Move standard_tracer's subscription observer to cpp

[frmdstryr]

Moves the subscription observer in standard_tracer.py to cpp.

This gives a considerable performance increase if there are a lot of changes going on (~60%) but for most code it will likely only be a small improvement. My application sped up about 30% (even after fixing the original problem).

Ref #561

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 70.91%. Comparing base (3bcd2a4) to head (63507e2).
Report is 27 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #562      +/-   ##
==========================================
- Coverage   70.92%   70.91%   -0.01%     
==========================================
  Files         287      287              
  Lines       25710    25698      -12     
  Branches     3643     3641       -2     
==========================================
- Hits        18234    18224      -10     
  Misses       6385     6385              
+ Partials     1091     1089       -2     

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[MatthieuDartiailh]

I made several suggestions to simplify your handling of reference counts. I tested locally and it works but feel free to give it another spin.

Could you also add a changelog entry ?

[frmdstryr]

Thanks for the review.

I tried moving the whole standard_tracer to cpp here but this branch actually faster somehow so I think it's better to just go with this branch as it's also much simpler.

[MatthieuDartiailh]

Can you rebase on main to get the CI to pass ?

[MatthieuDartiailh]

I tried moving the whole standard_tracer to cpp here but this branch actually faster somehow so I think it's better to just go with this branch as it's also much simpler.

This is unexpected but I will take your word for it. No reason to add more cpp if there is no real gain.

[frmdstryr]

I was able to get the poor example in #561 even faster by moving all the PyImport_ImportModule into the modexec (from ~16sec to ~12sec).

[sccolbert]

I’d like to review the c++ code before this is merged.

…

On Tue, Jan 7, 2025 at 13:33 frmdstryr ***@***.***> wrote: I was able to get it event faster by moving all the PyImport_ImportModule into the modexec. — Reply to this email directly, view it on GitHub <#562 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AABBQSI6F46AQ37WUIBTFKD2JQTY7AVCNFSM6AAAAABUVUII22VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKNZWGA3TMOBZGA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[MatthieuDartiailh]

Sure, @sccolbert , I will wait for your approval before merging.

[frmdstryr]

The original example now runs in 9 seconds, down from the original 50 seconds.

[MatthieuDartiailh]

That's a really nice improvement !