argument list of a function/method call
// example: `f(1, x)` has 2 arguments, `t.tupleFirst()` has no arguments (though `t` is passed as `self`)
// example: `f(mutate arg)` has 1 argument with `passed_as_mutate` flag
// (without `mutate` keyword, the entity "argument" could be replaced just by "any expression")
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_argument> final : ASTExprUnary {
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
bool passed_as_mutate;
[Tolk] Get rid of ~tilda with `mutate` and `self` methods This is a very big change. If FunC has `.methods()` and `~methods()`, Tolk has only dot, one and only way to call a `.method()`. A method may mutate an object, or may not. It's a behavioral and semantic difference from FunC. - `cs.loadInt(32)` modifies a slice and returns an integer - `b.storeInt(x, 32)` modifies a builder - `b = b.storeInt()` also works, since it not only modifies, but returns - chained methods also work, they return `self` - everything works exactly as expected, similar to JS - no runtime overhead, exactly same Fift instructions - custom methods are created with ease - tilda `~` does not exist in Tolk at all
2024-10-31 07:18:54 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
AnyExprV get_expr() const { return child; }
[Tolk] Get rid of ~tilda with `mutate` and `self` methods This is a very big change. If FunC has `.methods()` and `~methods()`, Tolk has only dot, one and only way to call a `.method()`. A method may mutate an object, or may not. It's a behavioral and semantic difference from FunC. - `cs.loadInt(32)` modifies a slice and returns an integer - `b.storeInt(x, 32)` modifies a builder - `b = b.storeInt()` also works, since it not only modifies, but returns - chained methods also work, they return `self` - everything works exactly as expected, similar to JS - no runtime overhead, exactly same Fift instructions - custom methods are created with ease - tilda `~` does not exist in Tolk at all
2024-10-31 07:18:54 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex(SrcLocation loc, AnyExprV expr, bool passed_as_mutate)
: ASTExprUnary(ast_argument, loc, expr)
, passed_as_mutate(passed_as_mutate) {}
[Tolk] Get rid of ~tilda with `mutate` and `self` methods This is a very big change. If FunC has `.methods()` and `~methods()`, Tolk has only dot, one and only way to call a `.method()`. A method may mutate an object, or may not. It's a behavioral and semantic difference from FunC. - `cs.loadInt(32)` modifies a slice and returns an integer - `b.storeInt(x, 32)` modifies a builder - `b = b.storeInt()` also works, since it not only modifies, but returns - chained methods also work, they return `self` - everything works exactly as expected, similar to JS - no runtime overhead, exactly same Fift instructions - custom methods are created with ease - tilda `~` does not exist in Tolk at all
2024-10-31 07:18:54 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_argument_list contains N arguments of a function/method call
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_argument_list> final : ASTExprVararg {
const std::vector<AnyExprV>& get_arguments() const { return children; }
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
auto get_arg(int i) const { return child(i)->as<ast_argument>(); }
[Tolk] Get rid of ~tilda with `mutate` and `self` methods This is a very big change. If FunC has `.methods()` and `~methods()`, Tolk has only dot, one and only way to call a `.method()`. A method may mutate an object, or may not. It's a behavioral and semantic difference from FunC. - `cs.loadInt(32)` modifies a slice and returns an integer - `b.storeInt(x, 32)` modifies a builder - `b = b.storeInt()` also works, since it not only modifies, but returns - chained methods also work, they return `self` - everything works exactly as expected, similar to JS - no runtime overhead, exactly same Fift instructions - custom methods are created with ease - tilda `~` does not exist in Tolk at all
2024-10-31 07:18:54 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex(SrcLocation loc, std::vector<AnyExprV> arguments)
: ASTExprVararg(ast_argument_list, loc, std::move(arguments)) {}
[Tolk] Get rid of ~tilda with `mutate` and `self` methods This is a very big change. If FunC has `.methods()` and `~methods()`, Tolk has only dot, one and only way to call a `.method()`. A method may mutate an object, or may not. It's a behavioral and semantic difference from FunC. - `cs.loadInt(32)` modifies a slice and returns an integer - `b.storeInt(x, 32)` modifies a builder - `b = b.storeInt()` also works, since it not only modifies, but returns - chained methods also work, they return `self` - everything works exactly as expected, similar to JS - no runtime overhead, exactly same Fift instructions - custom methods are created with ease - tilda `~` does not exist in Tolk at all
2024-10-31 07:18:54 +00:00
};
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_dot_access is "object before dot, identifier + optional <T> after dot"
// examples: `tensorVar.0` / `obj.field` / `getObj().method` / `t.tupleFirst<int>`
// from traversing point of view, it's an unary expression: only obj is expression, field name is not
// note, that `obj.method()` is a function call with "dot access `obj.method`" callee
struct Vertex<ast_dot_access> final : ASTExprUnary {
private:
V<ast_identifier> identifier; // `0` / `field` / `method`
V<ast_instantiationT_list> instantiationTs; // not null if `<int>`, otherwise nullptr
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
public:
[Tolk] Support syntax `tensorVar.0` and `tupleVar.0` It works both for reading and writing: > var t = (1, 2); > t.0; // 1 > t.0 = 5; > t; // (5, 2) It also works for typed/untyped tuples, producing INDEX and SETINDEX. Global tensors and tuples works. Nesting `t.0.1.2` works. `mutate` works. Even mixing tuples inside tensors inside a global for writing works.
2025-01-27 07:29:17 +00:00
typedef std::variant<
const FunctionData*, // for `t.tupleAt` target is `tupleAt` global function
int // for `t.0` target is "indexed access" 0
> DotTarget;
DotTarget target = static_cast<FunctionData*>(nullptr); // filled at type inferring
bool is_target_fun_ref() const { return std::holds_alternative<const FunctionData*>(target); }
bool is_target_indexed_access() const { return std::holds_alternative<int>(target); }
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
AnyExprV get_obj() const { return child; }
auto get_identifier() const { return identifier; }
bool has_instantiationTs() const { return instantiationTs != nullptr; }
auto get_instantiationTs() const { return instantiationTs; }
std::string_view get_field_name() const { return identifier->name; }
[Tolk] Get rid of ~tilda with `mutate` and `self` methods This is a very big change. If FunC has `.methods()` and `~methods()`, Tolk has only dot, one and only way to call a `.method()`. A method may mutate an object, or may not. It's a behavioral and semantic difference from FunC. - `cs.loadInt(32)` modifies a slice and returns an integer - `b.storeInt(x, 32)` modifies a builder - `b = b.storeInt()` also works, since it not only modifies, but returns - chained methods also work, they return `self` - everything works exactly as expected, similar to JS - no runtime overhead, exactly same Fift instructions - custom methods are created with ease - tilda `~` does not exist in Tolk at all
2024-10-31 07:18:54 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex* mutate() const { return const_cast<Vertex*>(this); }
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
void assign_target(const DotTarget& target);
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
Vertex(SrcLocation loc, AnyExprV obj, V<ast_identifier> identifier, V<ast_instantiationT_list> instantiationTs)
: ASTExprUnary(ast_dot_access, loc, obj)
, identifier(identifier), instantiationTs(instantiationTs) {}
[Tolk] Get rid of ~tilda with `mutate` and `self` methods This is a very big change. If FunC has `.methods()` and `~methods()`, Tolk has only dot, one and only way to call a `.method()`. A method may mutate an object, or may not. It's a behavioral and semantic difference from FunC. - `cs.loadInt(32)` modifies a slice and returns an integer - `b.storeInt(x, 32)` modifies a builder - `b = b.storeInt()` also works, since it not only modifies, but returns - chained methods also work, they return `self` - everything works exactly as expected, similar to JS - no runtime overhead, exactly same Fift instructions - custom methods are created with ease - tilda `~` does not exist in Tolk at all
2024-10-31 07:18:54 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_function_call is "calling some lhs with parenthesis", lhs is arbitrary expression (callee)
// example: `globalF()` then callee is reference
// example: `globalF<int>()` then callee is reference (with instantiation Ts filled)
// example: `local_var()` then callee is reference (points to local var, filled at resolve identifiers)
// example: `getF()()` then callee is another func call (which type is TypeDataFunCallable)
// example: `obj.method()` then callee is dot access (resolved while type inferring)
struct Vertex<ast_function_call> final : ASTExprBinary {
const FunctionData* fun_maybe = nullptr; // filled while type inferring for `globalF()` / `obj.f()`; remains nullptr for `local_var()` / `getF()()`
[Tolk] Get rid of ~tilda with `mutate` and `self` methods This is a very big change. If FunC has `.methods()` and `~methods()`, Tolk has only dot, one and only way to call a `.method()`. A method may mutate an object, or may not. It's a behavioral and semantic difference from FunC. - `cs.loadInt(32)` modifies a slice and returns an integer - `b.storeInt(x, 32)` modifies a builder - `b = b.storeInt()` also works, since it not only modifies, but returns - chained methods also work, they return `self` - everything works exactly as expected, similar to JS - no runtime overhead, exactly same Fift instructions - custom methods are created with ease - tilda `~` does not exist in Tolk at all
2024-10-31 07:18:54 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
AnyExprV get_callee() const { return lhs; }
bool is_dot_call() const { return lhs->type == ast_dot_access; }
AnyExprV get_dot_obj() const { return lhs->as<ast_dot_access>()->get_obj(); }
[Tolk] Get rid of ~tilda with `mutate` and `self` methods This is a very big change. If FunC has `.methods()` and `~methods()`, Tolk has only dot, one and only way to call a `.method()`. A method may mutate an object, or may not. It's a behavioral and semantic difference from FunC. - `cs.loadInt(32)` modifies a slice and returns an integer - `b.storeInt(x, 32)` modifies a builder - `b = b.storeInt()` also works, since it not only modifies, but returns - chained methods also work, they return `self` - everything works exactly as expected, similar to JS - no runtime overhead, exactly same Fift instructions - custom methods are created with ease - tilda `~` does not exist in Tolk at all
2024-10-31 07:18:54 +00:00
auto get_arg_list() const { return rhs->as<ast_argument_list>(); }
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
int get_num_args() const { return rhs->as<ast_argument_list>()->size(); }
auto get_arg(int i) const { return rhs->as<ast_argument_list>()->get_arg(i); }
Vertex* mutate() const { return const_cast<Vertex*>(this); }
void assign_fun_ref(const FunctionData* fun_ref);
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
Vertex(SrcLocation loc, AnyExprV lhs_f, V<ast_argument_list> arguments)
: ASTExprBinary(ast_function_call, loc, lhs_f, arguments) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_underscore represents `_` symbol used for left side of assignment
// example: `(cs, _) = cs.loadAndReturn()`
// though it's the only correct usage, using _ as rvalue like `var x = _;` is correct from AST point of view
// note, that for declaration `var _ = 1` underscore is a regular local var declared (with empty name)
// but for `_ = 1` (not declaration) it's underscore; it's because `var _:int` is also correct
struct Vertex<ast_underscore> final : ASTExprLeaf {
explicit Vertex(SrcLocation loc)
: ASTExprLeaf(ast_underscore, loc) {}
};
[Tolk] Compilation pipeline, register global symbols in advance Since I've implemented AST, now I can drop forward declarations. Instead, I traverse AST of all files and register global symbols (functions, constants, global vars) as a separate step, in advance. That's why, while converting AST to Expr/Op, all available symbols are already registered. This greatly simplifies "intermediate state" of yet unknown functions and checking them afterward. Redeclaration of local variables (inside the same scope) is now also prohibited.
2024-10-31 07:04:58 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
template<>
// ast_assign represents assignment "lhs = rhs"
// examples: `a = 4` / `var a = 4` / `(cs, b, mode) = rhs` / `f() = g()`
// note, that `a = 4` lhs is ast_reference, `var a = 4` lhs is ast_local_vars_declaration
struct Vertex<ast_assign> final : ASTExprBinary {
AnyExprV get_lhs() const { return lhs; }
AnyExprV get_rhs() const { return rhs; }
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
explicit Vertex(SrcLocation loc, AnyExprV lhs, AnyExprV rhs)
: ASTExprBinary(ast_assign, loc, lhs, rhs) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_set_assign represents assignment-and-set operation "lhs <op>= rhs"
// examples: `a += 4` / `b <<= c`
struct Vertex<ast_set_assign> final : ASTExprBinary {
const FunctionData* fun_ref = nullptr; // filled at type inferring, points to `_+_` built-in for +=
std::string_view operator_name; // without equal sign, "+" for operator +=
TokenType tok; // tok_set_*
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
AnyExprV get_lhs() const { return lhs; }
AnyExprV get_rhs() const { return rhs; }
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex* mutate() const { return const_cast<Vertex*>(this); }
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
void assign_fun_ref(const FunctionData* fun_ref);
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
Vertex(SrcLocation loc, std::string_view operator_name, TokenType tok, AnyExprV lhs, AnyExprV rhs)
: ASTExprBinary(ast_set_assign, loc, lhs, rhs)
, operator_name(operator_name), tok(tok) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_unary_operator is "some operator over one expression"
// examples: `-1` / `~found`
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_unary_operator> final : ASTExprUnary {
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
const FunctionData* fun_ref = nullptr; // filled at type inferring, points to some built-in function
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
std::string_view operator_name;
TokenType tok;
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
AnyExprV get_rhs() const { return child; }
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
Vertex* mutate() const { return const_cast<Vertex*>(this); }
void assign_fun_ref(const FunctionData* fun_ref);
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex(SrcLocation loc, std::string_view operator_name, TokenType tok, AnyExprV rhs)
: ASTExprUnary(ast_unary_operator, loc, rhs)
, operator_name(operator_name), tok(tok) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_binary_operator is "some operator over two expressions"
// examples: `a + b` / `x & true` / `(a, b) << g()`
// note, that `a = b` is NOT a binary operator, it's ast_assign, also `a += b`, it's ast_set_assign
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_binary_operator> final : ASTExprBinary {
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
const FunctionData* fun_ref = nullptr; // filled at type inferring, points to some built-in function
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
std::string_view operator_name;
TokenType tok;
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
AnyExprV get_lhs() const { return lhs; }
AnyExprV get_rhs() const { return rhs; }
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
Vertex* mutate() const { return const_cast<Vertex*>(this); }
void assign_fun_ref(const FunctionData* fun_ref);
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex(SrcLocation loc, std::string_view operator_name, TokenType tok, AnyExprV lhs, AnyExprV rhs)
: ASTExprBinary(ast_binary_operator, loc, lhs, rhs)
, operator_name(operator_name), tok(tok) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_ternary_operator is a traditional ternary construction
// example: `cond ? a : b`
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_ternary_operator> final : ASTExprVararg {
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
AnyExprV get_cond() const { return child(0); }
AnyExprV get_when_true() const { return child(1); }
AnyExprV get_when_false() const { return child(2); }
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex(SrcLocation loc, AnyExprV cond, AnyExprV when_true, AnyExprV when_false)
: ASTExprVararg(ast_ternary_operator, loc, {cond, when_true, when_false}) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_cast_as_operator is explicit casting with "as" keyword
// examples: `arg as int` / `null as cell` / `t.tupleAt(2) as slice`
struct Vertex<ast_cast_as_operator> final : ASTExprUnary {
AnyExprV get_expr() const { return child; }
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
TypePtr cast_to_type;
Vertex* mutate() const { return const_cast<Vertex*>(this); }
void assign_resolved_type(TypePtr cast_to_type);
Vertex(SrcLocation loc, AnyExprV expr, TypePtr cast_to_type)
: ASTExprUnary(ast_cast_as_operator, loc, expr)
, cast_to_type(cast_to_type) {}
};
//
// ---------------------------------------------------------
// statements
//
template<>
// ast_empty_statement is very similar to "empty sequence" but has a special treatment
// example: `;` (just semicolon)
// example: body of `builtin` function is empty statement (not a zero sequence)
struct Vertex<ast_empty_statement> final : ASTStatementVararg {
explicit Vertex(SrcLocation loc)
: ASTStatementVararg(ast_empty_statement, loc, {}) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_sequence is "some sequence of statements"
// example: function body is a sequence
// example: do while body is a sequence
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_sequence> final : ASTStatementVararg {
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
SrcLocation loc_end;
const std::vector<AnyV>& get_items() const { return children; }
AnyV get_item(int i) const { return children.at(i); }
Vertex(SrcLocation loc, SrcLocation loc_end, std::vector<AnyV> items)
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
: ASTStatementVararg(ast_sequence, loc, std::move(items))
, loc_end(loc_end) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_return_statement is "return something from a function"
// examples: `return a` / `return any_expr()()` / `return;`
// note, that for `return;` (without a value, meaning "void"), in AST, it's stored as empty expression
struct Vertex<ast_return_statement> : ASTStatementUnary {
AnyExprV get_return_value() const { return child_as_expr(); }
bool has_return_value() const { return child->type != ast_empty_expression; }
Vertex(SrcLocation loc, AnyExprV child)
: ASTStatementUnary(ast_return_statement, loc, child) {}
};
template<>
// ast_if_statement is a traditional if statement, probably followed by an else branch
// examples: `if (cond) { ... } else { ... }` / `if (cond) { ... }`
// when else branch is missing, it's stored as empty statement
// for "else if", it's just "if statement" inside a sequence of else branch
struct Vertex<ast_if_statement> final : ASTStatementVararg {
bool is_ifnot; // if(!cond), to generate more optimal fift code
AnyExprV get_cond() const { return child_as_expr(0); }
auto get_if_body() const { return children.at(1)->as<ast_sequence>(); }
auto get_else_body() const { return children.at(2)->as<ast_sequence>(); } // always exists (when else omitted, it's empty)
Vertex(SrcLocation loc, bool is_ifnot, AnyExprV cond, V<ast_sequence> if_body, V<ast_sequence> else_body)
: ASTStatementVararg(ast_if_statement, loc, {cond, if_body, else_body})
, is_ifnot(is_ifnot) {}
};
template<>
// ast_repeat_statement is "repeat something N times"
// example: `repeat (10) { ... }`
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_repeat_statement> final : ASTStatementVararg {
AnyExprV get_cond() const { return child_as_expr(0); }
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
auto get_body() const { return children.at(1)->as<ast_sequence>(); }
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex(SrcLocation loc, AnyExprV cond, V<ast_sequence> body)
: ASTStatementVararg(ast_repeat_statement, loc, {cond, body}) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_while_statement is a standard "while" loop
// example: `while (x > 0) { ... }`
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_while_statement> final : ASTStatementVararg {
AnyExprV get_cond() const { return child_as_expr(0); }
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
auto get_body() const { return children.at(1)->as<ast_sequence>(); }
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex(SrcLocation loc, AnyExprV cond, V<ast_sequence> body)
: ASTStatementVararg(ast_while_statement, loc, {cond, body}) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_do_while_statement is a standard "do while" loop
// example: `do { ... } while (x > 0);`
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_do_while_statement> final : ASTStatementVararg {
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
auto get_body() const { return children.at(0)->as<ast_sequence>(); }
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
AnyExprV get_cond() const { return child_as_expr(1); }
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex(SrcLocation loc, V<ast_sequence> body, AnyExprV cond)
: ASTStatementVararg(ast_do_while_statement, loc, {body, cond}) {}
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_throw_statement is throwing an exception, it accepts excNo and optional arg
// examples: `throw 10` / `throw (ERR_LOW_BALANCE)` / `throw (1001, incomingAddr)`
// when thrown arg is missing, it's stored as empty expression
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_throw_statement> final : ASTStatementVararg {
AnyExprV get_thrown_code() const { return child_as_expr(0); }
bool has_thrown_arg() const { return child_as_expr(1)->type != ast_empty_expression; }
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
AnyExprV get_thrown_arg() const { return child_as_expr(1); }
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex(SrcLocation loc, AnyExprV thrown_code, AnyExprV thrown_arg)
: ASTStatementVararg(ast_throw_statement, loc, {thrown_code, thrown_arg}) {}
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_assert_statement is "assert that cond is true, otherwise throw an exception"
// examples: `assert (balance > 0, ERR_ZERO_BALANCE)` / `assert (balance > 0) throw (ERR_ZERO_BALANCE)`
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_assert_statement> final : ASTStatementVararg {
AnyExprV get_cond() const { return child_as_expr(0); }
AnyExprV get_thrown_code() const { return child_as_expr(1); }
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex(SrcLocation loc, AnyExprV cond, AnyExprV thrown_code)
: ASTStatementVararg(ast_assert_statement, loc, {cond, thrown_code}) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_try_catch_statement is a standard try catch (finally block doesn't exist)
// example: `try { ... } catch (excNo) { ... }`
// there are two formal "arguments" of catch: excNo and arg, but both can be omitted
// when omitted, they are stored as underscores, so len of a catch tensor is always 2
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_try_catch_statement> final : ASTStatementVararg {
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
auto get_try_body() const { return children.at(0)->as<ast_sequence>(); }
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
auto get_catch_expr() const { return children.at(1)->as<ast_tensor>(); } // (excNo, arg), always len 2
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
auto get_catch_body() const { return children.at(2)->as<ast_sequence>(); }
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
Vertex(SrcLocation loc, V<ast_sequence> try_body, V<ast_tensor> catch_expr, V<ast_sequence> catch_body)
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
: ASTStatementVararg(ast_try_catch_statement, loc, {try_body, catch_expr, catch_body}) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_asm_body is a body of `asm` function — a set of strings, and optionally stack order manipulations
// example: `fun skipMessageOp... asm "32 PUSHINT" "SDSKIPFIRST";`
// user can specify "arg order"; example: `fun store(self: builder, op: int) asm (op self)` then [1, 0]
// user can specify "ret order"; example: `fun modDiv... asm(-> 1 0) "DIVMOD";` then [1, 0]
struct Vertex<ast_asm_body> final : ASTStatementVararg {
std::vector<int> arg_order;
std::vector<int> ret_order;
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
const std::vector<AnyV>& get_asm_commands() const { return children; } // ast_string_const[]
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
Vertex(SrcLocation loc, std::vector<int> arg_order, std::vector<int> ret_order, std::vector<AnyV> asm_commands)
: ASTStatementVararg(ast_asm_body, loc, std::move(asm_commands))
, arg_order(std::move(arg_order)), ret_order(std::move(ret_order)) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
//
// ---------------------------------------------------------
// other
//
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_genericsT_item is generics T at declaration
// example: `fun f<T1, T2>` has a list of 2 generic Ts
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_genericsT_item> final : ASTOtherLeaf {
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
std::string_view nameT;
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
Vertex(SrcLocation loc, std::string_view nameT)
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
: ASTOtherLeaf(ast_genericsT_item, loc)
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
, nameT(nameT) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_genericsT_list is a container for generics T at declaration
// example: see above
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_genericsT_list> final : ASTOtherVararg {
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
std::vector<AnyV> get_items() const { return children; }
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
auto get_item(int i) const { return children.at(i)->as<ast_genericsT_item>(); }
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
Vertex(SrcLocation loc, std::vector<AnyV> genericsT_items)
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
: ASTOtherVararg(ast_genericsT_list, loc, std::move(genericsT_items)) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
int lookup_idx(std::string_view nameT) const;
};
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
template<>
// ast_instantiationT_item is manual substitution of generic T used in code, mostly for func calls
// examples: `g<int>()` / `t.tupleFirst<slice>()` / `f<(int, slice), builder>()`
struct Vertex<ast_instantiationT_item> final : ASTOtherLeaf {
TypePtr substituted_type;
Vertex* mutate() const { return const_cast<Vertex*>(this); }
void assign_resolved_type(TypePtr substituted_type);
Vertex(SrcLocation loc, TypePtr substituted_type)
: ASTOtherLeaf(ast_instantiationT_item, loc)
, substituted_type(substituted_type) {}
};
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_instantiationT_list is a container for generic T substitutions used in code
// examples: see above
struct Vertex<ast_instantiationT_list> final : ASTOtherVararg {
std::vector<AnyV> get_items() const { return children; }
auto get_item(int i) const { return children.at(i)->as<ast_instantiationT_item>(); }
Vertex(SrcLocation loc, std::vector<AnyV> instantiationTs)
: ASTOtherVararg(ast_instantiationT_list, loc, std::move(instantiationTs)) {}
};
template<>
// ast_parameter is a parameter of a function in its declaration
// example: `fun f(a: int, mutate b: slice)` has 2 parameters
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_parameter> final : ASTOtherLeaf {
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
const LocalVarData* param_ref = nullptr; // filled on resolve identifiers
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
std::string_view param_name;
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
TypePtr declared_type;
bool declared_as_mutate; // declared as `mutate param_name`
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
bool is_underscore() const { return param_name.empty(); }
Vertex* mutate() const { return const_cast<Vertex*>(this); }
void assign_param_ref(const LocalVarData* param_ref);
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
void assign_resolved_type(TypePtr declared_type);
[Tolk] Compilation pipeline, register global symbols in advance Since I've implemented AST, now I can drop forward declarations. Instead, I traverse AST of all files and register global symbols (functions, constants, global vars) as a separate step, in advance. That's why, while converting AST to Expr/Op, all available symbols are already registered. This greatly simplifies "intermediate state" of yet unknown functions and checking them afterward. Redeclaration of local variables (inside the same scope) is now also prohibited.
2024-10-31 07:04:58 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
Vertex(SrcLocation loc, std::string_view param_name, TypePtr declared_type, bool declared_as_mutate)
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
: ASTOtherLeaf(ast_parameter, loc)
, param_name(param_name), declared_type(declared_type), declared_as_mutate(declared_as_mutate) {}
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_parameter_list is a container of parameters
// example: see above
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_parameter_list> final : ASTOtherVararg {
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
const std::vector<AnyV>& get_params() const { return children; }
auto get_param(int i) const { return children.at(i)->as<ast_parameter>(); }
Vertex(SrcLocation loc, std::vector<AnyV> params)
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
: ASTOtherVararg(ast_parameter_list, loc, std::move(params)) {}
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
int lookup_idx(std::string_view param_name) const;
[Tolk] Get rid of ~tilda with `mutate` and `self` methods This is a very big change. If FunC has `.methods()` and `~methods()`, Tolk has only dot, one and only way to call a `.method()`. A method may mutate an object, or may not. It's a behavioral and semantic difference from FunC. - `cs.loadInt(32)` modifies a slice and returns an integer - `b.storeInt(x, 32)` modifies a builder - `b = b.storeInt()` also works, since it not only modifies, but returns - chained methods also work, they return `self` - everything works exactly as expected, similar to JS - no runtime overhead, exactly same Fift instructions - custom methods are created with ease - tilda `~` does not exist in Tolk at all
2024-10-31 07:18:54 +00:00
int get_mutate_params_count() const;
bool has_mutate_params() const { return get_mutate_params_count() > 0; }
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_annotation is @annotation above a declaration
// example: `@pure fun ...`
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_annotation> final : ASTOtherVararg {
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
AnnotationKind kind;
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
auto get_arg() const { return children.at(0)->as<ast_tensor>(); }
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
static AnnotationKind parse_kind(std::string_view name);
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
Vertex(SrcLocation loc, AnnotationKind kind, V<ast_tensor> arg_probably_empty)
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
: ASTOtherVararg(ast_annotation, loc, {arg_probably_empty})
, kind(kind) {}
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_function_declaration is declaring a function/method
// methods are still global functions, just accepting "self" first parameter
// example: `fun f() { ... }`
// functions can be generic, `fun f<T>(params) { ... }`
// their body is either sequence (regular code function), or `asm`, or `builtin`
struct Vertex<ast_function_declaration> final : ASTOtherVararg {
auto get_identifier() const { return children.at(0)->as<ast_identifier>(); }
int get_num_params() const { return children.at(1)->as<ast_parameter_list>()->size(); }
auto get_param_list() const { return children.at(1)->as<ast_parameter_list>(); }
auto get_param(int i) const { return children.at(1)->as<ast_parameter_list>()->get_param(i); }
AnyV get_body() const { return children.at(2); } // ast_sequence / ast_asm_body
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
const FunctionData* fun_ref = nullptr; // filled after register
TypePtr declared_return_type; // filled at ast parsing; if unspecified (nullptr), means "auto infer"
V<ast_genericsT_list> genericsT_list; // for non-generics it's nullptr
td::RefInt256 method_id; // specified via @method_id annotation
int flags; // from enum in FunctionData
bool is_asm_function() const { return children.at(2)->type == ast_asm_body; }
bool is_code_function() const { return children.at(2)->type == ast_sequence; }
bool is_builtin_function() const { return children.at(2)->type == ast_empty_statement; }
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex* mutate() const { return const_cast<Vertex*>(this); }
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
void assign_fun_ref(const FunctionData* fun_ref);
void assign_resolved_type(TypePtr declared_return_type);
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
Vertex(SrcLocation loc, V<ast_identifier> name_identifier, V<ast_parameter_list> parameters, AnyV body, TypePtr declared_return_type, V<ast_genericsT_list> genericsT_list, td::RefInt256 method_id, int flags)
: ASTOtherVararg(ast_function_declaration, loc, {name_identifier, parameters, body})
, declared_return_type(declared_return_type), genericsT_list(genericsT_list), method_id(std::move(method_id)), flags(flags) {}
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_global_var_declaration is declaring a global var, outside a function
// example: `global g: int;`
// note, that globals don't have default values, since there is no single "entrypoint" for a contract
struct Vertex<ast_global_var_declaration> final : ASTOtherVararg {
const GlobalVarData* var_ref = nullptr; // filled after register
TypePtr declared_type; // filled always, typing globals is mandatory
auto get_identifier() const { return children.at(0)->as<ast_identifier>(); }
[Tolk] v0.6 syntax: `fun`, `import`, `var`, types on the right, etc. Lots of changes, actually. Most noticeable are: - traditional //comments - #include -> import - a rule "import what you use" - ~ found -> !found (for -1/0) - null() -> null - is_null?(v) -> v == null - throw is a keyword - catch with swapped arguments - throw_if, throw_unless -> assert - do until -> do while - elseif -> else if - drop ifnot, elseifnot - drop rarely used operators A testing framework also appears here. All tests existed earlier, but due to significant syntax changes, their history is useless.
2024-10-31 07:11:41 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
Vertex* mutate() const { return const_cast<Vertex*>(this); }
void assign_var_ref(const GlobalVarData* var_ref);
void assign_resolved_type(TypePtr declared_type);
Vertex(SrcLocation loc, V<ast_identifier> name_identifier, TypePtr declared_type)
: ASTOtherVararg(ast_global_var_declaration, loc, {name_identifier})
, declared_type(declared_type) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_constant_declaration is declaring a global constant, outside a function
// example: `const op = 0x123;`
struct Vertex<ast_constant_declaration> final : ASTOtherVararg {
const GlobalConstData* const_ref = nullptr; // filled after register
TypePtr declared_type; // not null for `const op: int = ...`
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
auto get_identifier() const { return children.at(0)->as<ast_identifier>(); }
AnyExprV get_init_value() const { return child_as_expr(1); }
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex* mutate() const { return const_cast<Vertex*>(this); }
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
void assign_const_ref(const GlobalConstData* const_ref);
void assign_resolved_type(TypePtr declared_type);
[Tolk] Compilation pipeline, register global symbols in advance Since I've implemented AST, now I can drop forward declarations. Instead, I traverse AST of all files and register global symbols (functions, constants, global vars) as a separate step, in advance. That's why, while converting AST to Expr/Op, all available symbols are already registered. This greatly simplifies "intermediate state" of yet unknown functions and checking them afterward. Redeclaration of local variables (inside the same scope) is now also prohibited.
2024-10-31 07:04:58 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
Vertex(SrcLocation loc, V<ast_identifier> name_identifier, TypePtr declared_type, AnyExprV init_value)
: ASTOtherVararg(ast_constant_declaration, loc, {name_identifier, init_value})
, declared_type(declared_type) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_tolk_required_version is a preamble fixating compiler's version at the top of the file
// example: `tolk 0.6`
// when compiler version mismatches, it means, that another compiler was earlier for that sources, a warning is emitted
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_tolk_required_version> final : ASTOtherLeaf {
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
std::string_view semver;
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex(SrcLocation loc, std::string_view semver)
: ASTOtherLeaf(ast_tolk_required_version, loc)
, semver(semver) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_import_directive is an import at the top of the file
// examples: `import "another.tolk"` / `import "@stdlib/tvm-dicts"`
struct Vertex<ast_import_directive> final : ASTOtherVararg {
const SrcFile* file = nullptr; // assigned after imports have been resolved, just after parsing a file to ast
[Tolk] Compilation pipeline, register global symbols in advance Since I've implemented AST, now I can drop forward declarations. Instead, I traverse AST of all files and register global symbols (functions, constants, global vars) as a separate step, in advance. That's why, while converting AST to Expr/Op, all available symbols are already registered. This greatly simplifies "intermediate state" of yet unknown functions and checking them afterward. Redeclaration of local variables (inside the same scope) is now also prohibited.
2024-10-31 07:04:58 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
auto get_file_leaf() const { return children.at(0)->as<ast_string_const>(); }
[Tolk] Compilation pipeline, register global symbols in advance Since I've implemented AST, now I can drop forward declarations. Instead, I traverse AST of all files and register global symbols (functions, constants, global vars) as a separate step, in advance. That's why, while converting AST to Expr/Op, all available symbols are already registered. This greatly simplifies "intermediate state" of yet unknown functions and checking them afterward. Redeclaration of local variables (inside the same scope) is now also prohibited.
2024-10-31 07:04:58 +00:00
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
std::string get_file_name() const { return static_cast<std::string>(children.at(0)->as<ast_string_const>()->str_val); }
[Tolk] Compilation pipeline, register global symbols in advance Since I've implemented AST, now I can drop forward declarations. Instead, I traverse AST of all files and register global symbols (functions, constants, global vars) as a separate step, in advance. That's why, while converting AST to Expr/Op, all available symbols are already registered. This greatly simplifies "intermediate state" of yet unknown functions and checking them afterward. Redeclaration of local variables (inside the same scope) is now also prohibited.
2024-10-31 07:04:58 +00:00
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
Vertex* mutate() const { return const_cast<Vertex*>(this); }
void assign_src_file(const SrcFile* file);
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
[Tolk] Compilation pipeline, register global symbols in advance Since I've implemented AST, now I can drop forward declarations. Instead, I traverse AST of all files and register global symbols (functions, constants, global vars) as a separate step, in advance. That's why, while converting AST to Expr/Op, all available symbols are already registered. This greatly simplifies "intermediate state" of yet unknown functions and checking them afterward. Redeclaration of local variables (inside the same scope) is now also prohibited.
2024-10-31 07:04:58 +00:00
Vertex(SrcLocation loc, V<ast_string_const> file_name)
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
: ASTOtherVararg(ast_import_directive, loc, {file_name}) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
template<>
[Tolk] Rewrite the type system from Hindley-Milner to static typing FunC's (and Tolk's before this PR) type system is based on Hindley-Milner. This is a common approach for functional languages, where types are inferred from usage through unification. As a result, type declarations are not necessary: () f(a,b) { return a+b; } // a and b now int, since `+` (int, int) While this approach works for now, problems arise with the introduction of new types like bool, where `!x` must handle both int and bool. It will also become incompatible with int32 and other strict integers. This will clash with structure methods, struggle with proper generics, and become entirely impractical for union types. This PR completely rewrites the type system targeting the future. 1) type of any expression is inferred and never changed 2) this is available because dependent expressions already inferred 3) forall completely removed, generic functions introduced (they work like template functions actually, instantiated while inferring) 4) instantiation `<...>` syntax, example: `t.tupleAt<int>(0)` 5) `as` keyword, for example `t.tupleAt(0) as int` 6) methods binding is done along with type inferring, not before ("before", as worked previously, was always a wrong approach)
2024-12-30 15:31:27 +00:00
// ast_tolk_file represents a whole parsed input .tolk file
// with functions, constants, etc.
// particularly, it contains imports that lead to loading other files
// a whole program consists of multiple parsed files, each of them has a parsed ast tree (stdlib is also parsed)
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
struct Vertex<ast_tolk_file> final : ASTOtherVararg {
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
const SrcFile* const file;
const std::vector<AnyV>& get_toplevel_declarations() const { return children; }
Vertex(const SrcFile* file, std::vector<AnyV> toplevel_declarations)
[Tolk] AST-based semantic analysis, get rid of Expr This is a huge refactoring focusing on untangling compiler internals (previously forked from FunC). The goal is to convert AST directly to Op (a kind of IR representation), doing all code analysis at AST level. Noteable changes: - AST-based semantic kernel includes: registering global symbols, scope handling and resolving local/global identifiers, lvalue/rvalue calc and check, implicit return detection, mutability analysis, pure/impure validity checks, simple constant folding - values of `const` variables are calculated NOT based on CodeBlob, but via a newly-introduced AST-based constant evaluator - AST vertices are now inherited from expression/statement/other; expression vertices have common properties (TypeExpr, lvalue/rvalue) - symbol table is rewritten completely, SymDef/SymVal no longer exist, lexer now doesn't need to register identifiers - AST vertices have references to symbols, filled at different stages of pipeline - the remaining "FunC legacy part" is almost unchanged besides Expr which was fully dropped; AST is converted to Ops (IR) directly
2024-12-16 18:19:45 +00:00
: ASTOtherVararg(ast_tolk_file, SrcLocation(file), std::move(toplevel_declarations))
, file(file) {}
[Tolk] Implement AST: intermediate representation of tolk files Now, the whole .tolk file can be loaded as AST tree and then converted to Expr/Op. This gives a great ability to implement AST transformations. In the future, more and more code analysis will be moved out of legacy to AST-level.
2024-10-31 07:03:33 +00:00
};
} // namespace tolk
Reference in a new issue Copy permalink