When hovering on an attribute name like 'value' in 'instance.value',
the minimal covering node is the Identifier, but we need the Attribute
expression to get the type. Update find_covering_node to track both
the minimal node and minimal expression, preferring the expression.
This fixes hover on attribute accesses. All hover tests now pass.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Add hover.md with examples of hover assertions across different
expression types. Some tests still need arrow alignment fixes, but
many sections are passing including basic literals, function definitions,
comprehensions, and the simple hover test file.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
When hovering, if we find a statement node (like StmtExpr), extract the
expression from within it. This allows hover assertions to work on
standalone expressions like variable references.
Also add a simple working mdtest to demonstrate hover assertions.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Update SortedCheckOutputs to store references to CheckOutput instead
of owned values, matching the design of the previous SortedDiagnostics
implementation. This avoids unnecessary cloning and makes the API more
consistent.
Changes:
- SortedCheckOutputs now stores Vec<&CheckOutput>
- new() takes IntoIterator<Item = &CheckOutput>
- LineCheckOutputs.outputs is now &[&CheckOutput]
- Implement Unmatched and UnmatchedWithColumn for &CheckOutput
- Update match_line to take &[&CheckOutput]
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Elide unnecessary explicit lifetimes in find_covering_node
- Add backticks around CheckOutputs in doc comment
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Update match_line signature to accept &[CheckOutput] instead of
&LineCheckOutputs for consistency with the previous API that took
&[Diagnostic].
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Store the hover assertion column as OneIndexed since that's the type
returned by line_column() and required by SourceLocation. This
eliminates unnecessary conversions between zero-based and one-based
indexing.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Instead of manually calculating character offsets, use line_column()
which does all the work for us:
1. Find the byte offset of the arrow in the comment text
2. Add that to the comment's TextRange to get the arrow's absolute position
3. Call line_column() on that position to get the character column
This is much simpler and lets line_column handle all the UTF-8/UTF-32
conversion complexity.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
The column field was being treated inconsistently - calculated as a
character offset but used as a byte offset. This would break on any
line with multi-byte UTF-8 characters before the hover position.
Fixes:
1. Use chars().position() instead of find() to get character offset
2. Use LineIndex::offset() with PositionEncoding::Utf32 to properly
convert character offset to byte offset (TextSize)
3. Document that column is a UTF-32 character offset
This ensures hover assertions work correctly with Unicode text.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Convert the Hover variant from tuple-style to named fields for better
clarity and self-documentation.
Before: Hover(&'a str, &'a str, TextRange)
After: Hover { expected_type, full_comment, range }
This makes the code more readable and easier to maintain.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Move the column calculation from generate_hover_outputs into
HoverAssertion::from_str() by passing LineIndex and SourceText to
the parse() method.
This simplifies the code and centralizes the column calculation logic
in one place during parsing, rather than spreading it across multiple
locations. The HoverAssertion now directly stores the final column
position in the line, ready to use.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
The previous implementation calculated the down arrow position within
the comment text, but then incorrectly used that as an offset into the
target line. This only worked if the comment started at column 0.
Now we properly calculate the column position by:
1. Finding the arrow offset within the comment text
2. Getting the comment's column position in its line
3. Adding these together to get the arrow's column in the line
4. Using that column to index into the target line
This fixes hover assertions when comments are indented or don't start
at the beginning of the line.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Change the column field in HoverAssertion from OneIndexed to usize,
storing it as a zero-based index. This matches how it's used and
eliminates unnecessary conversions between zero-based and one-based
indexing.
The arrow position from find() is already zero-based, and TextSize
uses zero-based offsets, so this is more natural.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Import InlineFileAssertions, ParsedAssertion, and UnparsedAssertion
at the module level instead of using fully qualified crate::assertion
names throughout the code.
This makes the code cleaner and more idiomatic.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Move the column calculation from generate_hover_outputs to the
HoverAssertion parsing logic. This makes better use of the existing
column field in HoverAssertion.
Changes:
- UnparsedAssertion::Hover now stores both the expected type and the
full comment text
- HoverAssertion::from_str() now takes both parameters and calculates
the column from the down arrow position in the full comment
- generate_hover_outputs() now reads the column from the parsed
assertion instead of recalculating it
This eliminates redundant calculations and makes the column field
actually useful.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Implement UnmatchedWithColumn for CheckOutput and update the column()
method to work with CheckOutput instead of just Diagnostic.
This allows eliminating the match statement when formatting unmatched
outputs - we can now just call unmatched_with_column() on all outputs
uniformly.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Move the HoverOutput type from check_output.rs to hover.rs where it
logically belongs. The check_output module should only contain the
CheckOutput enum and sorting infrastructure, while hover-specific
types belong in the hover module.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Create a dedicated HoverOutput struct to hold hover result data,
replacing the inline fields in CheckOutput::Hover variant.
This allows implementing Unmatched and UnmatchedWithColumn traits
directly on HoverOutput, simplifying the CheckOutput implementations
to simple delegation.
Benefits:
- Better separation of concerns
- Cleaner trait implementations
- More consistent with Diagnostic handling
- Easier to extend HoverOutput in the future
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
SortedDiagnostics is now replaced by SortedCheckOutputs, which handles
both diagnostics and hover results. This refactoring:
- Renames diagnostic.rs to check_output.rs to better reflect its purpose
- Moves CheckOutput and SortedCheckOutputs definitions from matcher.rs
to check_output.rs where they belong
- Removes the now-unused SortedDiagnostics infrastructure
- Ports the test to use SortedCheckOutputs instead of SortedDiagnostics
- Updates all imports throughout the codebase
The check_output module now serves as the central location for sorting
and grouping all types of check outputs (diagnostics and hover results)
by line number.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Replace nested if-let blocks with let-else + continue to reduce
indentation and improve readability.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
The previous implementation incorrectly assumed the target line was
always line_number + 1, which breaks when multiple assertion comments
are stacked on consecutive lines.
Now generate_hover_outputs accepts the parsed InlineFileAssertions,
which already correctly associates each assertion with its target line
number (accounting for stacked comments).
For the column position, we extract it directly from the down arrow
position in the UnparsedAssertion::Hover text, avoiding the need to
parse the assertion ourselves (parsing happens later in the matcher).
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Replace the leave_node() approach with a simpler strategy: just compare
the range lengths and keep the smallest node that covers the offset.
This is more direct and easier to understand than tracking when we
leave nodes. The minimal covering node is simply the one with the
shortest range that contains the offset.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
The previous implementation had a bug: it would overwrite `found` for
every matching node in source order, which could incorrectly select a
sibling node instead of the minimal covering node.
Now use the same approach as ty_ide's covering_node:
- Use leave_node() to detect when we've finished traversing a subtree
- Set a `found` flag when leaving the minimal node to prevent further
updates
- This ensures we return the deepest (most specific) node that covers
the offset, not just the last one visited in source order
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Replace the large match statement over all AnyNodeRef expression variants
with a simple call to as_expr_ref(). This helper method handles all
expression-related variants automatically, reducing the function from
~60 lines to just 6 lines.
Also removes statement-related variants (StmtFunctionDef, StmtClassDef,
StmtExpr) to focus only on expression nodes, which is the primary use
case for hover assertions.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Move hover-related functions from lib.rs into a new hover.rs module
to improve code organization:
- find_covering_node() - locate AST nodes at specific offsets
- infer_type_at_position() - get inferred types using SemanticModel
- generate_hover_outputs() - scan for hover assertions and generate results
This keeps lib.rs focused on the main test execution flow.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
All core hover assertion functionality is now implemented and compiling!
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add find_covering_node() to locate AST nodes at specific positions
- Add infer_type_at_position() to get type information using ty_python_semantic
- Add generate_hover_outputs() to scan for hover assertions and generate results
- Integrate hover outputs into check flow in run_test()
- Implement hover matching logic in matcher.rs to compare types
- Avoid adding ty_ide dependency; instead use ty_python_semantic directly
The implementation extracts hover assertions from comments, computes the target
position from the down arrow location, infers the type at that position using
the AST and SemanticModel, and matches it against the expected type in the
assertion.
Hover assertions now work end-to-end in mdtest files!
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Create CheckOutput enum with Diagnostic and Hover variants
- Replace SortedDiagnostics with SortedCheckOutputs in matcher
- Update match_file to accept &[CheckOutput] instead of &[Diagnostic]
- Update matching logic to handle both diagnostics and hover results
- Implement Unmatched trait for CheckOutput
- Convert diagnostics to CheckOutput in lib.rs before matching
This infrastructure allows hover assertions to be matched against hover
results without polluting the DiagnosticId enum with test-specific IDs.
The hover matching logic will be implemented in the next commit.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add Hover variant to UnparsedAssertion and ParsedAssertion enums
- Create HoverAssertion struct with column and expected_type fields
- Add HoverAssertionParseError enum for validation errors
- Update from_comment() to recognize '# hover:' and '# ↓ hover:' patterns
- Simplified design: down arrow must appear immediately before 'hover' keyword
- Add placeholder matching logic in matcher.rs (to be completed)
- Add PLAN.md to track implementation progress
The hover assertion syntax uses a down arrow to indicate column position:
# ↓ hover: expected_type
expression_to_hover
This will enable testing hover functionality in mdtest files, similar to
how ty_ide tests work with <CURSOR> markers.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
This is still early days, but I hope the framework introduced here makes
it very easy to add new truth data. Truth data should be seen as a form
of regression test for non-ideal ranking of completion suggestions.
I think it would help to read `crates/ty_completion_eval/README.md`
first to get an idea of what you're reviewing.
## Summary
We have the following test in `protocols.md`:
```py
class HasX(Protocol):
x: int
# […]
class Foo:
x: int
# […]
class FooBool(Foo):
x: bool
static_assert(not is_subtype_of(FooBool, HasX))
static_assert(not is_assignable_to(FooBool, HasX))
```
If `Foo` was indeed intended to be a base class of `FooBool`, then `x:
bool` should be reported as a Liskov violation. And then it's a matter
of definition whether or not these assertions should hold true or not
(should the incorrect override take precedence or not?). So it looks to
me like this is just an oversight, probably a copy-paste error from
another test right before it, where `FooSub` is indeed intended to be a
subclass of `Foo`.
I am fixing this because this test started to fail on a branch of mine
that changes how attribute lookup in inheritance chains works.
## Summary
Fixes [astral-sh/ty#1307](https://github.com/astral-sh/ty/issues/1307)
Unions with length <= 5 are unaffected to minimize test churn
Unions with length > 5 will only display the first 3 elements + "...
omitted x union elements"
Here "length" is defined as the number of elements after condensation to
literals
Edit: we no longer truncate in revel case.
Before:
> info: Attempted to call union type `(def f1() -> int) | (def f2(name:
str) -> int) | (def f3(a: int, b: int) -> int) | (def f4[T](x: T@f4) ->
int) | Literal[5] | (Overload[() -> None, (x: str) -> str]) |
(Overload[() -> None, (x: str, y: str) -> str]) | PossiblyNotCallable`
After:
> info: Attempted to call union type `(def f1() -> int) | (def f2(name:
str) -> int) | (def f3(a: int, b: int) -> int) | ... omitted 5 union
elements`
The below comparisons are outdated, but left here as a reference.
Before:
```reveal_type(x) # revealed: Literal[1, 2] | A | B | C | D | E | F | G```
```reveal_type(x) # revealed: Result1A | Result1B | Result2A | Result2B
| Result3 | Result4```
After:
```reveal_type(x) # revealed: Literal[1, 2] | A | B | ... omitted 5 union elements```
```reveal_type(x) # revealed: Result1A | Result1B | Result2A | ...
omitted 3 union elements```
This formatting is consistent with
`crates/ty_python_semantic/src/types/call/bind.rs` line 2992
## Test Plan
Cosmetic only, covered and verified by changes in mdtest
## Summary
Bump the latest supported Python version of ty to 3.14 and updates some
references from 3.13 to 3.14.
This also fixes a bug with `dataclasses.field` on 3.14 (which adds a new
keyword-only parameter to that function, breaking our previously naive
matching on the parameter structure of that function).
## Test Plan
A `ty check` on a file with template strings (without any further
configuration) doesn't raise errors anymore.
## Summary
Typevar attributes (bound/constraints/default) can be either lazily
evaluated or eagerly evaluated. Currently they are lazily evaluated for
PEP 695 typevars, and eager for legacy and synthetic typevars.
https://github.com/astral-sh/ruff/pull/20598 will make them lazy also
for legacy typevars, and the ecosystem report on that PR surfaced the
issue fixed here (because legacy typevars are much more common in the
ecosystem than PEP 695 typevars.)
Applying a transform to a typevar (normalization, materialization, or
mark-inferable) will reify all lazy attributes and create a new typevar
with eager attributes. In terms of Salsa identity, this transformed
typevar will be considered different from the original typevar, whether
or not the attributes were actually transformed.
In general, this is not a problem, since all typevars in a given generic
context will be transformed, or not, together.
The exception to this was implicit-self vs explicit Self annotations.
The typevar we created for implicit self was created initially using
inferable typevars, whereas an explicit Self annotation is initially
non-inferable, then transformed via mark-inferable when accessed as part
of a function signature. If the containing class (which becomes the
upper bound of `Self`) is generic, and has e.g. a lazily-evaluated
default, then the explicit-Self annotation will reify that default in
the upper bound, and the implicit-self would not, leading them to be
treated as different typevars, and causing us to fail to solve a call to
a method such as `def method(self) -> Self` correctly.
The fix here is to treat implicit-self more like explicit-Self,
initially creating it as non-inferable and then using the mark-inferable
transform on it. This is less efficient, but restores the invariant that
all typevars in a given generic context are transformed together, or
not, fixing the bug.
In the improved-constraint-solver work, the separation of typevars into
"inferable" and "non-inferable" is expected to disappear, along with the
mark-inferable transform, which would render both this bug and the fix
moot. So this fix is really just temporary until that lands.
There is a performance regression, but not a huge one: 1-2% on most
projects, 5% on one outlier. This seems acceptable, given that it should
be fully recovered by removing the mark-inferable transform.
## Test Plan
Added mdtests that failed before this change.
Resolves https://github.com/astral-sh/ruff/issues/20694
This PR updates the `zip_without_explicit_strict` and
`map_without_explicit_strict` rules so their fixes are always marked
unsafe, following Brent's guidance that adding `strict=False` can
silently preserve buggy behaviour when inputs differ. The fix safety
docs now spell out that reasoning, the applicability drops to `Unsafe`,
and the snapshots were refreshed so Ruff clearly warns users before
applying the edit.
Douglas Creager
Andrew Gallant
David Peter
Mark Z. Ding
Takayuki Maeda
Carl Meyer
Alex Waygood
Dan Parizher
liam