Closed rasmushenningsson closed 2 months ago
Oh... yes. I see. I can investigate this. I think I know what is going on.
However it is possible I will want to merge #55 first, depending on how much of a lift this is to fix, as that PR is a pretty substantial change to the implementation and API.
No hurry. It makes sense to merge #55 first.
Note: in the final design of #55, @ericphanson and I eventually opted to always exclude the module information, as it is often considered an implementation detail of a package what module it belongs to.
The new documentation (merged in #58) aims to clearly communicate that the module information is ignored, and how you can opt in to not ignoring it. @rasmushenningsson, if you have an opportunity to take a look through the section on how Function's are hashed, I'd appreciate any feedback regarding whether you think it is sufficient to close #54.
This makes total sense to me. This issue was mainly about consistency. Closes as solved by #58.
It seems like the module information is ignored for
ComposedFunction
s.As expected, these give different hashes:
But these give the same hashes, even though the objects are distinct:
It seems like it would be better if they would return different hashes.