The motivation for this rule is solid; it's been in preview for a long
time; the implementation and tests seem sound; there are no open issues
regarding it, and as far as I can tell there never have been any.
The only issue I see is that the docs don't really describe the rule
accurately right now; I fix that in this PR.
## Summary
This PR migrates our release workflow to
[`cargo-dist`](https://github.com/axodotdev/cargo-dist). The primary
motivation here is that we want to ship dedicated installers for Ruff
that work across platforms, and `cargo-dist` gives us those installers
out-of-the-box. The secondary motivation is that `cargo-dist` formalizes
some of the patterns that we've built up over time in our own release
process.
At a high level:
- The `release.yml` file is generated by `cargo-dist` with `cargo dist
generate`. It doesn't contain any modifications vis-a-vis the generated
file. (If it's edited out of band from generation, the release fails.)
- Our customizations are inserted as custom steps within the
`cargo-dist` workflow. Specifically, `build-binaries` builds the wheels
and packages them into binaries (as on `main`), while `build-docker.yml`
builds the Docker image. `publish-pypi.yml` publishes the wheels to
PyPI. This is effectively our `release.yaml` (on `main`), broken down
into individual workflows rather than steps within a single workflow.
### Changes from `main`
The workflow is _nearly_ unchanged. We kick off a release manually via
the GitHub Action by providing a tag. If the tag doesn't match the
`Cargo.toml`, the release fails. If the tag matches an already-existing
release, the release fails.
The release proceeds by (in order):
0. Doing some upfront validation via `cargo-dist`.
1. Creating the wheels and archives.
2. Building and pushing the Docker image.
3. Publishing to PyPI (if it's not a "dry run").
4. Creating the GitHub Release (if it's not a "dry run").
5. Notifying `ruff-pre-commit` (if it's not a "dry run").
There are a few changes in the workflow as compared to `main`:
- **We no longer validate the SHA** (just the tag). It's not an input to
the job. The Axo team is considering whether / how to support this.
- **Releases are now published directly** (rather than as draft). Again,
the Axo team is considering whether / how to support this. The downside
of drafts is that the URLs aren't stable, so the installers don't work
_as long as the release is in draft_. This is fine for our workflow. It
seems like the Axo team will add it.
- Releases already contain the latest entry from the changelog (we don't
need to copy it over). This "Just Works", which is nice, though we'll
still want to edit them to add contributors.
There are also a few **breaking changes** for consumers of the binaries:
- **We no longer include the version tag in the file name**. This
enables users to install via `/latest` URLs on GitHub, and is part of
the cargo-dist paradigm.
- **Archives now include an extra level of nesting,** which you can
remove with `--strip-components=1` when untarring.
Here's an example release that I created -- I omitted all the artifacts
since I was just testing a workflow, so none of the installers or links
work, but it gives you a sense for what the release looks like:
https://github.com/charliermarsh/cargodisttest/releases/tag/0.1.13.
### Test Plan
I ran a successful release to completion last night, and installed Ruff
via the installer:
The piece I'm least confident about is the Docker push. We build the
image, but the push fails in my test repo since I haven't wired up the
credentials.
## Summary
This rule removes `PLR1701` and redirects it to `SIM101`.
In addition to that, the `SIM101` autofix has been fixed to add padding
if required.
### `PLR1701` has bugs
It also seems that the implementation of `PLR1701` is incorrect in
multiple scenarios. For example, the following code snippet:
```py
# There are two _different_ variables `a` and `b`
if isinstance(a, int) or isinstance(b, bool) or isinstance(a, float):
pass
# There's another condition `or 1`
if isinstance(self.k, int) or isinstance(self.k, float) or 1:
pass
```
is fixed to:
```py
# Fixed to only considering variable `a`
if isinstance(a, (float, int)):
pass
# The additional condition is not present in the fix
if isinstance(self.k, (float, int)):
pass
```
Playground: https://play.ruff.rs/6cfbdfb7-f183-43b0-b59e-31e728b34190
## Documentation Preview
### `PLR1701`
<img width="1397" alt="Screenshot 2024-06-25 at 11 14 40"
src="779ee84d-7c4d-4bb8-a3a4-c2b23a313eba">
## Test Plan
Remove the test cases for `PLR1701`, port the padding test case to
`SIM101` and update the snapshot.
## Summary
This PR splits the re-lexing logic into two parts:
1. `TokenSource`: The token source will be responsible to find the
position the lexer needs to be moved to
2. `Lexer`: The lexer will be responsible to reduce the nesting level
and move itself to the new position if recovered from a parenthesized
context
This split makes it easy to find the new lexer position without needing
to implement the backwards lexing logic again which would need to handle
cases involving:
* Different kinds of newlines
* Line continuation character(s)
* Comments
* Whitespaces
### F-strings
This change did reveal one thing about re-lexing f-strings. Consider the
following example:
```py
f'{'
# ^
f'foo'
```
Here, the quote as highlighted by the caret (`^`) is the start of a
string inside an f-string expression. This is unterminated string which
means the token emitted is actually `Unknown`. The parser tries to
recover from it but there's no newline token in the vector so the new
logic doesn't recover from it. The previous logic does recover because
it's looking at the raw characters instead.
The parser would be at `FStringStart` (the one for the second line) when
it calls into the re-lexing logic to recover from an unterminated
f-string on the first line. So, moving backwards the first character
encountered is a newline character but the first token encountered is an
`Unknown` token.
This is improved with #12067fixes: #12046fixes: #12036
## Test Plan
Update the snapshot and validate the changes.
## Summary
This PR fixes the lexer logic to **not** consume the newline character
for an unterminated string literal.
Currently, the lexer would consume it to be part of the string itself
but that would be bad for recovery because then the lexer wouldn't emit
the newline token ever. This PR fixes that to avoid consuming the
newline character in that case.
This was discovered during https://github.com/astral-sh/ruff/pull/12060.
## Test Plan
Update the snapshots and validate them.
## Summary
This PR fixes a bug introduced in
https://github.com/astral-sh/ruff/pull/12008 which didn't consider the
two character newline after the line continuation character.
For example, consider the following code highlighted with whitespaces:
```py
call(foo # comment \\r\n
\r\n
def bar():\r\n
....pass\r\n
```
The lexer is at `def` when it's running the re-lexing logic and trying
to move back to a newline character. It encounters `\n` and it's being
escaped (incorrect) but `\r` is being escaped, so it moves the lexer to
`\n` character. This creates an overlap in token ranges which causes the
panic.
```
Name 0..4
Lpar 4..5
Name 5..8
Comment 9..20
NonLogicalNewline 20..22 <-- overlap between
Newline 21..22 <-- these two tokens
NonLogicalNewline 22..23
Def 23..26
...
```
fixes: #12028
## Test Plan
Add a test case with line continuation and windows style newline
character.
## Summary
(I'm pretty sure I added this in the parser re-write but must've got
lost in the rebase?)
This PR raises a syntax error if the type parameter list is empty.
As per the grammar, there should be at least one type parameter:
```
type_params:
| invalid_type_params
| '[' type_param_seq ']'
type_param_seq: ','.type_param+ [',']
```
Verified via the builtin `ast` module as well:
```console
$ python3.13 -m ast parser/_.py
Traceback (most recent call last):
[..]
File "parser/_.py", line 1
def foo[]():
^
SyntaxError: Type parameter list cannot be empty
```
## Test Plan
Add inline test cases and update the snapshots.
## Summary
Right now, it's inconsistent... We sometimes match against the name, and
sometimes against the alias (`asname`). I could see a case for always
matching against the name, but matching against both seems fine too,
since the rule is really about the combination of the two?
Closes https://github.com/astral-sh/ruff/issues/12031.
## Summary
This PR fixes a bug where Ruff would raise `E203` for f-string debug
expression. This isn't valid because whitespaces are important for debug
expressions.
fixes: #12023
## Test Plan
Add test case and make sure there are no snapshot changes.
## Summary
This PR updates the parser test infrastructure to validate the token
ranges.
From the code documentation:
```
/// Verifies that:
/// * the ranges are strictly increasing when loop the tokens in insertion order
/// * all ranges are within the length of the source code
```
Follow-up from #12016 and #12017resolves: #11938
## Test Plan
Make sure that there are no failures.
## Summary
This PR updates the unterminated string error range to not include the
final newline character.
This is a follow-up to #12016 and required for #12019
This is not done for when the unterminated string goes till the end of
file (not a newline character). The unterminated f-string range is
correct.
### Why is this required for #12019 ?
Because otherwise the token ranges will overlap. For example:
```py
f"{"
f"{foo!r"
```
Here, the re-lexing logic recovers from an unterminated f-string and
thus emitting a `Newline` token for the one at the end of the first
line. But, currently the `Unknown` and the `Newline` token would overlap
because the `Unknown` token (unterminated string literal) range would
include the newline character.
## Test Plan
Update and validate the snapshot.
## Summary
This PR fixes the range highlighted for the line continuation error.
Previously, it would highlight an incorrect range:
```
1 | call(a, b, \\\
| ^^ Syntax Error: unexpected character after line continuation character
2 |
3 | def bar():
|
```
And now:
```
|
1 | call(a, b, \\\
| ^ Syntax Error: unexpected character after line continuation character
2 |
3 | def bar():
|
```
This is implemented by avoiding to update the token range for the
`Unknown` token which is emitted when there's a lexical error. Instead,
the `push_error` helper method will be responsible to update the range
to the error location.
This actually becomes a requirement which can be seen in follow-up PRs.
## Test Plan
Update and validate the snapshot.
## Summary
This PR fixes a bug where the re-lexing logic didn't consider the line
continuation character being present before the newline character. This
meant that the lexer was being moved back to the newline character which
is actually ignored via `\`.
Considering the following code:
```py
f'middle {'string':\
'format spec'}
```
The old token stream is:
```
...
Colon 18..19
FStringMiddle 19..29 (flags = F_STRING)
Newline 20..21
Indent 21..29
String 29..42
Rbrace 42..43
...
```
Notice how the ranges are overlapping between the `FStringMiddle` token
and the tokens emitted after moving the lexer backwards.
After this fix, the new token stream which is without moving the lexer
backwards in this scenario:
```
FStringStart 0..2 (flags = F_STRING)
FStringMiddle 2..9 (flags = F_STRING)
Lbrace 9..10
String 10..18
Colon 18..19
FStringMiddle 19..29 (flags = F_STRING)
FStringEnd 29..30 (flags = F_STRING)
Name 30..36
Name 37..41
Unknown 41..44
Newline 44..45
```
fixes: #12004
## Test Plan
Add test cases and update the snapshots.
## Summary
This PR updates `F811` rule to include assignment as possible shadowed
binding. This will fix issue: #11828 .
## Test Plan
Add a test file, F811_30.py, which includes a redefinition after an
assignment and a verified snapshot file.
## Summary
Addresses #11974 to add a `RUF` rule to replace `print` expressions in
`assert` statements with the inner message.
An autofix is available, but is considered unsafe as it changes
behaviour of the execution, notably:
- removal of the printout in `stdout`, and
- `AssertionError` instance containing a different message.
While the detection of the condition is a straightforward matter,
deciding how to resolve the print arguments into a string literal can be
a relatively subjective matter. The implementation of this PR chooses to
be as tolerant as possible, and will attempt to reformat any number of
`print` arguments containing single or concatenated strings or variables
into either a string literal, or a f-string if any variables or
placeholders are detected.
## Test Plan
`cargo test`.
## Examples
For ease of discussion, this is the diff for the tests:
```diff
# Standard Case
# Expects:
# - single StringLiteral
-assert True, print("This print is not intentional.")
+assert True, "This print is not intentional."
# Concatenated string literals
# Expects:
# - single StringLiteral
-assert True, print("This print" " is not intentional.")
+assert True, "This print is not intentional."
# Positional arguments, string literals
# Expects:
# - single StringLiteral concatenated with " "
-assert True, print("This print", "is not intentional")
+assert True, "This print is not intentional"
# Concatenated string literals combined with Positional arguments
# Expects:
# - single stringliteral concatenated with " " only between `print` and `is`
-assert True, print("This " "print", "is not intentional.")
+assert True, "This print is not intentional."
# Positional arguments, string literals with a variable
# Expects:
# - single FString concatenated with " "
-assert True, print("This", print.__name__, "is not intentional.")
+assert True, f"This {print.__name__} is not intentional."
# Mixed brackets string literals
# Expects:
# - single StringLiteral concatenated with " "
-assert True, print("This print", 'is not intentional', """and should be removed""")
+assert True, "This print is not intentional and should be removed"
# Mixed brackets with other brackets inside
# Expects:
# - single StringLiteral concatenated with " " and escaped brackets
-assert True, print("This print", 'is not "intentional"', """and "should" be 'removed'""")
+assert True, "This print is not \"intentional\" and \"should\" be 'removed'"
# Positional arguments, string literals with a separator
# Expects:
# - single StringLiteral concatenated with "|"
-assert True, print("This print", "is not intentional", sep="|")
+assert True, "This print|is not intentional"
# Positional arguments, string literals with None as separator
# Expects:
# - single StringLiteral concatenated with " "
-assert True, print("This print", "is not intentional", sep=None)
+assert True, "This print is not intentional"
# Positional arguments, string literals with variable as separator, needs f-string
# Expects:
# - single FString concatenated with "{U00A0}"
-assert True, print("This print", "is not intentional", sep=U00A0)
+assert True, f"This print{U00A0}is not intentional"
# Unnecessary f-string
# Expects:
# - single StringLiteral
-assert True, print(f"This f-string is just a literal.")
+assert True, "This f-string is just a literal."
# Positional arguments, string literals and f-strings
# Expects:
# - single FString concatenated with " "
-assert True, print("This print", f"is not {'intentional':s}")
+assert True, f"This print is not {'intentional':s}"
# Positional arguments, string literals and f-strings with a separator
# Expects:
# - single FString concatenated with "|"
-assert True, print("This print", f"is not {'intentional':s}", sep="|")
+assert True, f"This print|is not {'intentional':s}"
# A single f-string
# Expects:
# - single FString
-assert True, print(f"This print is not {'intentional':s}")
+assert True, f"This print is not {'intentional':s}"
# A single f-string with a redundant separator
# Expects:
# - single FString
-assert True, print(f"This print is not {'intentional':s}", sep="|")
+assert True, f"This print is not {'intentional':s}"
# Complex f-string with variable as separator
# Expects:
# - single FString concatenated with "{U00A0}", all placeholders preserved
condition = "True is True"
maintainer = "John Doe"
-assert True, print("Unreachable due to", condition, f", ask {maintainer} for advice", sep=U00A0)
+assert True, f"Unreachable due to{U00A0}{condition}{U00A0}, ask {maintainer} for advice"
# Empty print
# Expects:
# - `msg` entirely removed from assertion
-assert True, print()
+assert True
# Empty print with separator
# Expects:
# - `msg` entirely removed from assertion
-assert True, print(sep=" ")
+assert True
# Custom print function that actually returns a string
# Expects:
@@ -100,4 +100,4 @@
# Use of `builtins.print`
# Expects:
# - single StringLiteral
-assert True, builtins.print("This print should be removed.")
+assert True, "This print should be removed."
```
## Known Issues
The current implementation resolves all arguments and separators of the
`print` expression into a single string, be it
`StringLiteralValue::single` or a `FStringValue::single`. This:
- potentially joins together strings well beyond the ideal character
limit for each line, and
- does not preserve multi-line strings in their original format, in
favour of a single line `"...\n...\n..."` format.
These are purely formatting issues only occurring in unusual scenarios.
Additionally, the autofix will tolerate `print` calls that were
previously invalid:
```python
assert True, print("this", "should not be allowed", sep=42)
```
This will be transformed into
```python
assert True, f"this{42}should not be allowed"
```
which some could argue is an alteration of behaviour.
---------
Co-authored-by: Charlie Marsh <charlie.r.marsh@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
Documentation mentions:
> PEP 563 enabled the use of a number of convenient type annotations,
such as `list[str]` instead of `List[str]`
but it meant [PEP 585](https://peps.python.org/pep-0585/) instead.
[PEP 563](https://peps.python.org/pep-0563/) is the one defining `from
__future__ import annotations`.
## Test Plan
No automated test required, just verify that
https://peps.python.org/pep-0585/ is the correct reference.
## Summary
I look at the token stream a lot, not specifically in the playground but
in the terminal output and it's annoying to scroll a lot to find
specific location. Most of the information is also redundant.
The final format we end up with is: `<kind> <range> (flags = ...)` e.g.,
`String 0..4 (flags = BYTE_STRING)` where the flags part is only
populated if there are any flags set.
## Summary
Fixes#11651.
Fixes#11851.
We were double-closing a notebook document from the index, once in
`textDocument/didClose` and then in the `notebookDocument/didClose`
handler. The second time this happens, taking a snapshot fails.
I've rewritten how we handle snapshots for closing notebooks / notebook
cells so that any failure is simply logged instead of propagating
upwards. This implementation works consistently even if we don't receive
`textDocument/didClose` notifications for each specific cell, since they
get closed (and the diagnostics get cleared) in the notebook document
removal process.
## Test Plan
1. Open an untitled, unsaved notebook with the `Create: New Jupyter
Notebook` command from the VS Code command palette (`Ctrl/Cmd + Shift +
P`)
2. Without saving the document, close it.
3. No error popup should appear.
4. Run the debug command (`Ruff: print debug information`) to confirm
that there are no open documents
