fix: tokenization of ArgsKwargsPackedFunction

[pfackeldey]

Fixes #553.

Now, the token is generated deterministically and thus the dak_cache is hit correctly for multiple repetitive invocations of dak.map_partitions with the same arguments.

Unfortunately I couldn't use __dask_tokenize__ because one needs to repack the args and kwargs correctly.

Not sure how much of a performance gain there is, would be nice to measure it somehow...

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.49%. Comparing base (8cb8994) to head (d68b470).
Report is 194 commits behind head on main.

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #555      +/-   ##
==========================================
- Coverage   93.06%   92.49%   -0.58%     
==========================================
  Files          23       22       -1     
  Lines        3290     3439     +149     
==========================================
+ Hits         3062     3181     +119     
- Misses        228      258      +30     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[pfackeldey]

@lgray do you want to add something else here? I would add a test tomorrow, and then the PR is ready from my side 👍

[pfackeldey]

This one is ready from my side @martindurant and @lgray 👍

[lgray]

@pfackeldey with #558 you'll need to map_partitions a function that has structured inputs to get it to pack it into ArgsKwargsPackedFunction. Just want to make sure your test stays pertinent!

[martindurant]

+1

[lgray]

i.e. you'll have to pass a dict or something like that which contains dask collections.

[pfackeldey]

I hope this is now correct @lgray and @martindurant

edit: I don't know what the failing test is about, but I don't think it's related to my changes? I'm confused...

[lgray]

Actually @martindurant I'm going to revert that change. Dealing with all the cases is worse than the string manipulation.

[martindurant]

Sorry :|

[lgray]

that one test is weirdly flakey...