if sys.version_info blocks without an else get rewritten via the pre-commit hook about pyupgrade

[asottile]

this was intentionally changed at some point. there are a few cases where it still leaves an else less one but when it deems it safe it can

[jvavrek]

this was intentionally changed at some point. there are a few cases where it still leaves an else less one but when it deems it safe it can

Is it intentional that the behavior differs between the pre-commit hook and the command line, though?

Similarly, is it intentional/safe that it deletes the next comment?

[asottile]

you are comparing 3.3.1 vs 3.15.0 what do you expect?

[jvavrek]

you are comparing 3.3.1 vs 3.15.0 what do you expect?

Ok, if I update my local pyupgrade to 3.15.0, I get consistent behavior between the two---both delete the if block. However I still submit that there are two issues:

1. The documentation is out of date
2. pyupgrade incorrectly deletes the next comment after the if sys.version_info block

[asottile]

the end of the block is not well defined when there's trialing comments -- pyupgrade picks the approach based on the information the tokenization hands it -- there isn't a correct choice because people do all sorts of flavors of:

if ...:
    code
    # comment

etc.

the docs aren't really out of date since there are still cases where it will leave an elseless branch alone -- but enumerating them in the docs isn't useful

[asottile]

the case the docs is specifically referring to (where it will introduce a syntax error) is:

if True:
    if sys.version_info ...:
        ...

[jvavrek]

Fair points, thanks.