Change codegen to treat typing.Union[Foo, NoneType] and typing.Optional[Foo] as the same #508
Conversation
facebook-github-bot
added
the
CLA Signed
…al[Foo] as the same
| if typestr.startswith("<class '") and typestr.endswith("'>"): | ||
| typestr = typestr[8:-2] | ||
| typestr = re.sub(CLASS_RE, r"\1", typestr) | ||
| typestr = re.sub(OPTIONAL_RE, r"typing.Optional[\1]", typestr) |
There was a problem hiding this comment.
This won't work for Union[A, B, NoneType] or Union[A, None]. While this might not cause an issue today, I don't want to be on the debugging end if it pops up in the future :)
There was a problem hiding this comment.
Why don't it work for Union[A, B, NoneType], my regex rejects that.
I'm not sure I understand the difference between Union[A, NoneType] and Union[A, None]
There was a problem hiding this comment.
Hm, fair enough. apparently Option[Union[A, B]] always gets flattened to Union[A, B, NoneType] by dataclasses, so it doesn't even matter.
| if typestr.startswith("<class '") and typestr.endswith("'>"): | ||
| typestr = typestr[8:-2] | ||
| typestr = re.sub(CLASS_RE, r"\1", typestr) | ||
| typestr = re.sub(OPTIONAL_RE, r"typing.Optional[\1]", typestr) |
There was a problem hiding this comment.
Hm, fair enough. apparently Option[Union[A, B]] always gets flattened to Union[A, B, NoneType] by dataclasses, so it doesn't even matter.
Summary
This is a fix for #467. As per this comment, the type of optional fields in dataclasses has changed in python 3.9.
This change makes the codegen rewrite the old type to the new type, so it works the same in all versions.
Test Plan
Codegen tests