feat: Change fractional custom op from percentage-based to relative weighting.

[aepfli]

relates: open-feature/flagd#1282

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.45%. Comparing base (d8e10c7) to head (992b7b9).
Report is 34 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #91      +/-   ##
==========================================
+ Coverage   90.55%   92.45%   +1.90%     
==========================================
  Files           8       12       +4     
  Lines         180      464     +284     
==========================================
+ Hits          163      429     +266     
- Misses         17       35      +18     

Flag	Coverage Δ
unittests	?

Flags with carried forward coverage won't be shown. Click here to find out more.

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

[matthewelwell]

I've added a few styling related comments, but this PR will also need additional test coverage.

I also find the logic quite difficult to follow. Can we try to avoid multiple nested if statements? And can we perhaps make the error logs more descriptive about why they are error cases?

[aepfli]

I've added a few styling related comments, but this PR will also need additional test coverage.

I also find the logic quite difficult to follow. Can we try to avoid multiple nested if statements? And can we perhaps make the error logs more descriptive about why they are error cases?

I am working on adding further tests to the test harness. But i can also try to add here more tests. python, as you might have guessed, is not my strong suit. but i will try to adapt

[aepfli]

the python sdk only logs and error and continues its targeting evaluation - compared to the other sdks, eg java or js - is this a unintended difference between those, should we also unify this, and instead of logging an error raise an exception?

[Kavindu-Dodan]

the python sdk only logs and error and continues its targeting evaluation - compared to the other sdks, eg java or js - is this a unintended difference between those, should we also unify this, and instead of logging an error raise an exception?

+1 for this proposal. I had the same doubt when reviewing the PR. To be consistent, yes we should raise exceptions

[gruebel]

nice work, added just some minor comments

[gruebel]

thanks, looks good 🍻

[gruebel]

@beeme1mr do you have any idea about the failing uploads to codecov? I thought we fixed that issue or is it something new?

[beeme1mr]

@beeme1mr do you have any idea about the failing uploads to codecov? I thought we fixed that issue or is it something new?

I'm not sure what's wrong. The logs indicate that the codecov secret isn't set but the configuration looks correct and it was working last week.

[aepfli]

Any help for codecov is highly appreciated - it seems that this is the only reason the builds are failing

[beeme1mr]

@aepfli could you please rebase? I think I've resolved the CodeCov issue with this PR.

[aepfli]

done i am not sure if anything else is missing here ;) but please let me know