printf-string-formatting (UP031) - fix should f-string instead of str.format

[DetachHead]

when both UP031 and UP032 are enabled, the fix for UP031 causes UP032:

foo = "foo"
"%s" % (foo,) # error: UP031

after applying the quick fix:

foo = "foo"
"{}".format(foo) # error: UP032

https://play.ruff.rs/c4a825b4-12e5-4479-9167-c0f4388a17c2

this means i have to apply 2 fixes to get to the correct code. it would be nice if UP031 suggested f strings instead.

[dhruvmanila]

We iteratively apply fixes until the source code is stable so if you've select both UP031 and UP032, the final diff should be the f-string code:

$ ruff check --unsafe-fixes --fix --no-cache --diff --select=UP031,UP032 src/autofix.py
--- src/autofix.py
+++ src/autofix.py
@@ -1,2 +1,2 @@
 foo = "foo"
-"%s" % (foo,)
+f"{foo}"

Would fix 2 errors.

Can you tell me how are you running ruff?

[DetachHead]

i meant when applying the quick fix from vscode (or in this case the playground):

that only applies one quick fix at a time. IMO, since f strings are the preferred method of interpolating strings, i think UP031's message and fix should be to change it to an f string

[dhruvmanila]

I think astral-sh/ruff-vscode#451 is related.

For this specific case, maybe we could prioritize UP032 if both UP031 and UP032 are enabled similar to what @charliermarsh did in #11173.

[dhruvmanila]

Oh, but that might not be possible because UP032 isn't triggered until UP031 is fixed.

[charliermarsh]

Yeah this is roughly by design though I understand why it's not ideal in an editor. IIRC we only support translating % to .format, and .format to f-string.

[autinerd]

Maybe it would be a good idea to merge UP031 and UP032. I don't know if ruff has a minimum version, but f-strings exist since Python 3.6 🤔

[Avasam]

@autinerd UP031 and UP032 have different concerns. .format can still be used to avoid very long lines or multiline string concatenation (styling preferences).