use _compression.BaseStream for lzma.LZMAFile #11166
Conversation
|
According to mypy_primer, this change has no effect on the checked open source code. 🤖🎉 |
| @@ -104,7 +104,7 @@ class LZMACompressor: | |||
|
|
|||
| class LZMAError(Exception): ... | |||
|
|
|||
| class LZMAFile(io.BufferedIOBase, IO[bytes]): # type: ignore[misc] # incompatible definitions of writelines in the base classes | |||
| class LZMAFile(BaseStream, IO[bytes]): # type: ignore[misc] # incompatible definitions of writelines in the base classes | |||
There was a problem hiding this comment.
Unrelated to this MR:
I stared at this for a while, and I can't quite figure out what the purpose of inheriting from IO[bytes] here is. I found the discussion of typing.IO here python/typing#829 which seems to say typeshed's policy is to avoid the use of typing.IO.
I guess the concern is that because typing.IO isn't actually a Protocol, functions typed to accept a typing.IO argument require classes in the standard library to be subclasses of typing.IO to work as expected?
There was a problem hiding this comment.
Yes, yes, and yes :) Most of our uses of IO are legacy uses from years back, before we came up with the policy of mostly using narrower protocols for I/O related things. Nowadays, we recommend that our users don't use typing.IO, and we wouldn't usually start using it in a new class we were adding stubs for.
We could possibly think about changing this -- but, as you say, it comes at the risk of breaking user code that's already using typing.IO. So, best to consider it separately in another PR, if at all.
related to #3968