test: lower low boundary for accuracy in test_calculate_context_similarity_on_non_matching_contexts

[ZanSara]

Related Issues

• failing main: https://github.com/deepset-ai/haystack/runs/8300467040?check_suite_focus=true

Proposed Changes:

• Lower low boundary for accuracy in test_calculate_context_similarity_on_non_matching_contexts from 0.99 to 0.98

How did you test it?

• The test passes locally (with accuracy=1) and on CI (with accuracy=0.9860646599777034)

Notes for the reviewer

• I attempted to use a more refined solution with pytest.approx, but that works with equality checks only, not with larger-than checks.

Checklist

• I have read the contributors guidelines and the code of conduct
• I have updated the related issue with new insights and changes
• I added tests that demonstrate the correct behavior of the change
• I've used the conventional commit convention for my PR title
• I documented my code
• I ran pre-commit hooks and fixed any issue

[tstadel]

For a quick fix of main this is fine.
However, that's interesting. When did that start to fail?
I suppose there has been a new rapidfuzz version, that is causing this. Let's make sure this is a feature and not a bug

[tstadel]

Ok, in the failing tests we're using rapidfuzz 2.8.0. Last passing test had rapidfuzz 2.6.1.

[tstadel]

As this is clearly caused by the new rapidfuzz version 2.8.0, let's better add a version pin restriction of <2.8.0. We also need to create an issue at https://github.com/maxbachmann/RapidFuzz. I'll take care of this. After the issue has been resolved or it turns out this is a feature we can remove the version restriction.

[tstadel]

Ok, I looked into the changes of rapidfuzz 2.8.0. This is actually a feature and should enable us to get rid of the code I wrote to boost split overlaps (calculate_context_similarity's boost_split_overlap option). I created an issue for that as it would need some tinkering regarding a proper threshold: #3202
Let's reference this issue in the pin comment and merge :-)

[tstadel]

Just saw that this dependency entry is a duplicate. Let's merge it with the existing one and add a comment about the issue