RFE: Warn about implicit Unicode conversions which could raise an exception

[Herst]

What I'm looking look for is basically a solution to the following SO question concerning problematic implicit conversions into unicode (in Python 2, I guess in Python 3 the equivalent problem with bytes to string conversions exists).

Summary of the SO question: I have a large existing projects where people weren't thinking much about Unicode and happily mixing encoded strs and unicode together without thinking twice about telling the decoder what the encoding used is and what to do with byte sequences which can't be decoded. Apparently I'm not the only one with the problem if you do a SO search or search for the horrible hack I mentioned in the SO question.

Now it would be great if e.g. there was a mypy parameter with which all implicit conversions into unicode (which also means that the programmer didn't explicitly state any encoding) would raise a warning, including the examples from the SO post:

# potentially problematic cases if someStr can also have non-ASCII characters in it
u"foo" + someStr
u"foo{}".format(someStr)

Note that I don't think that mypy should be making guesses about the content of the strings. Also, this RFE is probably related to #1141.

/edit: I just noticed that in python/typing#208 such solutions were already being discussed. Sorry for not noticing it earlier, think of this post here as an argument in favor of the strict approach then.

[gvanrossum]

Thanks! I think we have enough issues open for this topic already (you found both of them) so I'm going to close this one. We're going slowly here but I have a good hope that we'll end up in a good spot with support for your use case.