feat(typeck): implement constant folding for integer literal type preservation #649
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
onbjerg
added
C-enhancement
This was referenced Dec 17, 2025
DaniPopes
requested changes
Dec 17, 2025
| BitXor => Self::new(self.data ^ r.data), | ||
| Shr => Self::new(self.data.wrapping_shr(r.data.try_into().unwrap_or(usize::MAX))), | ||
| Shl => Self::new(self.data.wrapping_shl(r.data.try_into().unwrap_or(usize::MAX))), | ||
| Sar => Self::new(self.data.arithmetic_shr(r.data.try_into().unwrap_or(usize::MAX))), |
Member
There was a problem hiding this comment.
this is >>> which only works on fixed types, we should rename this to sometihng so we dont forget its useless
| BitOr => Self::new(self.data | r.data), | ||
| BitAnd => Self::new(self.data & r.data), | ||
| BitXor => Self::new(self.data ^ r.data), | ||
| Shr => Self::new(self.data.wrapping_shr(r.data.try_into().unwrap_or(usize::MAX))), |
Member
There was a problem hiding this comment.
this should be arithmetic_shr depending on sign
| } | ||
|
|
||
| /// Returns whether the value is negative. | ||
| pub fn is_negative(&self) -> bool { |
Member
There was a problem hiding this comment.
this should probably be is_signed and be a constant value. then data should be I256 and is_negative will be I256's is_negative (last bit set). this way all ops are easier to keep track of, since they become regular two's complement
DaniPopes
pushed a commit
that referenced
this pull request
Dec 22, 2025
This enables distinguishing edge cases like 127 (fits in int8) from 128 (doesn't fit), which is needed for correct implicit and explicit integer conversions. - Changed internal storage from `u8` (bytes 0-32) to `u16` (bits 0-256) - `TypeSize::new()` now takes bits directly - Added `new_fb_bytes()` for fixed-bytes types that take byte counts - `bits()` returns stored value, `bytes()` computes `ceil(bits/8)` - `mk_ty_int_literal` no longer rounds to multiples of 8 Closes #650 Unblocks #647 #648 #649
onbjerg
force-pushed
the
onbjerg/int-literal-negative
branch
from
addc964 to
e4c688d
Compare
onbjerg
force-pushed
the
onbjerg/int-literal-const-fold
branch
from
d3a3036 to
a1f3e12
Compare
onbjerg
force-pushed
the
onbjerg/int-literal-negative
branch
from
e4c688d to
d621ddf
Compare
onbjerg
force-pushed
the
onbjerg/int-literal-const-fold
branch
from
a1f3e12 to
9ee451e
Compare
onbjerg
added a commit
that referenced
this pull request
Dec 22, 2025
Add implicit type coercion for integer literals (IntLiteral) to typed integers (uint/int). The coercion rules are: - IntLiteral -> uint: Allowed if the literal is non-negative and fits in the target size (size.bits() <= target.bits()) - IntLiteral -> int: Allowed with strict inequality (size.bits() < target.bits()) for non-negative values due to TypeSize rounding, and non-strict for negative values TypeSize stores ceil(bit_len/8), so int_literal[1] covers 0-255. This means we can't distinguish edge cases like 127 (fits in int8) from 128 (doesn't fit), so we conservatively require int16+ for int_literal[1]. Note: Negative literal support requires additional work in the type checker to propagate negativity through unary negation. This is provided in a follow up. Supercedes #564 and closes #627 (closes #564) Stack: - #647 (this) - #648 - #649
onbjerg
force-pushed
the
onbjerg/int-literal-negative
branch
from
d621ddf to
6860eb9
Compare
onbjerg
force-pushed
the
onbjerg/int-literal-const-fold
branch
from
9ee451e to
44ebc93
Compare
onbjerg
added a commit
that referenced
this pull request
Dec 22, 2025
Implement proper handling of negative integer literals in the type checker. Previously, negative literals like -42 were broken because: 1. The parser represents them as unary negation applied to a positive literal, so type_of_lit only sees the positive value 2. The negativity flag in IntLiteral was always false This commit fixes the issue by: - Allowing unary negation on IntLiteral types (they can always be negated since the result is just a negative literal) - Propagating the negativity flag when applying unary negation to an IntLiteral, flipping neg from false to true - Not propagating the expected type through negation when targeting signed types, to avoid premature type mismatch errors on the inner expression Also simplifies the coercion rule to use strict inequality for both positive and negative literals, since TypeSize rounding means we can't reliably distinguish edge cases in either direction. On top of #647 Supercedes #566 (will mark @mablr as a co-author to commend effort) and closes #560 (closes #566) Stack: - #647 - #648 (this) - #649 Co-Authored-By: Mablr <[email protected]>
Implement proper handling of negative integer literals in the type checker. Previously, negative literals like -42 were broken because: 1. The parser represents them as unary negation applied to a positive literal, so type_of_lit only sees the positive value 2. The negativity flag in IntLiteral was always false This commit fixes the issue by: - Allowing unary negation on IntLiteral types (they can always be negated since the result is just a negative literal) - Propagating the negativity flag when applying unary negation to an IntLiteral, flipping neg from false to true - Not propagating the expected type through negation when targeting signed types, to avoid premature type mismatch errors on the inner expression Also simplifies the coercion rule to use strict inequality for both positive and negative literals, since TypeSize rounding means we can't reliably distinguish edge cases in either direction.
…servation When both operands of a binary operation are IntLiteral, evaluate the expression at compile time to preserve literal type semantics. This allows expressions like `-(1 + 2)` to work correctly, since the result remains an IntLiteral that can be negated. Changes: - Extended IntScalar with explicit sign tracking and signed arithmetic - Added try_eval_int_literal_binop to evaluate literal binary expressions - Updated eval_array_len to reject negative array lengths
onbjerg
force-pushed
the
onbjerg/int-literal-const-fold
branch
from
44ebc93 to
522c0bb
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
When both operands of a binary operation are IntLiteral, evaluate the expression at compile time to preserve literal type semantics. This allows expressions like
-(1 + 2)to work correctly, since the result remains an IntLiteral that can be negated.Changes:
On top of #648
Stack: