[pylint] Implement consider-using-assignment-expr (R6103)
#13196
Conversation
CodSpeed Performance ReportMerging #13196 will not alter performanceComparing Summary
|
vincevannoort
changed the title
R6103R6103
vincevannoort
force-pushed
the
consider-using-assignment-expr
branch
from
b422d5c
to
dbefe63
Compare
|
| code | total | + violation | - violation | + fix | - fix |
|---|---|---|---|---|---|
| PLR6103 | 5110 | 5110 | 0 | 0 | 0 |
vincevannoort
force-pushed
the
consider-using-assignment-expr
branch
from
84c2180
to
4d891db
Compare
| if checker.enabled(Rule::UnnecessaryAssignment) { | ||
| pylint::rules::unnecessary_assignment(checker, if_); | ||
| } |
There was a problem hiding this comment.
I applied this rule by checking whether the previous statement from a IfStmt is an AssignStmt, however I am wondering if it is more desirable to do check if the next statement of a AssignStmt is an IfStmt?
There was a problem hiding this comment.
What benefits do you see in testing the next statement after an AssignStmt.
My intuition here is that there are probably more assignment than if statements. Therefore, running the rules on if nodes might overall be faster?
There was a problem hiding this comment.
I agree that assignments are far more common than if statements.
I think the answer depends on the costs for checking previous_statement and next_statement.
My thought here is that retrieving the previous statement using the newly added previous_statement might be more expensive because it has to iterate over all previous statements using previous_statements to find the previous statement (if there is a better way, please let me know).
While checking an assignment, then checking the next_statement might be a cheap operation.
Do you have any idea? If they have equal cost I think the current implementation is fine. 😄
vincevannoort
changed the title
R6103consider-using-assignment-expr (R6103)
vincevannoort
changed the title
consider-using-assignment-expr (R6103)consider-using-assignment-expr (R6103)
| pub fn previous_statement(&self, stmt: &'a Stmt) -> Option<&Stmt> { | ||
| self.previous_statements(stmt)?.next() | ||
| } |
There was a problem hiding this comment.
I needed something like this function, not sure if there is an existing better way?
There was a problem hiding this comment.
That makes sense. I think we can make it more efficient with #13895
which reduces an extra find and collect
vincevannoort
force-pushed
the
consider-using-assignment-expr
branch
from
07d2997
to
964c3d6
Compare
MichaReiser
added
rule
There was a problem hiding this comment.
Thanks for this contribution. My plane is about to land. I've to finish the review at a later time.
I only had enough time to quickly glance over the Rust code. We should look into removing the many node.clone() calls because it is fairly expensive to clone nodes and probably unnecessary. Let me know if you need some guidance on how to remove the clone calls (It probably requires adding some lifetimes)
Regarding the rule's naming.
- I did a quick search to see how we referred to
:=in other rules. There are not many usages butnamed expression (walrus operator)is the most common form. - The rule name seems too generic to me and its name is very similar to
unnecessary-assign(which we should rename to `unnecessary-assignment). Reading through the examples the rule mainly is about assigning a value that is then only used in an if condition. I need to think a bit more about what a good rule name could be. Maybe you have an idea?
| bad5 = ( | ||
| 'example', | ||
| 'example', | ||
| 'example', | ||
| 'example', | ||
| 'example', | ||
| 'example', | ||
| 'example', | ||
| 'example', | ||
| 'example', | ||
| 'example', | ||
| ) |
There was a problem hiding this comment.
I like the example because it shows a potentially controversial use case. I would probably prefer the assignment to keep the if smaller.
crates/ruff_linter/resources/test/fixtures/pylint/unnecessary_assignment.py
Show resolved
Hide resolved
| if checker.enabled(Rule::UnnecessaryAssignment) { | ||
| pylint::rules::unnecessary_assignment(checker, if_); | ||
| } |
There was a problem hiding this comment.
What benefits do you see in testing the next statement after an AssignStmt.
My intuition here is that there are probably more assignment than if statements. Therefore, running the rules on if nodes might overall be faster?
crates/ruff_linter/src/rules/pylint/rules/unnecessary_assignment.rs
Outdated
Show resolved
Hide resolved
crates/ruff_linter/src/rules/pylint/rules/unnecessary_assignment.rs
Outdated
Show resolved
Hide resolved
Thanks for the review, much appreciated! 👍
I have tried and was able to remove almost all clone calls, except a few which I think are needed for returning the diagnostic. Could you take a look at the remaining ones and see if any of the 3 can still be removed?
I agree, here are some possible options:
These seem close to what the lint is trying to prevent, do you have other ideas in mind? |
|
Hey @MichaReiser, once you have the time, would you mind giving this pull request another look? 😄 |
|
Hey @MichaReiser, could you or someone else from the team have a look? 😄 |
There was a problem hiding this comment.
Thanks for the ping and sorry for the late review. I pushed a few smaller refactors to avoid unnecessary collects.
I created a PR that should allow us to implement a more efficient previous_statement here.
I think the rule has to become cleverer, at least when we want to support handling elif cases because today's implementation can result in changes that fail at runtime.
| }; | ||
| } | ||
|
|
||
| // case - elif else clauses (`elif test1:`) |
There was a problem hiding this comment.
Is there a reason why compare expressions aren't handled inside elif_else clauses?
| errors.extend( | ||
| stmt.elif_else_clauses | ||
| .iter() | ||
| .filter(|elif_else_clause| elif_else_clause.test.is_some()) | ||
| .filter_map(|elif_else_clause| { | ||
| let elif_check = elif_else_clause.test.as_ref().unwrap(); | ||
| find_assignment_before_if_stmt(semantic, elif_check, elif_check) | ||
| }) | ||
| .collect::<Vec<AssignmentBeforeIfStmt>>(), |
There was a problem hiding this comment.
Changing assignments to walrus operators in elif else branches is semantically incorrect if the variable is used afterwards
>>> if True: ...
... elif x :=10: ...
...
Ellipsis
>>> x
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
NameError: name 'x' is not defined| error: AssignmentBeforeIfStmt, | ||
| ) -> Diagnostic { | ||
| let (origin, expr_name, assignment) = error; | ||
| let assignment_expr = generator.expr(&assignment.value); |
There was a problem hiding this comment.
What's the motivation for calling into the generator here? We try to avoid using the generator because it removes comments. Could we instead take the assignment value as it is in the source (using locator)? Note: We have to be careful about parenthesized expressions.
| pub fn previous_statement(&self, stmt: &'a Stmt) -> Option<&Stmt> { | ||
| self.previous_statements(stmt)?.next() | ||
| } |
There was a problem hiding this comment.
That makes sense. I think we can make it more efficient with #13895
which reduces an extra find and collect
| bad7 = 'example' | ||
| if bad7 == 'something': # [consider-using-assignment-expr] | ||
| pass | ||
| elif bad7 == 'something else': | ||
| pass |
There was a problem hiding this comment.
This is an interesting example and possibly controversial. I would prefer the existing solution because the assignment in the if is very subtle and I can see how it can be confusing when trying to figure out what the value of bad7 is in the elif branch
|
Thanks for the review @MichaReiser 👍 , just a heads up: I will be travelling without my laptop for the coming 5 weeks, so will only get back to pull request this after. |
Summary
This pull request implements the
R6103pylint rule: pylint documentation.It checks assignments which are directly followed by if statements using that expression:
And suggest to use the
:=operator toTest Plan
I have added test cases, and checked some of the
ruff ecosystemresults.