Consolidate implementation of calculate_diff

[Swatinem]

This cleans up the duplicated implementations of Filtered/Report.calculate_diff and moves it to a new well typed file.

---

I’m a bit surprised that this does not yield any speedup in the diff calculation benchmarks (compared to #525).

I think the difference between the base and FilteredReport is based on the actual filtering and line-mapping code that happens, instead of the different and less optimized version that was previously there.

                                          Benchmark Results                                          
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━┳━━━━━━━━┓
┃                                       Benchmark ┃   Time (best) ┃ Rel. StdDev ┃ Run time ┃  Iters ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━╇━━━━━━━━┩
│            test_report_diff_calculation[Report] │         803ns │        0.9% │    2.92s │ 15,975 │
│    test_report_diff_calculation[FilteredReport] │       1,886ns │        1.1% │    2.92s │ 10,437 │
└─────────────────────────────────────────────────┴───────────────┴─────────────┴──────────┴────────┘

[codspeed-hq]

CodSpeed Performance Report

Merging #528 will create unknown performance changes

_{Comparing swatinem/unify-calculate-diff (ca08e87) with main (0222015)}

Summary

⚠️ No benchmarks were detected in both the base of the PR and the PR.\

[giovanni-guidini]

LGTM. Thanks for adding more types to this part of the code 🥳

As you mentioned no speed improvement, but in the comparison the stddev for the Report class was greatly reduced... that might be a win? Having a more bounded code I guess. The speed report doesn't really report quantiles for us to know, though.

I wonder if using @dataclass instead of TypedDict could improve speed. But my guess is that it would be worse (cause you need to transform from dict into a class instance).

[giovanni-guidini]

A comment here like # (base_start_line, base_length, head_start_line, head_length) would be useful for people that don't know what the numbers represent

Or breaking up the header further, but might be overkill

[giovanni-guidini]

do we know what the possible types are?

[joseph-sentry]

also lgtm:

to answer one of @giovanni-guidini's questions from above:

shared/shared/torngit/base.py

Line 142 in 3bcf3c9

def diff_to_json(self, diff):

I got confused because i was looking at the response schemas from the Github API and I couldn't find "type" or "segments"

[Swatinem]

Thanks for the suggestions. I improved the types and added doc comments.

On that note, I just hate how python doc comments come after the item they are documenting. This is so unintuitive compared to pretty much every other language that does it differently.

[overwatch-beta]

✅ Sentry found no issues in your recent changes ✅