DEBUG-3182 Rework DI loading

[p-datadog]

What does this PR do?

This PR separates the part of DI that is required for code tracker (which is the code tracker itself and the DI.current_component group of methods, plus exception classes) to be self-contained. Currently in master loading datadog/di/init would bring in datadog/di as well which is too much (for example, that would cause DI to attempt to load ActiveRecord contrib, which is definitely not the right time for it).

Motivation:

Ensuring DI code tracker is loaded early without causing the rest of dd-trace-rb to be loaded early.

Change log entry

Yes: Improved early loading mechanism of dynamic instrumentation (datadog/di/init).

Additional Notes:

While this PR changes the behavior of datadog/di/init, this behavior has not yet been documented anywhere and should not be used by customers.

How to test the change?

Integration tests specifically for datadog/di/init will be added in a follow-up PR. The existing unit tests verify that basic DI functionality continues to operate correctly.

[datadog-datadog-prod-us1]

Datadog Report

Branch report: di-loading
Commit report: a482c0f
Test service: dd-trace-rb

✅ 0 Failed, 22113 Passed, 1476 Skipped, 5m 57.55s Total Time

[pr-commenter]

Benchmarks

Benchmark execution time: 2025-01-09 06:59:09

Comparing candidate commit a482c0f in PR branch di-loading with baseline commit 2e98fc4 in branch master.

Found 0 performance improvements and 0 performance regressions! Performance is the same for 31 metrics, 2 unstable metrics.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 79.24528% with 11 lines in your changes missing coverage. Please review.

Project coverage is 97.73%. Comparing base (2e98fc4) to head (a482c0f).

Files with missing lines	Patch %	Lines
lib/datadog/di/base.rb	71.42%	8 Missing ⚠️
spec/datadog/di/init_spec.rb	87.50%	2 Missing ⚠️
lib/datadog/di/code_tracker.rb	50.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #4239      +/-   ##
==========================================
- Coverage   97.73%   97.73%   -0.01%     
==========================================
  Files        1352     1354       +2     
  Lines       82424    82449      +25     
  Branches     4232     4236       +4     
==========================================
+ Hits        80558    80581      +23     
- Misses       1866     1868       +2     

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

[ivoanjo]

While this PR changes the behavior of datadog/di/init, this behavior has not yet been documented anywhere and should not be used by customers.

If I can offer some advice on this, is as much as possible don't expose any API that's expected to be used directly by customers in DI. Specifically, since DI is mostly about "turn it on, then go to the datadog UX to clicky clicky", it's worth taking advantage of this by not allowing too much configuration/fiddling/and manual operations.

This is somewhat similar to profiling: it's supported to be -- turn it on, then go look at the results. There's one or two public APIs but very very little surface, which has made it much easier to move fast and improve things without needing to walk on egg shells (or needing to put extra work or add extra overhead to comply with some specific public APIs)

[ivoanjo]

Left a few notes!

[ivoanjo]

👍

We spoke via dm and while I'm somewhat concerned about the complexity around tracking the current component (locks, state to be tracked, potential memory leaks if not used correctly), this PR only moves something that already existed, so we can tackle that separately.