## Summary
This PR refactors unary expression parsing with the following changes:
* Ability to get `OperatorPrecedence` from a unary operator (`UnaryOp`)
* Implement methods on `TokenKind`
* Add `as_unary_operator` which returns an `Option<UnaryOp>`
* Add `as_unary_arithmetic_operator` which returns an `Option<UnaryOp>`
(used for pattern parsing)
* Rename `is_unary` to `is_unary_arithmetic_operator` (used in the
linter)
resolves: #10752
## Test Plan
Verify that the existing test cases pass, no ecosystem changes, run the
Python based fuzzer on 3000 random inputs and run it on dozens of
open-source repositories.
## Summary
This PR refactors the binary expression parsing in a way to make it
readable and easy to understand. It draws inspiration from the suggested
edits in the linked messages in #10752.
### Changes
* Ability to get the precedence of an operator
* From a boolean operator (`BinOp`) to `OperatorPrecedence`
* From a binary operator (`Operator`) to `OperatorPrecedence`
* No comparison operator because all of them have the same precedence
* Implement methods on `TokenKind` to convert it to an appropriate
operator enum
* Add `as_boolean_operator` which returns an `Option<BoolOp>`
* Add `as_binary_operator` which returns an `Option<Operator>`
* No `as_comparison_operator` because it requires lookahead and I'm not
sure if `token.as_comparison_operator(peek)` is a good way to implement
it
* Introduce `BinaryLikeOperator`
* Constructed from two tokens using the methods from the second point
* Add `precedence` method using the conversion methods mentioned in the
first point
* Make most of the functions in `TokenKind` private to the module
* Use `self` instead of `&self` for `TokenKind`
fixes: #11072
## Test Plan
Refer #11088
## Summary
This PR does a few things but the main change is that is makes
associativity a property of operator precedence.
1. Rename `Precedence` -> `OperatorPrecedence`
2. Rename `parse_expression_with_precedence` ->
`parse_binary_expression_or_higher`
3. Move `current_binding_power` to `OperatorPrecedence::try_from_tokens`
[^1]
4. Add a `OperatorPrecedence::is_right_associative` method
5. Move from `increment_precedence` to using `<=` / `<` to check if the
parsing loop needs to stop [^2]
[^1]: Another alternative would be to have two separate methods to avoid
lookahead as it's required only for once case (`not in`). So,
`try_from_current_token(current).or_else(|| try_from_next_token(current,
peek))`
[^2]: This will allow us to easily make the refactors mentioned in
#10752
## Test Plan
Make sure the precedence parsing algorithm is still correct by running
the test suite, fuzz testing it and running it against a dozen or so
open-source repositories.
## Summary
This PR adds a new `ExpressionContext` struct which is used in
expression parsing.
This solves the following problem:
1. Allowing starred expression with different precedence
2. Allowing yield expression in certain context
3. Remove ambiguity with `in` keyword when parsing a `for ... in`
statement
For context, (1) was solved by adding `parse_star_expression_list` and
`parse_star_expression_or_higher` in #10623, (2) was solved by by adding
`parse_yield_expression_or_else` in #10809, and (3) was fixed in #11009.
All of the mentioned functions have been removed in favor of the context
flags.
As mentioned in #11009, an ideal solution would be to implement an
expression context which is what this PR implements. This is passed
around as function parameter and the call stack is used to automatically
reset the context.
### Recovery
How should the parser recover if the target expression is invalid when
an expression can consume the `in` keyword?
1. Should the `in` keyword be part of the target expression?
2. Or, should the expression parsing stop as soon as `in` keyword is
encountered, no matter the expression?
For example:
```python
for yield x in y: ...
# Here, should this be parsed as
for (yield x) in (y): ...
# Or
for (yield x in y): ...
# where the `in iter` part is missing
```
Or, for binary expression parsing:
```python
for x or y in z: ...
# Should this be parsed as
for (x or y) in z: ...
# Or
for (x or y in z): ...
# where the `in iter` part is missing
```
This need not be solved now, but is very easy to change. For context
this PR does the following:
* For binary, comparison, and unary expressions, stop at `in`
* For lambda, yield expressions, consume the `in`
## Test Plan
1. Add test cases for the `for ... in` statement and verify the
snapshots
2. Make sure the existing test suite pass
3. Run the fuzzer for around 3000 generated source code
4. Run the updated logic on a dozen or so open source repositories
(codename "parser-checkouts")
## Summary
I happened to notice that we box `TypeParams` on `StmtClassDef` but not
on `StmtFunctionDef` and wondered why, since `StmtFunctionDef` is bigger
and sets the size of `Stmt`.
@charliermarsh found that at the time we started boxing type params on
classes, classes were the largest statement type (see #6275), but that's
no longer true.
So boxing type-params also on functions reduces the overall size of
`Stmt`.
## Test Plan
The `<=` size tests are a bit irritating (since their failure doesn't
tell you the actual size), but I manually confirmed that the size is
actually 120 now.
## Summary
This fixes a bug where the parser would panic when there is a "gap" in
the token source.
What's a gap?
The reason it's `<=` instead of just `==` is because there could be
whitespaces between
the two tokens. For example:
```python
# last token end
# | current token (newline) start
# v v
def foo \n
# ^
# assume there's trailing whitespace here
```
Or, there could tokens that are considered "trivia" and thus aren't
emitted by the token
source. These are comments and non-logical newlines. For example:
```python
# last token end
# v
def foo # comment\n
# ^ current token (newline) start
```
In either of the above cases, there's a "gap" between the end of the
last token and start
of the current token.
## Test Plan
Add test cases and update the snapshots.
## Summary
This PR adds a new `Clause::Case` and uses it to parse the body of a
`case` block. Earlier, it was using `Match` which would give an
incorrect error message like:
```
|
1 | match subject:
2 | case 1:
3 | case 2: ...
| ^^^^ Syntax Error: Expected an indented block after `match` statement
|
```
## Test Plan
Add test case and update the snapshot.
## Summary
This PR fixes the bug in with items parsing where it would fail to
recognize that the parenthesized expression is part of a large binary
expression.
## Test Plan
Add test cases and verified the snapshots.
## Summary
This PR fixes the bug in parenthesized with items parsing where the `if`
expression would result into a syntax error.
The reason being that once we identify that the ambiguous left
parenthesis belongs to the context expression, the parser converts the
parsed with item into an equivalent expression. Then, the parser
continuous to parse any postfix expressions. Now, attribute, subscript,
and call are taken into account as they're grouped in
`parse_postfix_expression` but `if` expression has it's own parsing
function.
Use `parse_if_expression` once all postfix expressions have been parsed.
Ideally, I think that `if` could be included in postfix expression
parsing as they can be chained as well (`x if True else y if True else
z`).
## Test Plan
Add test cases and verified the snapshots.
## Summary
This PR fixes a bug in the new parser which involves the parser context
w.r.t. for statement. This is specifically around the `in` keyword which
can be present in the target expression and shouldn't be considered to
be part of the `for` statement header. Ideally it should use a context
which is passed between functions, thus using a call stack to set /
unset a specific variant which will be done in a follow-up PR as it
requires some amount of refactor.
## Test Plan
Add test cases and update the snapshots.
(Supersedes #9152, authored by @LaBatata101)
## Summary
This PR replaces the current parser generated from LALRPOP to a
hand-written recursive descent parser.
It also updates the grammar for [PEP
646](https://peps.python.org/pep-0646/) so that the parser outputs the
correct AST. For example, in `data[*x]`, the index expression is now a
tuple with a single starred expression instead of just a starred
expression.
Beyond the performance improvements, the parser is also error resilient
and can provide better error messages. The behavior as seen by any
downstream tools isn't changed. That is, the linter and formatter can
still assume that the parser will _stop_ at the first syntax error. This
will be updated in the following months.
For more details about the change here, refer to the PR corresponding to
the individual commits and the release blog post.
## Test Plan
Write _lots_ and _lots_ of tests for both valid and invalid syntax and
verify the output.
## Acknowledgements
- @MichaReiser for reviewing 100+ parser PRs and continuously providing
guidance throughout the project
- @LaBatata101 for initiating the transition to a hand-written parser in
#9152
- @addisoncrump for implementing the fuzzer which helped
[catch](https://github.com/astral-sh/ruff/pull/10903)
[a](https://github.com/astral-sh/ruff/pull/10910)
[lot](https://github.com/astral-sh/ruff/pull/10966)
[of](https://github.com/astral-sh/ruff/pull/10896)
[bugs](https://github.com/astral-sh/ruff/pull/10877)
---------
Co-authored-by: Victor Hugo Gomes <labatata101@linuxmail.org>
Co-authored-by: Micha Reiser <micha@reiser.io>
## Summary
I used `codespell` and `gramma` to identify mispellings and grammar
errors throughout the codebase and fixed them. I tried not to make any
controversial changes, but feel free to revert as you see fit.
## Summary
Fix#10282
This PR updates the Python grammar to include the `*` character in
`*args` `**kwargs` in the range of the `Parameter`
```
def f(*args, **kwargs): pass
# ~~~~ ~~~~~~ <-- range before the PR
# ^^^^^ ^^^^^^^^ <-- range after
```
The invalid syntax `def f(*, **kwargs): ...` is also now correctly
reported.
## Test Plan
Test cases were added to `function.rs`.
This PR modifies our AST so that nodes for string literals, bytes literals and f-strings all retain the following information:
- The quoting style used (double or single quotes)
- Whether the string is triple-quoted or not
- Whether the string is raw or not
This PR is a followup to #10256. Like with that PR, this PR does not, in itself, fix any bugs. However, it means that we will have the necessary information to preserve quoting style and rawness of strings in the `ExprGenerator` in a followup PR, which will allow us to provide a fix for https://github.com/astral-sh/ruff/issues/7799.
The information is recorded on the AST nodes using a bitflag field on each node, similarly to how we recorded the information on `Tok::String`, `Tok::FStringStart` and `Tok::FStringMiddle` tokens in #10298. Rather than reusing the bitflag I used for the tokens, however, I decided to create a custom bitflag for each AST node.
Using different bitflags for each node allows us to make invalid states unrepresentable: it is valid to set a `u` prefix on a string literal, but not on a bytes literal or an f-string. It also allows us to have better debug representations for each AST node modified in this PR.
The expression types in our AST are called `ExprYield`, `ExprAwait`,
`ExprStringLiteral` etc, except `ExprNamedExpr`, `ExprIfExpr` and
`ExprGenratorExpr`. This seems to align with [Python AST's
naming](https://docs.python.org/3/library/ast.html) but feels
inconsistent and excessive.
This PR removes the `Expr` postfix from `ExprNamedExpr`, `ExprIfExpr`,
and `ExprGeneratorExpr`.
## Summary
This PR fixes the `DebugText` implementation to use the expression range
instead of the parenthesized range.
Taking the following code snippet as an example:
```python
x = 1
print(f"{ ( x ) = }")
```
The output of running it would be:
```
( x ) = 1
```
Notice that the whitespace between the parentheses and the expression is
preserved as is.
Currently, we don't preserve this information in the AST which defeats
the purpose of `DebugText` as the main purpose of the struct is to
preserve whitespaces _around_ the expression.
This is also problematic when generating the code from the AST node as
then the generator has no information about the parentheses the
whitespaces between them and the expression which would lead to the
removal of the parentheses in the generated code.
I noticed this while working on the f-string formatting where the debug
text would be used to preserve the text surrounding the expression in
the presence of debug expression. The parentheses were being dropped
then which made me realize that the problem is instead in the parser.
## Test Plan
1. Add a test case for the parser
2. Add a test case for the generator
## Summary
This PR reduces the size of `Expr` from 80 to 64 bytes, by reducing the
sizes of...
- `ExprCall` from 72 to 56 bytes, by using boxed slices for `Arguments`.
- `ExprCompare` from 64 to 48 bytes, by using boxed slices for its
various vectors.
In testing, the parser gets a bit faster, and the linter benchmarks
improve quite a bit.
## Summary
Given:
```python
F"{"ڤ
```
We try to locate the "unclosed left brace" error by subtracting the
quote size from the lexer offset -- so we subtract 1 from the end of the
source, which puts us in the middle of a Unicode character. I don't
think we should try to adjust the offset in this way, since there can be
content _after_ the quote. For example, with the advent of PEP 701, this
string could reasonably be fixed as:
```python
F"{"ڤ"}"
````
Closes https://github.com/astral-sh/ruff/issues/9379.
## Summary
This PR modifies our `Cargo.toml` files to use workspace dependencies
for _all_ dependencies, rather than the status quo of sporadically
trying to use workspace dependencies for those dependencies that are
used across multiple crates. I find the current situation more confusing
and harder to manage, since we have a mix of workspace and crate-local
dependencies, whereas this setup consistently uses the same approach for
all dependencies.
## Summary
I always found it odd that we had to pass this in, since it's really
higher-level context for the error. The awkwardness is further evidenced
by the fact that we pass in fake values everywhere (even outside of
tests). The source path isn't actually used to display the error; it's
only accessed elsewhere to _re-display_ the error in certain cases. This
PR modifies to instead pass the path directly in those cases.
## Summary
This helps a bit with (but does not close) the issues described in
https://github.com/astral-sh/ruff/issues/9311. E.g., now, we at least
see: `error: Failed to format main.py: source contains syntax errors:
invalid syntax. Got unexpected token '=' at byte offset 20`.
## Summary
This PR adds some helper structs to the linter paths to enable passing
in the pre-computed tokens and parsed source code during benchmarking,
to remove lexing and parsing from the overall linter benchmark
measurement. We already remove parsing for the formatter, and we have
separate benchmarks for the lexer and the parser, so this should make it
much easier to measure linter performance changes.
This sets `lto = "thin"` instead of using "fat" LTO, and sets
`codegen-units = 16`. These are the defaults for Cargo's `release`
profile, and I think it may give us faster iteration times, especially
when benchmarking. The point of this PR is to see what kind of impact
this has on benchmarks. It is expected that benchmarks may regress to
some extent.
I did some quick ad hoc experiments to quantify this change in compile
times. Namely, I ran:
cargo build --profile release -p ruff_cli
Then I ran
touch crates/ruff_python_formatter/src/expression/string/docstring.rs
(because that's where i've been working lately) and re-ran
cargo build --profile release -p ruff_cli
This last command is what I timed, since it reflects how much time one
has to wait between making a change and getting a compiled artifact.
Here are my results:
* With status quo `release` profile, build takes 77s
* with `release` but `lto = "thin"`, build takes 41s
* with `release`, but `lto = false`, build takes 19s
* with `release`, but `lto = false` **and** `codegen-units = 16`, build
takes 7s
* with `release`, but `lto = "thin"` **and** `codegen-units = 16`, build
takes 16s (i believe this is the default `release` configuration)
This PR represents the last option. It's not the fastest to compile, but
it's nearly a whole minute faster! The idea is that with `codegen-units
= 16`, we still make use of parallelism, but keep _some_ level of LTO on
to try and re-gain what we lose by increasing the number of codegen
units.
Rebase of #6365 authored by @davidszotten.
## Summary
This PR updates the AST structure for an f-string elements.
The main **motivation** behind this change is to have a dedicated node
for the string part of an f-string. Previously, the existing
`ExprStringLiteral` node was used for this purpose which isn't exactly
correct. The `ExprStringLiteral` node should include the quotes as well
in the range but the f-string literal element doesn't include the quote
as it's a specific part within an f-string. For example,
```python
f"foo {x}"
# ^^^^
# This is the literal part of an f-string
```
The introduction of `FStringElement` enum is helpful which represent
either the literal part or the expression part of an f-string.
### Rule Updates
This means that there'll be two nodes representing a string depending on
the context. One for a normal string literal while the other is a string
literal within an f-string. The AST checker is updated to accommodate
this change. The rules which work on string literal are updated to check
on the literal part of f-string as well.
#### Notes
1. The `Expr::is_literal_expr` method would check for
`ExprStringLiteral` and return true if so. But now that we don't
represent the literal part of an f-string using that node, this improves
the method's behavior and confines to the actual expression. We do have
the `FStringElement::is_literal` method.
2. We avoid checking if we're in a f-string context before adding to
`string_type_definitions` because the f-string literal is now a
dedicated node and not part of `Expr`.
3. Annotations cannot use f-string so we avoid changing any rules which
work on annotation and checks for `ExprStringLiteral`.
## Test Plan
- All references of `Expr::StringLiteral` were checked to see if any of
the rules require updating to account for the f-string literal element
node.
- New test cases are added for rules which check against the literal
part of an f-string.
- Check the ecosystem results and ensure it remains unchanged.
## Performance
There's a performance penalty in the parser. The reason for this remains
unknown as it seems that the generated assembly code is now different
for the `__reduce154` function. The reduce function body is just popping
the `ParenthesizedExpr` on top of the stack and pushing it with the new
location.
- The size of `FStringElement` enum is the same as `Expr` which is what
it replaces in `FString::format_spec`
- The size of `FStringExpressionElement` is the same as
`ExprFormattedValue` which is what it replaces
I tried reducing the `Expr` enum from 80 bytes to 72 bytes but it hardly
resulted in any performance gain. The difference can be seen here:
- Original profile: https://share.firefox.dev/3Taa7ES
- Profile after boxing some node fields:
https://share.firefox.dev/3GsNXpD
### Backtracking
I tried backtracking the changes to see if any of the isolated change
produced this regression. The problem here is that the overall change is
so small that there's only a single checkpoint where I can backtrack and
that checkpoint results in the same regression. This checkpoint is to
revert using `Expr` to the `FString::format_spec` field. After this
point, the change would revert back to the original implementation.
## Review process
The review process is similar to #7927. The first set of commits update
the node structure, parser, and related AST files. Then, further commits
update the linter and formatter part to account for the AST change.
---------
Co-authored-by: David Szotten <davidszotten@gmail.com>
<!--
Thank you for contributing to Ruff! To help us out with reviewing,
please consider the following:
- Does this pull request include a summary of the change? (See below.)
- Does this pull request include a descriptive title?
- Does this pull request include references to any relevant issues?
-->
## Summary
Our `SoftKeywordTokenizer` only respected soft keywords in compound
statement positions -- for example, at the start of a logical line:
```python
type X = int
```
However, type aliases can also appear in simple statement positions,
like:
```python
class Class: type X = int
```
(Note that `match` and `case` are _not_ valid keywords in such
positions.)
This PR upgrades the tokenizer to track both kinds of valid positions.
Closes https://github.com/astral-sh/ruff/issues/8900.
Closes https://github.com/astral-sh/ruff/issues/8899.
## Test Plan
`cargo test`
## Summary
Given `with (a := b): pass`, we truncate the `WithItem` range by one on
both sides such that the parentheses are part of the statement, rather
than the item. However, for `with (a := b) as x: pass`, we want to avoid
this trick.
Closes https://github.com/astral-sh/ruff/issues/8913.
