allow is and is not for type comparisons (E721)
#1086
Conversation
| @@ -134,8 +134,10 @@ | |||
| r'\s*(?(1)|(None|False|True))\b') | |||
| COMPARE_NEGATIVE_REGEX = re.compile(r'\b(?<!is\s)(not)\s+[^][)(}{ ]+\s+' | |||
| r'(in|is)\s') | |||
| COMPARE_TYPE_REGEX = re.compile(r'(?:[=!]=|is(?:\s+not)?)\s+type(?:s.\w+Type' | |||
There was a problem hiding this comment.
I can't find the rationale for removing types.*Type comparisons in the linked issues. As far as I can tell, those should still be included in the checks, shouldn't they?
There was a problem hiding this comment.
I think those got removed because the name types isn't necessarily the stdlib types module -- not 100% sure though
There was a problem hiding this comment.
the original patch is here: #1041 -- this is just a revert-revert of that with an additional commit on top
There was a problem hiding this comment.
I suppose the precise reasoning is the following, then:
This check also breaks many ligimitate cases (such as having your own types module with an Enum class for instance, and comparing a value to the enum member).
which I can relate to. You can, of course, also shadow the built-in type function, but that's just not something people do in general and having a custom types module is much more likely, although also ending your enum in Type is most likely less common.
asottile
force-pushed
the
E721-allow-is-v2
branch
from
4c8dd67 to
571d3ec
Compare
asottile
force-pushed
the
E721-allow-is-v2
branch
from
571d3ec to
c30c06b
Compare
asottile
force-pushed
the
E721-allow-is-v2
branch
from
c30c06b to
ac223bd
Compare
see #1084 and #1041 for more details
resolves #1084
resolves #830