[ruff] Implement check for Decimal called with a float literal (RUF032)

[kbaskett248]

Summary

This PR adds a new check for a Decimal call with a float argument. As described in #11840, this is a problem when precision is important, which is usually why Decimal is used in the first place. To pull an example by @mayanyax from the issue:

>>> Decimal(0.3) == Decimal(0.1) + Decimal(0.1) + Decimal(0.1)
False
>>> Decimal('0.3') == Decimal('0.1') + Decimal('0.1') + Decimal('0.1')
True

The check supports positive and negative floats. It also includes a fix that wraps the float in quotes.

Test Plan

Test cases were added for the new rule.

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

ℹ️ ecosystem check detected linter changes. (+25 -0 violations, +0 -0 fixes in 4 projects; 50 projects unchanged)

RasaHQ/rasa (+14 -0 violations, +0 -0 fixes)

ruff check --no-cache --exit-zero --ignore RUF9 --output-format concise --preview

+ tests/core/test_utils.py:41:29: RUF032 `Decimal()` called with float literal argument
+ tests/core/test_utils.py:49:28: RUF032 `Decimal()` called with float literal argument
+ tests/core/test_utils.py:49:52: RUF032 `Decimal()` called with float literal argument
+ tests/core/test_utils.py:49:77: RUF032 `Decimal()` called with float literal argument
+ tests/core/test_utils.py:55:52: RUF032 `Decimal()` called with float literal argument
+ tests/core/test_utils.py:55:76: RUF032 `Decimal()` called with float literal argument
+ tests/core/test_utils.py:56:40: RUF032 `Decimal()` called with float literal argument
+ tests/core/test_utils.py:71:18: RUF032 `Decimal()` called with float literal argument
+ tests/core/test_utils.py:77:19: RUF032 `Decimal()` called with float literal argument
+ tests/core/test_utils.py:77:33: RUF032 `Decimal()` called with float literal argument
+ tests/core/test_utils.py:77:48: RUF032 `Decimal()` called with float literal argument
+ tests/core/test_utils.py:80:43: RUF032 `Decimal()` called with float literal argument
+ tests/core/test_utils.py:80:57: RUF032 `Decimal()` called with float literal argument
+ tests/core/test_utils.py:80:82: RUF032 `Decimal()` called with float literal argument

apache/airflow (+3 -0 violations, +0 -0 fixes)

ruff check --no-cache --exit-zero --ignore RUF9 --output-format concise --preview --select ALL

+ tests/providers/amazon/aws/transfers/test_dynamodb_to_s3.py:94:21: RUF032 `Decimal()` called with float literal argument
+ tests/providers/google/cloud/transfers/test_cassandra_to_gcs.py:132:27: RUF032 `Decimal()` called with float literal argument
+ tests/www/views/test_views_trigger_dag.py:104:28: RUF032 `Decimal()` called with float literal argument

ibis-project/ibis (+1 -0 violations, +0 -0 fixes)

ruff check --no-cache --exit-zero --ignore RUF9 --output-format concise --preview

+ ibis/expr/datatypes/tests/test_value.py:48:26: RUF032 `Decimal()` called with float literal argument

pandas-dev/pandas (+7 -0 violations, +0 -0 fixes)

ruff check --no-cache --exit-zero --ignore RUF9 --output-format concise --preview

+ pandas/tests/dtypes/cast/test_downcast.py:36:39: RUF032 `Decimal()` called with float literal argument
+ pandas/tests/dtypes/cast/test_downcast.py:38:39: RUF032 `Decimal()` called with float literal argument
+ pandas/tests/dtypes/test_missing.py:326:21: RUF032 `Decimal()` called with float literal argument
+ pandas/tests/tools/test_to_numeric.py:195:40: RUF032 `Decimal()` called with float literal argument
+ pandas/tests/tools/test_to_numeric.py:210:31: RUF032 `Decimal()` called with float literal argument
+ pandas/tests/tools/test_to_numeric.py:210:60: RUF032 `Decimal()` called with float literal argument
+ pandas/tests/tools/test_to_numeric.py:213:37: RUF032 `Decimal()` called with float literal argument

Changes by rule (1 rules affected)

code	total	+ violation	- violation	+ fix	- fix
RUF032	25	25	0	0	0

[MichaReiser]

Nice. I'd love to get some input from @AlexWaygood but this looks great

[MichaReiser]

What's your reasoning for making it an unsafe edit. Is it because it changes the precision of the values and, therefore, could result in runtime changes?

[AlexWaygood]

I think an unsafe edit is appropriate here... in most cases, you'll be changing the underlying value of the Decimal object being constructed. There might be code elsewhere that implicitly depended on the previous, "wrong" value that was caused by constructing the Decimal instance from the float, so it's entirely possible that fixing this error could unexpectedly cause your code to break.

@kbaskett248, it would be great if you could add one or two sentences to the docs explaining why this is an unsafe fix, like how we do with other rules.

[kbaskett248]

Honestly it was mostly unfamiliarity with the apis, but @AlexWaygood brings up a good point.

[codspeed-hq]

CodSpeed Performance Report

Merging #12909 will not alter performance

_{Comparing kbaskett248:rule/float-literal-decimal (9a8fd54) with main (65de8f2)}

Summary

✅ 32 untouched benchmarks

[AlexWaygood]

Merging #12909 will degrade performances by 5.27%

You can ignore that, it's been super flaky recently

[kbaskett248]

Thanks for your help with this @AlexWaygood and @MichaReiser! Any other updates you'd like me to make?

[AlexWaygood]

One last thing. Here we should use the quote style that the user is generally using for the rest of their file, which won't necessarily be double quotes if they're not using autoformatting. We can do this with the checker.stylist():

Suggested change

	fn fix_float_literal(range: TextRange, float_literal: &str) -> Fix {
	let content = format!("\"{float_literal}\"");
	Fix::unsafe_edit(Edit::range_replacement(content, range))
	fn fix_float_literal(range: TextRange, float_literal: &str, stylist: &Stylist) -> Fix {
	let quote = stylist.quote();
	let content = format!("{quote}{float_literal}{quote}");
	Fix::unsafe_edit(Edit::range_replacement(content, range))

You'll just want to pass checker.stylist() into this function at the callsite, and you'll want to add use ruff_python_codegen::Stylist; to the top of the file so that you can use it here in the type signature

[kbaskett248]

Thank you both for your help getting this in!

[AlexWaygood]

Thank you both for your help getting this in!

Thanks for the great contribution!