Codecov Report

Attention: Patch coverage is 87.09677% with 8 lines in your changes missing coverage. Please review.

Project coverage is 70.84%. Comparing base (91e05d9) to head (15dc601).
Report is 29 commits behind head on main.

@@             Coverage Diff              @@
##               main    #1583      +/-   ##
============================================
- Coverage     70.86%   70.84%   -0.02%     
- Complexity     2765     2798      +33     
============================================
  Files           407      411       +4     
  Lines          8346     8424      +78     
============================================
+ Hits           5914     5968      +54     
- Misses         2432     2456      +24     

I wonder if this could also be used for limitting the recursive span scenario with templating languages we talked about. @brettmc Do you see that we could enhance this for looking at presence of attributes or instrumentation names?

Example:

Twig is rendering templates recursively

index.html.twig
- header.html.twig
- - meta.html.twig
- - meta.html.twig
- - meta.html.twig
- - ....
- body.html.twig
- - content.html.twig
- - - blog_list.html.twig
- - - - blog_article.html.twig
- - - - ...

As a library author, i would probably want to define

SpanSuppression::suppress('io.open-telemetry.php.twig')

or similar, so that i can capture the overall render time of the templates, but stop recursion on the top level.

Would prefer an injectable abstraction layer similar to Javas SpanSuppressionStrategy instead of static methods to allow users to opt-out of span suppression.

I've made a start on this. I think there are some improvements that could be made:

• storing the active strategies in context rather than a static variable?
• config from env var and/or declarative config
• what are all the possible inputs to a decision?

Do you see that we could enhance this for looking at presence of attributes or instrumentation names?

@cedricziel yes, I think we could. In the twig example, what feature of the recursive spans can be used to make the decision? The tracer name? One of the SemConv attributes? This could be helpful in working out the possible inputs to the decision.