## Summary
This PR adds (unsafe) fixes to the flake8-annotations rules that enforce
missing return types, offering to automatically insert type annotations
for functions with literal return values. The logic is smart enough to
generate simplified unions (e.g., `float` instead of `int | float`) and
deal with implicit returns (`return` without a value).
Closes https://github.com/astral-sh/ruff/issues/1640 (though we could
open a separate issue for referring parameter types).
Closes https://github.com/astral-sh/ruff/issues/8213.
## Test Plan
`cargo test`
## Summary
Adds an extra check to F632 to check for any `is` comparisons to a
mutable initialisers.
Implements #8589 .
Example:
```Python
named_var = {}
if named_var is {}: # F632 (fix)
pass
```
The if condition will always evaluate to False because it checks on
identity and it's impossible to take the same identity as a hard coded
list/set/dict initializer.
## Test Plan
Multiple test cases were added to ensure the rule works + doesn't flag
false positives + the fix works correctly.
## Summary
Adds `TRIO105` from the [flake8-trio
plugin](https://github.com/Zac-HD/flake8-trio). The `MethodName` logic
mirrors that of `TRIO100` to stay consistent within the plugin.
It is at 95% parity with the exception of upstream also checking for a
slightly more complex scenario where a call to `start()` on a
`trio.Nursery` context should also be immediately awaited. Upstream
plugin appears to just check for anything named `nursery` judging from
[the relevant issue](https://github.com/Zac-HD/flake8-trio/issues/56).
Unsure if we want to do so something similar or, alternatively, if there
is some capability in ruff to check for calls made on this context some
other way
## Test Plan
Added a new fixture, based on [the one from upstream
plugin](https://github.com/Zac-HD/flake8-trio/blob/main/tests/eval_files/trio105.py)
## Issue link
Refers: https://github.com/astral-sh/ruff/issues/8451
## Summary
If the value of `shell` wasn't literally `True`, we now show a message
describing it as truthy, rather than the (misleading) `shell=True`
literal in the diagnostic.
Closes https://github.com/astral-sh/ruff/issues/8310.
## Summary
This PR splits the `Constant` enum as individual literal nodes. It
introduces the following new nodes for each variant:
* `ExprStringLiteral`
* `ExprBytesLiteral`
* `ExprNumberLiteral`
* `ExprBooleanLiteral`
* `ExprNoneLiteral`
* `ExprEllipsisLiteral`
The main motivation behind this refactor is to introduce the new AST
node for implicit string concatenation in the coming PR. The elements of
that node will be either a string literal, bytes literal or a f-string
which can be implemented using an enum. This means that a string or
bytes literal cannot be represented by `Constant::Str` /
`Constant::Bytes` which creates an inconsistency.
This PR avoids that inconsistency by splitting the constant nodes into
it's own literal nodes, literal being the more appropriate naming
convention from a static analysis tool perspective.
This also makes working with literals in the linter and formatter much
more ergonomic like, for example, if one would want to check if this is
a string literal, it can be done easily using
`Expr::is_string_literal_expr` or matching against `Expr::StringLiteral`
as oppose to matching against the `ExprConstant` and enum `Constant`. A
few AST helper methods can be simplified as well which will be done in a
follow-up PR.
This introduces a new `Expr::is_literal_expr` method which is the same
as `Expr::is_constant_expr`. There are also intermediary changes related
to implicit string concatenation which are quiet less. This is done so
as to avoid having a huge PR which this already is.
## Test Plan
1. Verify and update all of the existing snapshots (parser, visitor)
2. Verify that the ecosystem check output remains **unchanged** for both
the linter and formatter
### Formatter ecosystem check
#### `main`
| project | similarity index | total files | changed files |
|----------------|------------------:|------------------:|------------------:|
| cpython | 0.75803 | 1799 | 1647 |
| django | 0.99983 | 2772 | 34 |
| home-assistant | 0.99953 | 10596 | 186 |
| poetry | 0.99891 | 317 | 17 |
| transformers | 0.99966 | 2657 | 330 |
| twine | 1.00000 | 33 | 0 |
| typeshed | 0.99978 | 3669 | 20 |
| warehouse | 0.99977 | 654 | 13 |
| zulip | 0.99970 | 1459 | 22 |
#### `dhruv/constant-to-literal`
| project | similarity index | total files | changed files |
|----------------|------------------:|------------------:|------------------:|
| cpython | 0.75803 | 1799 | 1647 |
| django | 0.99983 | 2772 | 34 |
| home-assistant | 0.99953 | 10596 | 186 |
| poetry | 0.99891 | 317 | 17 |
| transformers | 0.99966 | 2657 | 330 |
| twine | 1.00000 | 33 | 0 |
| typeshed | 0.99978 | 3669 | 20 |
| warehouse | 0.99977 | 654 | 13 |
| zulip | 0.99970 | 1459 | 22 |
## Summary
Implement
[`no-single-item-in`](https://github.com/dosisod/refurb/blob/master/refurb/checks/iterable/no_single_item_in.py)
as `single-item-membership-test` (`FURB171`).
Uses the helper function `generate_comparison` from the `pycodestyle`
implementations; this function should probably be moved, but I am not
sure where at the moment.
Update: moved it to `ruff_python_ast::helpers`.
Related to #1348.
## Test Plan
`cargo test`
## Summary
This is a follow-up to #7469 that attempts to achieve similar gains, but
without introducing malachite. Instead, this PR removes the `BigInt`
type altogether, instead opting for a simple enum that allows us to
store small integers directly and only allocate for values greater than
`i64`:
```rust
/// A Python integer literal. Represents both small (fits in an `i64`) and large integers.
#[derive(Clone, PartialEq, Eq, Hash)]
pub struct Int(Number);
#[derive(Debug, Clone, PartialEq, Eq, Hash)]
pub enum Number {
/// A "small" number that can be represented as an `i64`.
Small(i64),
/// A "large" number that cannot be represented as an `i64`.
Big(Box<str>),
}
impl std::fmt::Display for Number {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
match self {
Number::Small(value) => write!(f, "{value}"),
Number::Big(value) => write!(f, "{value}"),
}
}
}
```
We typically don't care about numbers greater than `isize` -- our only
uses are comparisons against small constants (like `1`, `2`, `3`, etc.),
so there's no real loss of information, except in one or two rules where
we're now a little more conservative (with the worst-case being that we
don't flag, e.g., an `itertools.pairwise` that uses an extremely large
value for the slice start constant). For simplicity, a few diagnostics
now show a dedicated message when they see integers that are out of the
supported range (e.g., `outdated-version-block`).
An additional benefit here is that we get to remove a few dependencies,
especially `num-bigint`.
## Test Plan
`cargo test`
## Motivation
The `ast::Arguments` for call argument are split into positional
arguments (args) and keywords arguments (keywords). We currently assume
that call consists of first args and then keywords, which is generally
the case, but not always:
```python
f(*args, a=2, *args2, **kwargs)
class A(*args, a=2, *args2, **kwargs):
pass
```
The consequence is accidentally reordering arguments
(https://github.com/astral-sh/ruff/pull/7268).
## Summary
`Arguments::args_and_keywords` returns an iterator of an `ArgOrKeyword`
enum that yields args and keywords in the correct order. I've fixed the
obvious `args` and `keywords` usages, but there might be some cases with
wrong assumptions remaining.
## Test Plan
The generator got new test cases, otherwise the stacked PR
(https://github.com/astral-sh/ruff/pull/7268) which uncovered this.
## Summary
The motivation here is that this enables us to implement `Ranged` in
crates that don't depend on `ruff_python_ast`.
Largely a mechanical refactor with a lot of regex, Clippy help, and
manual fixups.
## Test Plan
`cargo test`
## Summary
This PR introduces two new AST nodes to improve the representation of
`PatternMatchClass`. As a reminder, `PatternMatchClass` looks like this:
```python
case Point2D(0, 0, x=1, y=2):
...
```
Historically, this was represented as a vector of patterns (for the `0,
0` portion) and parallel vectors of keyword names (for `x` and `y`) and
values (for `1` and `2`). This introduces a bunch of challenges for the
formatter, but importantly, it's also really different from how we
represent similar nodes, like arguments (`func(0, 0, x=1, y=2)`) or
parameters (`def func(x, y)`).
So, firstly, we now use a single node (`PatternArguments`) for the
entire parenthesized region, making it much more consistent with our
other nodes. So, above, `PatternArguments` would be `(0, 0, x=1, y=2)`.
Secondly, we now have a `PatternKeyword` node for `x=1` and `y=2`. This
is much more similar to the how `Keyword` is represented within
`Arguments` for call expressions.
Closes https://github.com/astral-sh/ruff/issues/6866.
Closes https://github.com/astral-sh/ruff/issues/6880.
## Summary
Instead, we set an `is_star` flag on `Stmt::Try`. This is similar to the
pattern we've migrated towards for `Stmt::For` (removing
`Stmt::AsyncFor`) and friends. While these are significant differences
for an interpreter, we tend to handle these cases identically or nearly
identically.
## Test Plan
`cargo test`
## Summary
Per the discussion in
https://github.com/astral-sh/ruff/discussions/6183, this PR adds an
`implicit_concatenated` flag to the string and bytes constant variants.
It's not actually _used_ anywhere as of this PR, but it is covered by
the tests.
Specifically, we now use a struct for the string and bytes cases, along
with the `Expr::FString` node. That struct holds the value, plus the
flag:
```rust
#[derive(Clone, Debug, PartialEq, is_macro::Is)]
pub enum Constant {
Str(StringConstant),
Bytes(BytesConstant),
...
}
#[derive(Clone, Debug, PartialEq, Eq)]
pub struct StringConstant {
/// The string value as resolved by the parser (i.e., without quotes, or escape sequences, or
/// implicit concatenations).
pub value: String,
/// Whether the string contains multiple string tokens that were implicitly concatenated.
pub implicit_concatenated: bool,
}
impl Deref for StringConstant {
type Target = str;
fn deref(&self) -> &Self::Target {
self.value.as_str()
}
}
#[derive(Clone, Debug, PartialEq, Eq)]
pub struct BytesConstant {
/// The bytes value as resolved by the parser (i.e., without quotes, or escape sequences, or
/// implicit concatenations).
pub value: Vec<u8>,
/// Whether the string contains multiple string tokens that were implicitly concatenated.
pub implicit_concatenated: bool,
}
impl Deref for BytesConstant {
type Target = [u8];
fn deref(&self) -> &Self::Target {
self.value.as_slice()
}
}
```
## Test Plan
`cargo test`
## Summary
This PR renames the `MagicCommand` token to `IpyEscapeCommand` token and
`MagicKind` to `IpyEscapeKind` type to better reflect the purpose of the
token and type. Similarly, it renames the AST nodes from `LineMagic` to
`IpyEscapeCommand` prefixed with `Stmt`/`Expr` wherever necessary.
It also makes renames from using `jupyter_magic` to
`ipython_escape_commands` in various function names.
The mode value is still `Mode::Jupyter` because the escape commands are
part of the IPython syntax but the lexing/parsing is done for a Jupyter
notebook.
### Motivation behind the rename:
* IPython codebase defines it as "EscapeCommand" / "Escape Sequences":
* Escape Sequences:
292e3a2345/IPython/core/inputtransformer2.py (L329-L333)
* Escape command:
292e3a2345/IPython/core/inputtransformer2.py (L410-L411)
* The word "magic" is used mainly for the actual magic commands i.e.,
the ones starting with `%`/`%%`
(https://ipython.readthedocs.io/en/stable/interactive/reference.html#magic-command-system).
So, this avoids any confusion between the Magic token (`%`, `%%`) and
the escape command itself.
## Test Plan
* `cargo test` to make sure all renames are done correctly.
* `grep` for `jupyter_escape`/`magic` to make sure all renames are done
correctly.
## Summary
Per the suggestion in
https://github.com/astral-sh/ruff/discussions/6183, this PR removes
`AsyncWith`, `AsyncFor`, and `AsyncFunctionDef`, replacing them with an
`is_async` field on the non-async variants of those structs. Unlike an
interpreter, we _generally_ have identical handling for these nodes, so
separating them into distinct variants adds complexity from which we
don't really benefit. This can be seen below, where we get to remove a
_ton_ of code related to adding generic `Any*` wrappers, and a ton of
duplicate branches for these cases.
## Test Plan
`cargo test` is unchanged, apart from parser snapshots.
## Summary
Historically, we've stored "qualified names" on our
`BindingKind::Import`, `BindingKind::SubmoduleImport`, and
`BindingKind::ImportFrom` structs. In Ruff, a "qualified name" is a
dot-separated path to a symbol. For example, given `import foo.bar`, the
"qualified name" would be `"foo.bar"`; and given `from foo.bar import
baz`, the "qualified name" would be `foo.bar.baz`.
This PR modifies the `BindingKind` structs to instead store _call paths_
rather than qualified names. So in the examples above, we'd store
`["foo", "bar"]` and `["foo", "bar", "baz"]`. It turns out that this
more efficient given our data access patterns. Namely, we frequently
need to convert the qualified name to a call path (whenever we call
`resolve_call_path`), and it turns out that we do this operation enough
that those conversations show up on benchmarks.
There are a few other advantages to using call paths, rather than
qualified names:
1. The size of `BindingKind` is reduced from 32 to 24 bytes, since we no
longer need to store a `String` (only a boxed slice).
2. All three import types are more consistent, since they now all store
a boxed slice, rather than some storing an `&str` and some storing a
`String` (for `BindingKind::ImportFrom`, we needed to allocate a
`String` to create the qualified name, but the call path is a slice of
static elements that don't require that allocation).
3. A lot of code gets simpler, in part because we now do call path
resolution "earlier". Most notably, for relative imports (`from .foo
import bar`), we store the _resolved_ call path rather than the relative
call path, so the semantic model doesn't have to deal with that
resolution. (See that `resolve_call_path` is simpler, fewer branches,
etc.)
In my testing, this change improves the all-rules benchmark by another
4-5% on top of the improvements mentioned in #6047.
## Summary
Update `F841` autofix to not remove line magic expr
## Test Plan
Added test case for assignment statement with and without type
annotation
fixes: #6116
## Summary
Similar to #6279, moving some helpers onto the struct in the name of
reducing the number of random undiscoverable utilities we have in
`helpers.rs`.
Most of the churn is migrating rules to take `ast::ExprCall` instead of
the spread call arguments.
## Test Plan
`cargo test`
## Summary
This PR removes a now-unnecessary abstraction from `helper.rs`
(`CallArguments`), in favor of adding methods to `Arguments` directly,
which helps with discoverability.
## Summary
This PR boxes the `TypeParams` and `Arguments` fields on the class
definition node. These fields are optional and often emitted, and given
that class definition is our largest enum variant, we pay the cost of
including them for every statement in the AST. Boxing these types
reduces the statement size by 40 bytes, which seems like a good tradeoff
given how infrequently these are accessed.
## Test Plan
Need to benchmark, but no behavior changes.
## Summary
Similar to #6259, this PR adds a `TypeParams` node to the AST, to
capture the list of type parameters with their surrounding brackets.
If a statement lacks type parameters, the `type_params` field will be
`None`.
## Summary
This PR adds a new `Arguments` AST node, which we can use for function
calls and class definitions.
The `Arguments` node spans from the left (open) to right (close)
parentheses inclusive.
In the case of classes, the `Arguments` is an option, to differentiate
between:
```python
# None
class C: ...
# Some, with empty vectors
class C(): ...
```
In this PR, we don't really leverage this change (except that a few
rules get much simpler, since we don't need to lex to find the start and
end ranges of the parentheses, e.g.,
`crates/ruff/src/rules/pyupgrade/rules/lru_cache_without_parameters.rs`,
`crates/ruff/src/rules/pyupgrade/rules/unnecessary_class_parentheses.rs`).
In future PRs, this will be especially helpful for the formatter, since
we can track comments enclosed on the node itself.
## Test Plan
`cargo test`
## Summary
This PR renames...
- `Parameter#arg` to `Parameter#name`
- `ParameterWithDefault#def` to `ParameterWithDefault#parameter` (such
that `ParameterWithDefault` has a `default` and a `parameter`)
## Test Plan
`cargo test`
## Summary
This PR renames a few AST nodes for clarity:
- `Arguments` is now `Parameters`
- `Arg` is now `Parameter`
- `ArgWithDefault` is now `ParameterWithDefault`
For now, the attribute names that reference `Parameters` directly are
changed (e.g., on `StmtFunctionDef`), but the attributes on `Parameters`
itself are not (e.g., `vararg`). We may revisit that decision in the
future.
For context, the AST node formerly known as `Arguments` is used in
function definitions. Formally (outside of the Python context),
"arguments" typically refers to "the values passed to a function", while
"parameters" typically refers to "the variables used in a function
definition". E.g., if you Google "arguments vs parameters", you'll get
some explanation like:
> A parameter is a variable in a function definition. It is a
placeholder and hence does not have a concrete value. An argument is a
value passed during function invocation.
We're thus deviating from Python's nomenclature in favor of a scheme
that we find to be more precise.
## Summary
This PR adds the implementation for the new Jupyter AST nodes i.e.,
`ExprLineMagic` and `StmtLineMagic`.
## Test Plan
Add test cases for `unparse` containing magic commands
resolves: #6087
## Summary
My intuition is that it's faster to do these checks as-needed rather
than allocation new hash maps and vectors for the arguments. (We
typically only query once anyway.)
## Summary
This crate now contains utilities for dealing with trivia more broadly:
whitespace, newlines, "simple" trivia lexing, etc. So renaming it to
reflect its increased responsibilities.
To avoid conflicts, I've also renamed `Token` and `TokenKind` to
`SimpleToken` and `SimpleTokenKind`.
## Summary
Previously, `StmtIf` was defined recursively as
```rust
pub struct StmtIf {
pub range: TextRange,
pub test: Box<Expr>,
pub body: Vec<Stmt>,
pub orelse: Vec<Stmt>,
}
```
Every `elif` was represented as an `orelse` with a single `StmtIf`. This
means that this representation couldn't differentiate between
```python
if cond1:
x = 1
else:
if cond2:
x = 2
```
and
```python
if cond1:
x = 1
elif cond2:
x = 2
```
It also makes many checks harder than they need to be because we have to
recurse just to iterate over an entire if-elif-else and because we're
lacking nodes and ranges on the `elif` and `else` branches.
We change the representation to a flat
```rust
pub struct StmtIf {
pub range: TextRange,
pub test: Box<Expr>,
pub body: Vec<Stmt>,
pub elif_else_clauses: Vec<ElifElseClause>,
}
pub struct ElifElseClause {
pub range: TextRange,
pub test: Option<Expr>,
pub body: Vec<Stmt>,
}
```
where `test: Some(_)` represents an `elif` and `test: None` an else.
This representation is different tradeoff, e.g. we need to allocate the
`Vec<ElifElseClause>`, the `elif`s are now different than the `if`s
(which matters in rules where want to check both `if`s and `elif`s) and
the type system doesn't guarantee that the `test: None` else is actually
last. We're also now a bit more inconsistent since all other `else`,
those from `for`, `while` and `try`, still don't have nodes. With the
new representation some things became easier, e.g. finding the `elif`
token (we can use the start of the `ElifElseClause`) and formatting
comments for if-elif-else (no more dangling comments splitting, we only
have to insert the dangling comment after the colon manually and set
`leading_alternate_branch_comments`, everything else is taken of by
having nodes for each branch and the usual placement.rs fixups).
## Merge Plan
This PR requires coordination between the parser repo and the main ruff
repo. I've split the ruff part, into two stacked PRs which have to be
merged together (only the second one fixes all tests), the first for the
formatter to be reviewed by @michareiser and the second for the linter
to be reviewed by @charliermarsh.
* MH: Review and merge
https://github.com/astral-sh/RustPython-Parser/pull/20
* MH: Review and merge or move later in stack
https://github.com/astral-sh/RustPython-Parser/pull/21
* MH: Review and approve
https://github.com/astral-sh/RustPython-Parser/pull/22
* MH: Review and approve formatter PR
https://github.com/astral-sh/ruff/pull/5459
* CM: Review and approve linter PR
https://github.com/astral-sh/ruff/pull/5460
* Merge linter PR in formatter PR, fix ecosystem checks (ecosystem
checks can't run on the formatter PR and won't run on the linter PR, so
we need to merge them first)
* Merge https://github.com/astral-sh/RustPython-Parser/pull/22
* Create tag in the parser, update linter+formatter PR
* Merge linter+formatter PR https://github.com/astral-sh/ruff/pull/5459
---------
Co-authored-by: Micha Reiser <micha@reiser.io>
## Summary
I'm doing some unrelated profiling, and I noticed that this method is
actually measurable on the CPython benchmark -- it's > 1% of execution
time. We don't need to lex here, we already know the ranges of all
comments, so we can just do a simple binary search for overlap, which
brings the method down to 0%.
## Test Plan
`cargo test`
## Summary
This is really bad PR hygiene, but a mix of: using `Locator`-based fixes
in a few places (in lieu of `Generator`-based fixes), using match syntax
to avoid `.len() == 1` checks, using common helpers in more places, etc.
## Test Plan
`cargo test`
Support for `let…else` formatting was just merged to nightly
(rust-lang/rust#113225). Rerun `cargo fmt` with Rust nightly 2023-07-02
to pick this up. Followup to #939.
Signed-off-by: Anders Kaseorg <andersk@mit.edu>
## Summary
This PR modifies our statement deletion logic to delete any preceding
continuation lines.
For example, given:
```py
x = 1; \
import os
```
We'll now rewrite to:
```py
x = 1;
```
In addition, the logic can now handle multiple preceding continuations
(which is unlikely, but valid).
## Summary
This PR upgrade RustPython to pull in the changes to `Arguments` (zip
defaults with their identifiers) and all the renames to `CmpOp` and
friends.
## Summary
At present, when we store a binding, we include a `TextRange` alongside
it. The `TextRange` _sometimes_ matches the exact range of the
identifier to which the `Binding` is linked, but... not always.
For example, given:
```python
x = 1
```
The binding we create _will_ use the range of `x`, because the left-hand
side is an `Expr::Name`, which has a valid range on it.
However, given:
```python
try:
pass
except ValueError as e:
pass
```
When we create a binding for `e`, we don't have a `TextRange`... The AST
doesn't give us one. So we end up extracting it via lexing.
This PR extends that pattern to the rest of the binding kinds, to ensure
that whenever we create a binding, we always use the range of the bound
name. This leads to better diagnostics in cases like pattern matching,
whereby the diagnostic for "unused variable `x`" here used to include
`*x`, instead of just `x`:
```python
def f(provided: int) -> int:
match provided:
case [_, *x]:
pass
```
This is _also_ required for symbol renames, since we track writes as
bindings -- so we need to know the ranges of the bound symbols.
By storing these bindings precisely, we can also remove the
`binding.trimmed_range` abstraction -- since bindings already use the
"trimmed range".
To implement this behavior, I took some of our existing utilities (like
the code we had for `except ValueError as e` above), migrated them from
a full lexer to a zero-allocation lexer that _only_ identifies
"identifiers", and moved the behavior into a trait, so we can now do
`stmt.identifier(locator)` to get the range for the identifier.
Honestly, we might end up discarding much of this if we decide to put
ranges on all identifiers
(https://github.com/astral-sh/RustPython-Parser/pull/8). But even if we
do, this will _still_ be a good change, because the lexer introduced
here is useful beyond names (e.g., we use it find the `except` keyword
in an exception handler, to find the `else` after a `for` loop, and so
on). So, I'm fine committing this even if we end up changing our minds
about the right approach.
Closes#5090.
## Benchmarks
No significant change, with one statistically significant improvement
(-2.1654% on `linter/all-rules/large/dataset.py`):
```
linter/default-rules/numpy/globals.py
time: [73.922 µs 73.955 µs 73.986 µs]
thrpt: [39.882 MiB/s 39.898 MiB/s 39.916 MiB/s]
change:
time: [-0.5579% -0.4732% -0.3980%] (p = 0.00 < 0.05)
thrpt: [+0.3996% +0.4755% +0.5611%]
Change within noise threshold.
Found 6 outliers among 100 measurements (6.00%)
4 (4.00%) low severe
1 (1.00%) low mild
1 (1.00%) high mild
linter/default-rules/pydantic/types.py
time: [1.4909 ms 1.4917 ms 1.4926 ms]
thrpt: [17.087 MiB/s 17.096 MiB/s 17.106 MiB/s]
change:
time: [+0.2140% +0.2741% +0.3392%] (p = 0.00 < 0.05)
thrpt: [-0.3380% -0.2734% -0.2136%]
Change within noise threshold.
Found 4 outliers among 100 measurements (4.00%)
3 (3.00%) high mild
1 (1.00%) high severe
linter/default-rules/numpy/ctypeslib.py
time: [688.97 µs 691.34 µs 694.15 µs]
thrpt: [23.988 MiB/s 24.085 MiB/s 24.168 MiB/s]
change:
time: [-1.3282% -0.7298% -0.1466%] (p = 0.02 < 0.05)
thrpt: [+0.1468% +0.7351% +1.3461%]
Change within noise threshold.
Found 15 outliers among 100 measurements (15.00%)
1 (1.00%) low mild
2 (2.00%) high mild
12 (12.00%) high severe
linter/default-rules/large/dataset.py
time: [3.3872 ms 3.4032 ms 3.4191 ms]
thrpt: [11.899 MiB/s 11.954 MiB/s 12.011 MiB/s]
change:
time: [-0.6427% -0.2635% +0.0906%] (p = 0.17 > 0.05)
thrpt: [-0.0905% +0.2642% +0.6469%]
No change in performance detected.
Found 20 outliers among 100 measurements (20.00%)
1 (1.00%) low severe
2 (2.00%) low mild
4 (4.00%) high mild
13 (13.00%) high severe
linter/all-rules/numpy/globals.py
time: [148.99 µs 149.21 µs 149.42 µs]
thrpt: [19.748 MiB/s 19.776 MiB/s 19.805 MiB/s]
change:
time: [-0.7340% -0.5068% -0.2778%] (p = 0.00 < 0.05)
thrpt: [+0.2785% +0.5094% +0.7395%]
Change within noise threshold.
Found 2 outliers among 100 measurements (2.00%)
1 (1.00%) low mild
1 (1.00%) high severe
linter/all-rules/pydantic/types.py
time: [3.0362 ms 3.0396 ms 3.0441 ms]
thrpt: [8.3779 MiB/s 8.3903 MiB/s 8.3997 MiB/s]
change:
time: [-0.0957% +0.0618% +0.2125%] (p = 0.45 > 0.05)
thrpt: [-0.2121% -0.0618% +0.0958%]
No change in performance detected.
Found 11 outliers among 100 measurements (11.00%)
1 (1.00%) low severe
3 (3.00%) low mild
5 (5.00%) high mild
2 (2.00%) high severe
linter/all-rules/numpy/ctypeslib.py
time: [1.6879 ms 1.6894 ms 1.6909 ms]
thrpt: [9.8478 MiB/s 9.8562 MiB/s 9.8652 MiB/s]
change:
time: [-0.2279% -0.0888% +0.0436%] (p = 0.18 > 0.05)
thrpt: [-0.0435% +0.0889% +0.2284%]
No change in performance detected.
Found 5 outliers among 100 measurements (5.00%)
4 (4.00%) low mild
1 (1.00%) high severe
linter/all-rules/large/dataset.py
time: [7.1520 ms 7.1586 ms 7.1654 ms]
thrpt: [5.6777 MiB/s 5.6831 MiB/s 5.6883 MiB/s]
change:
time: [-2.5626% -2.1654% -1.7780%] (p = 0.00 < 0.05)
thrpt: [+1.8102% +2.2133% +2.6300%]
Performance has improved.
Found 2 outliers among 100 measurements (2.00%)
1 (1.00%) low mild
1 (1.00%) high mild
```
## Summary
Our current mechanism for handling deletions (e.g., `del x`) is to
remove the symbol from the scope's `bindings` table. This "does the
right thing", in that if we then reference a deleted symbol, we're able
to determine that it's unbound -- but it causes a variety of problems,
mostly in that it makes certain bindings and references unreachable
after-the-fact.
Consider:
```python
x = 1
print(x)
del x
```
If we analyze this code _after_ running the semantic model over the AST,
we'll have no way of knowing that `x` was ever introduced in the scope,
much less that it was bound to a value, read, and then deleted --
because we effectively erased `x` from the model entirely when we hit
the deletion.
In practice, this will make it impossible for us to support local symbol
renames. It also means that certain rules that we want to move out of
the model-building phase and into the "check dead scopes" phase wouldn't
work today, since we'll have lost important information about the source
code.
This PR introduces two new `BindingKind` variants to model deletions:
- `BindingKind::Deletion`, which represents `x = 1; del x`.
- `BindingKind::UnboundException`, which represents:
```python
try:
1 / 0
except Exception as e:
pass
```
In the latter case, `e` gets unbound after the exception handler
(assuming it's triggered), so we want to handle it similarly to a
deletion.
The main challenge here is auditing all of our existing `Binding` and
`Scope` usages to understand whether they need to accommodate deletions
or otherwise behave differently. If you look one commit back on this
branch, you'll see that the code is littered with `NOTE(charlie)`
comments that describe the reasoning behind changing (or not) each of
those call sites. I've also augmented our test suite in preparation for
this change over a few prior PRs.
### Alternatives
As an alternative, I considered introducing a flag to `BindingFlags`,
like `BindingFlags::UNBOUND`, and setting that at the appropriate time.
This turned out to be a much more difficult change, because we tend to
match on `BindingKind` all over the place (e.g., we have a bunch of code
blocks that only run when a `BindingKind` is
`BindingKind::Importation`). As a result, introducing these new
`BindingKind` variants requires only a few changes at the client sites.
Adding a flag would've required a much wider-reaching change.
