[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
|
|
|
/*
|
|
|
|
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/>.
|
|
|
|
*/
|
|
|
|
#include "generics-helpers.h"
|
|
|
|
#include "tolk.h"
|
|
|
|
#include "ast.h"
|
|
|
|
#include "ast-replicator.h"
|
|
|
|
#include "type-system.h"
|
|
|
|
#include "compiler-state.h"
|
|
|
|
#include "pipeline.h"
|
|
|
|
|
|
|
|
namespace tolk {
|
|
|
|
|
|
|
|
// given orig = "(int, T)" and substitutions = [slice], return "(int, slice)"
|
|
|
|
static TypePtr replace_genericT_with_deduced(TypePtr orig, const GenericsDeclaration* genericTs, const std::vector<TypePtr>& substitutionTs) {
|
|
|
|
if (!orig || !orig->has_genericT_inside()) {
|
|
|
|
return orig;
|
|
|
|
}
|
|
|
|
tolk_assert(genericTs->size() == substitutionTs.size());
|
|
|
|
|
|
|
|
return orig->replace_children_custom([genericTs, substitutionTs](TypePtr child) {
|
|
|
|
if (const TypeDataGenericT* asT = child->try_as<TypeDataGenericT>()) {
|
|
|
|
int idx = genericTs->find_nameT(asT->nameT);
|
|
|
|
if (idx == -1) {
|
|
|
|
throw Fatal("can not replace generic " + asT->nameT);
|
|
|
|
}
|
2025-02-11 18:49:26 +00:00
|
|
|
if (substitutionTs[idx] == nullptr) {
|
|
|
|
throw GenericDeduceError("can not deduce " + asT->nameT);
|
|
|
|
}
|
[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
|
|
|
return substitutionTs[idx];
|
|
|
|
}
|
|
|
|
return child;
|
|
|
|
});
|
|
|
|
}
|
|
|
|
|
2025-02-11 18:49:26 +00:00
|
|
|
GenericSubstitutionsDeduceForCall::GenericSubstitutionsDeduceForCall(const FunctionData* fun_ref)
|
|
|
|
: fun_ref(fun_ref) {
|
|
|
|
substitutionTs.resize(fun_ref->genericTs->size()); // filled with nullptr (nothing deduced)
|
|
|
|
}
|
|
|
|
|
|
|
|
void GenericSubstitutionsDeduceForCall::provide_deducedT(const std::string& nameT, TypePtr deduced) {
|
|
|
|
if (deduced == TypeDataNullLiteral::create() || deduced->has_unknown_inside()) {
|
|
|
|
return; // just 'null' doesn't give sensible info
|
|
|
|
}
|
|
|
|
|
|
|
|
int idx = fun_ref->genericTs->find_nameT(nameT);
|
|
|
|
if (substitutionTs[idx] == nullptr) {
|
|
|
|
substitutionTs[idx] = deduced;
|
|
|
|
} else if (substitutionTs[idx] != deduced) {
|
|
|
|
throw GenericDeduceError(nameT + " is both " + substitutionTs[idx]->as_human_readable() + " and " + deduced->as_human_readable());
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void GenericSubstitutionsDeduceForCall::provide_manually_specified(std::vector<TypePtr>&& substitutionTs) {
|
|
|
|
this->substitutionTs = std::move(substitutionTs);
|
|
|
|
this->manually_specified = true;
|
|
|
|
}
|
|
|
|
|
[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
|
|
|
// purpose: having `f<T>(value: T)` and call `f(5)`, deduce T = int
|
|
|
|
// generally, there may be many generic Ts for declaration, and many arguments
|
|
|
|
// for every argument, `consider_next_condition()` is called
|
|
|
|
// example: `f<T1, T2>(a: int, b: T1, c: (T1, T2))` and call `f(6, 7, (8, cs))`
|
|
|
|
// - `a` does not affect, it doesn't depend on generic Ts
|
|
|
|
// - next condition: param_type = `T1`, arg_type = `int`, deduce T1 = int
|
|
|
|
// - next condition: param_type = `(T1, T2)`, arg_type = `(int, slice)`, deduce T1 = int, T2 = slice
|
|
|
|
// for call `f(6, cs, (8, cs))` T1 will be both `slice` and `int`, fired an error
|
2025-02-11 18:49:26 +00:00
|
|
|
void GenericSubstitutionsDeduceForCall::consider_next_condition(TypePtr param_type, TypePtr arg_type) {
|
|
|
|
if (const auto* asT = param_type->try_as<TypeDataGenericT>()) {
|
|
|
|
// `(arg: T)` called as `f([1, 2])` => T is [int, int]
|
|
|
|
provide_deducedT(asT->nameT, arg_type);
|
|
|
|
} else if (const auto* p_nullable = param_type->try_as<TypeDataNullable>()) {
|
|
|
|
// `arg: T?` called as `f(nullableInt)` => T is int
|
|
|
|
if (const auto* a_nullable = arg_type->try_as<TypeDataNullable>()) {
|
|
|
|
consider_next_condition(p_nullable->inner, a_nullable->inner);
|
[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
|
|
|
}
|
2025-02-11 18:49:26 +00:00
|
|
|
// `arg: T?` called as `f(int)` => T is int
|
|
|
|
else {
|
|
|
|
consider_next_condition(p_nullable->inner, arg_type);
|
[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
|
|
|
}
|
2025-02-11 18:49:26 +00:00
|
|
|
} else if (const auto* p_tensor = param_type->try_as<TypeDataTensor>()) {
|
|
|
|
// `arg: (int, T)` called as `f((5, cs))` => T is slice
|
|
|
|
if (const auto* a_tensor = arg_type->try_as<TypeDataTensor>(); a_tensor && a_tensor->size() == p_tensor->size()) {
|
|
|
|
for (int i = 0; i < a_tensor->size(); ++i) {
|
|
|
|
consider_next_condition(p_tensor->items[i], a_tensor->items[i]);
|
[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
|
|
|
}
|
2025-02-11 18:49:26 +00:00
|
|
|
}
|
|
|
|
} else if (const auto* p_tuple = param_type->try_as<TypeDataTypedTuple>()) {
|
|
|
|
// `arg: [int, T]` called as `f([5, cs])` => T is slice
|
|
|
|
if (const auto* a_tuple = arg_type->try_as<TypeDataTypedTuple>(); a_tuple && a_tuple->size() == p_tuple->size()) {
|
|
|
|
for (int i = 0; i < a_tuple->size(); ++i) {
|
|
|
|
consider_next_condition(p_tuple->items[i], a_tuple->items[i]);
|
[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
|
|
|
}
|
2025-02-11 18:49:26 +00:00
|
|
|
}
|
|
|
|
} else if (const auto* p_callable = param_type->try_as<TypeDataFunCallable>()) {
|
|
|
|
// `arg: fun(TArg) -> TResult` called as `f(calcTupleLen)` => TArg is tuple, TResult is int
|
|
|
|
if (const auto* a_callable = arg_type->try_as<TypeDataFunCallable>(); a_callable && a_callable->params_size() == p_callable->params_size()) {
|
|
|
|
for (int i = 0; i < a_callable->params_size(); ++i) {
|
|
|
|
consider_next_condition(p_callable->params_types[i], a_callable->params_types[i]);
|
[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
|
|
|
}
|
2025-02-11 18:49:26 +00:00
|
|
|
consider_next_condition(p_callable->return_type, a_callable->return_type);
|
[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
|
|
|
}
|
|
|
|
}
|
2025-02-11 18:49:26 +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
|
|
|
|
2025-02-11 18:49:26 +00:00
|
|
|
TypePtr GenericSubstitutionsDeduceForCall::replace_by_manually_specified(TypePtr param_type) const {
|
|
|
|
return replace_genericT_with_deduced(param_type, fun_ref->genericTs, substitutionTs);
|
|
|
|
}
|
|
|
|
|
|
|
|
TypePtr GenericSubstitutionsDeduceForCall::auto_deduce_from_argument(SrcLocation loc, TypePtr param_type, TypePtr arg_type) {
|
|
|
|
try {
|
|
|
|
if (!manually_specified) {
|
|
|
|
consider_next_condition(param_type, arg_type);
|
[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
|
|
|
}
|
2025-02-11 18:49:26 +00:00
|
|
|
return replace_genericT_with_deduced(param_type, fun_ref->genericTs, substitutionTs);
|
|
|
|
} catch (const GenericDeduceError& ex) {
|
|
|
|
throw ParseError(loc, ex.message + " for generic function `" + fun_ref->as_human_readable() + "`");
|
[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
|
|
|
}
|
2025-02-11 18:49:26 +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
|
|
|
|
2025-02-11 18:49:26 +00:00
|
|
|
int GenericSubstitutionsDeduceForCall::get_first_not_deduced_idx() const {
|
|
|
|
for (int i = 0; i < static_cast<int>(substitutionTs.size()); ++i) {
|
|
|
|
if (substitutionTs[i] == nullptr) {
|
|
|
|
return i;
|
|
|
|
}
|
[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
|
|
|
}
|
2025-02-11 18:49:26 +00:00
|
|
|
return -1;
|
|
|
|
}
|
[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
|
|
|
|
|
|
|
// clone the body of `f<T>` replacing T everywhere with a substitution
|
|
|
|
// before: `fun f<T>(v: T) { var cp: [T] = [v]; }`
|
|
|
|
// after: `fun f<int>(v: int) { var cp: [int] = [v]; }`
|
|
|
|
// an instantiated function becomes a deep copy, all AST nodes are copied, no previous pointers left
|
|
|
|
class GenericFunctionReplicator final : public ASTReplicatorFunction {
|
|
|
|
const GenericsDeclaration* genericTs;
|
|
|
|
const std::vector<TypePtr>& substitutionTs;
|
|
|
|
|
|
|
|
protected:
|
|
|
|
using ASTReplicatorFunction::clone;
|
|
|
|
|
|
|
|
TypePtr clone(TypePtr t) override {
|
|
|
|
return replace_genericT_with_deduced(t, genericTs, substitutionTs);
|
|
|
|
}
|
|
|
|
|
|
|
|
public:
|
|
|
|
GenericFunctionReplicator(const GenericsDeclaration* genericTs, const std::vector<TypePtr>& substitutionTs)
|
|
|
|
: genericTs(genericTs)
|
|
|
|
, substitutionTs(substitutionTs) {
|
|
|
|
}
|
|
|
|
|
|
|
|
V<ast_function_declaration> clone_function_body(V<ast_function_declaration> v_function) override {
|
|
|
|
return createV<ast_function_declaration>(
|
|
|
|
v_function->loc,
|
|
|
|
clone(v_function->get_identifier()),
|
|
|
|
clone(v_function->get_param_list()),
|
|
|
|
clone(v_function->get_body()),
|
|
|
|
clone(v_function->declared_return_type),
|
|
|
|
nullptr, // a newly-created function is not generic
|
|
|
|
v_function->method_id,
|
|
|
|
v_function->flags
|
|
|
|
);
|
|
|
|
}
|
|
|
|
};
|
|
|
|
|
|
|
|
std::string GenericsDeclaration::as_human_readable() const {
|
|
|
|
std::string result = "<";
|
|
|
|
for (const GenericsItem& item : itemsT) {
|
|
|
|
if (result.size() > 1) {
|
|
|
|
result += ",";
|
|
|
|
}
|
|
|
|
result += item.nameT;
|
|
|
|
}
|
|
|
|
result += ">";
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
|
|
|
int GenericsDeclaration::find_nameT(std::string_view nameT) const {
|
|
|
|
for (int i = 0; i < static_cast<int>(itemsT.size()); ++i) {
|
|
|
|
if (itemsT[i].nameT == nameT) {
|
|
|
|
return i;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
// after creating a deep copy of `f<T>` like `f<int>`, its new and fresh body needs the previous pipeline to run
|
|
|
|
// for example, all local vars need to be registered as symbols, etc.
|
|
|
|
static void run_pipeline_for_instantiated_function(const FunctionData* inst_fun_ref) {
|
|
|
|
// these pipes are exactly the same as in tolk.cpp — all preceding (and including) type inferring
|
|
|
|
pipeline_resolve_identifiers_and_assign_symbols(inst_fun_ref);
|
|
|
|
pipeline_calculate_rvalue_lvalue(inst_fun_ref);
|
|
|
|
pipeline_detect_unreachable_statements(inst_fun_ref);
|
|
|
|
pipeline_infer_types_and_calls_and_fields(inst_fun_ref);
|
|
|
|
}
|
|
|
|
|
|
|
|
std::string generate_instantiated_name(const std::string& orig_name, const std::vector<TypePtr>& substitutions) {
|
|
|
|
// an instantiated function name will be "{orig_name}<{T1,T2,...}>"
|
|
|
|
std::string name = orig_name;
|
|
|
|
name += "<";
|
|
|
|
for (TypePtr subs : substitutions) {
|
|
|
|
if (name.size() > orig_name.size() + 1) {
|
|
|
|
name += ",";
|
|
|
|
}
|
|
|
|
name += subs->as_human_readable();
|
|
|
|
}
|
|
|
|
name.erase(std::remove(name.begin(), name.end(), ' '), name.end());
|
|
|
|
name += ">";
|
|
|
|
return name;
|
|
|
|
}
|
|
|
|
|
|
|
|
const FunctionData* instantiate_generic_function(SrcLocation loc, const FunctionData* fun_ref, const std::string& inst_name, std::vector<TypePtr>&& substitutionTs) {
|
|
|
|
tolk_assert(fun_ref->genericTs);
|
|
|
|
|
|
|
|
// if `f<int>` was earlier instantiated, return it
|
|
|
|
if (const auto* existing = lookup_global_symbol(inst_name)) {
|
|
|
|
const FunctionData* inst_ref = existing->try_as<FunctionData>();
|
|
|
|
tolk_assert(inst_ref);
|
|
|
|
return inst_ref;
|
|
|
|
}
|
|
|
|
|
|
|
|
std::vector<LocalVarData> parameters;
|
|
|
|
parameters.reserve(fun_ref->get_num_params());
|
|
|
|
for (const LocalVarData& orig_p : fun_ref->parameters) {
|
2024-12-18 16:26:26 +00:00
|
|
|
parameters.emplace_back(orig_p.name, orig_p.loc, replace_genericT_with_deduced(orig_p.declared_type, fun_ref->genericTs, substitutionTs), orig_p.flags, orig_p.param_idx);
|
[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_return_type = replace_genericT_with_deduced(fun_ref->declared_return_type, fun_ref->genericTs, substitutionTs);
|
|
|
|
const GenericsInstantiation* instantiationTs = new GenericsInstantiation(loc, std::move(substitutionTs));
|
|
|
|
|
|
|
|
if (fun_ref->is_asm_function()) {
|
|
|
|
FunctionData* inst_ref = new FunctionData(inst_name, fun_ref->loc, declared_return_type, std::move(parameters), fun_ref->flags, nullptr, instantiationTs, new FunctionBodyAsm, fun_ref->ast_root);
|
|
|
|
inst_ref->arg_order = fun_ref->arg_order;
|
|
|
|
inst_ref->ret_order = fun_ref->ret_order;
|
|
|
|
G.symtable.add_function(inst_ref);
|
|
|
|
G.all_functions.push_back(inst_ref);
|
|
|
|
run_pipeline_for_instantiated_function(inst_ref);
|
|
|
|
return inst_ref;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (fun_ref->is_builtin_function()) {
|
|
|
|
FunctionData* inst_ref = new FunctionData(inst_name, fun_ref->loc, declared_return_type, std::move(parameters), fun_ref->flags, nullptr, instantiationTs, fun_ref->body, fun_ref->ast_root);
|
|
|
|
inst_ref->arg_order = fun_ref->arg_order;
|
|
|
|
inst_ref->ret_order = fun_ref->ret_order;
|
|
|
|
G.symtable.add_function(inst_ref);
|
|
|
|
return inst_ref;
|
|
|
|
}
|
|
|
|
|
|
|
|
GenericFunctionReplicator replicator(fun_ref->genericTs, instantiationTs->substitutions);
|
|
|
|
V<ast_function_declaration> inst_root = replicator.clone_function_body(fun_ref->ast_root->as<ast_function_declaration>());
|
|
|
|
|
|
|
|
FunctionData* inst_ref = new FunctionData(inst_name, fun_ref->loc, declared_return_type, std::move(parameters), fun_ref->flags, nullptr, instantiationTs, new FunctionBodyCode, inst_root);
|
|
|
|
inst_ref->arg_order = fun_ref->arg_order;
|
|
|
|
inst_ref->ret_order = fun_ref->ret_order;
|
|
|
|
inst_root->mutate()->assign_fun_ref(inst_ref);
|
|
|
|
G.symtable.add_function(inst_ref);
|
|
|
|
G.all_functions.push_back(inst_ref);
|
|
|
|
run_pipeline_for_instantiated_function(inst_ref);
|
|
|
|
return inst_ref;
|
|
|
|
}
|
|
|
|
|
|
|
|
} // namespace tolk
|