Upgrade pylint to 1.8.2

[OttoWinter]

Description:

This PR upgrades pylint to 1.8.2 and fixes all resulting error checks.

Related issue (if applicable): #7108

bad-whitespace

PEP8 now recommends whitespace between an argument annotation and a default value around the = sign (see pylint-dev/pylint#238)

When combining an argument annotation with a default value, use spaces around the = sign (but only for those arguments that have both an annotation and a default).
Yes:

def munge(sep: AnyStr = None): ...
def munge(input: AnyStr, sep: AnyStr = None, limit=1000): ...

No:

def munge(input: AnyStr=None): ...
def munge(input: AnyStr, limit = 1000): ...

The code update however causes pylint 1.6.5 to complain about the version with the whitespace (i.e. the correct version) - so either the commit has to be reverted (and bad-whitespace needs to be disabled for 1.8.2) or bad-whitespace is disabled for pylint 1.6.5, which could potentially be a bad idea.

arguments-differ

Makes sure that the arguments for a overridden method match the super class. Also checks whether the names of the arguments match, which seems a bit verbose to me... Anyway, there's one problem though: DeviceScanner::get_device_name(self, mac: str) -> str should, as far as I understand, return the device name given a mac. Many implementations of this however don't use MAC-Addresses and call this argument device, thus triggering the pylint warning. I don't know what to do about this.

inconsistent-return-statements

Makes sure all return statements in a function either return something (e.g. None) or nothing. For example:

def foo():
    if bar():
        return 42
    else:
        return

would trigger this check. However, as described in pylint-dev/pylint#1782, this currently doesn't handle raise - and this erroneous flag is quite common in Home Assistant. Next, I believe correcting this would best be done together with static type checking updates.

All Changes

• len-as-condition
• no-else-return
• bad-whitespace (see comment)
• too-many-nested-blocks
• logging-not-lazy
• stop-iteration-return (locally disabled due to bug)
• useless-super-delegation
• trailing-comma-tuple (see commit message)
• redefine-argument-from-local
• consider-using-enumerate (see the one case, the length is already checked before, so that works)
• wrong-import-order (tried to be as consistent as possible)
• arguments-differ (device_tracker)
• no-member
• signatures-differ

Checklist:

• The code change is tested and works locally.
• Local tests with tox run successfully. Your PR cannot be merged unless tests pass

[houndci-bot]

line too long (80 > 79 characters)

[OttoWinter]

I believe this should be correct...

[balloob]

yes

[OttoWinter]

TEMP_CELSIUS is a constant. However appending constants is essentially the same as appending variables in terms of speed (*except for potential compiler optimizations, though the speed impact is negligible)

[MartinHjelmare]

The latest pylint version is 1.8.2. Any reason not updating to that instead of 1.8.1?

[OttoWinter]

Oh sorry, I wrote 1.8.1 in the description, but I'm running 1.8.2 locally - will update the description. Running with 1.8.1 and 1.8.2 seems to give the same results though.

[balloob]

Fine with for device tracker to rename the official argument to device instead of mac

[balloob]

That's too bad :(

[OttoWinter]

What about bad-whitespace? With the fix, pylint 1.6.5 now (incorrectly) complains itself about a bad-whitespace. So should we either

• Revert that commit and re-apply it later when pylint is actually upgraded
• or disable bad-whitespace by default (potentially a bad idea)?

Additionally, there's another issue that I forgot to mention: Some new checks in pylint 1.8.2, namely c-extension-no-member, stop-iteration-return and useless-super-delegation, don't exist in 1.6.5 - so 1.6.5 complains about a bad-option-value. I'm assuming it's okay to just ignore that locally.

[MartinHjelmare]

Are there too many changes needed to also upgrade pylint to 1.8.2 in this PR?

[OttoWinter]

It would work under 1.8.2, but I tried following the advice in this comment

[MartinHjelmare]

That comment is about Python versions. I don't see anything stopping us from upgrading to pylint 1.8.2 and run that on the Travis pylint tox job.