perf(serverless): lazy import botocore code

[purple4reina]

This PR lazy loads a couple of packages so they are only imported when needed.

Checklist

• PR author has checked that all the criteria below are met
• The PR description includes an overview of the change
• The PR description articulates the motivation for the change
• The change includes tests OR the PR description describes a testing strategy
• The PR description notes risks associated with the change, if any
• Newly-added code is easy to change
• The change follows the library release note guidelines
• The change includes or references documentation updates if necessary
• Backport labels are set (if applicable)

Reviewer Checklist

• Reviewer has checked that all the criteria below are met
• Title is accurate
• All changes are related to the pull request's stated goal
• Avoids breaking API changes
• Testing strategy adequately addresses listed risks
• Newly-added code is easy to change
• Release note makes sense to a user of the library
• If necessary, author has acknowledged and discussed the performance implications of this PR as reported in the benchmarks PR comment
• Backport labels are set in a manner that is consistent with the release branch maintenance policy

[github-actions]

CODEOWNERS have been resolved as:

ddtrace/_trace/trace_handlers.py                                        @DataDog/apm-sdk-api-python
tests/internal/test_serverless.py                                       @DataDog/apm-core-python @DataDog/apm-serverless

[github-actions]

Bootstrap import analysis

Comparison of import times between this PR and main.

Summary

The average import time in this PR is: 226 ± 2 ms.

The average import time in main is: 237 ± 2 ms.

The import time difference between this PR and main is: -10.74 ± 0.07 ms.

Import time breakdown

The following import paths have disappeared:

ddtrace.auto 9.149 ms (4.04%)

ddtrace.bootstrap.sitecustomize 9.149 ms (4.04%)

ddtrace._trace.trace_handlers 9.149 ms (4.04%)

ddtrace._trace.utils_botocore.span_tags 6.463 ms (2.85%)

ddtrace._trace.utils_botocore.aws_payload_tagging 5.902 ms (2.61%)

ddtrace.vendor.jsonpath_ng 3.999 ms (1.77%)

ddtrace.vendor.jsonpath_ng.jsonpath 2.315 ms (1.02%)

ddtrace.vendor.jsonpath_ng.lexer 1.643 ms (0.73%)

ddtrace.vendor.ply.lex 0.999 ms (0.44%)

ddtrace.vendor.ply 0.247 ms (0.11%)

ddtrace.vendor.jsonpath_ng.exceptions 0.296 ms (0.13%)

ddtrace.vendor.jsonpath_ng.parser 1.412 ms (0.62%)

ddtrace.vendor.ply.yacc 1.035 ms (0.46%)

decimal 1.589 ms (0.70%)

_decimal 1.358 ms (0.60%)

numbers 0.565 ms (0.25%)

ddtrace.ext.aws 0.271 ms (0.12%)

ddtrace._trace.utils_botocore.span_pointers 2.686 ms (1.19%)

ddtrace._trace.utils_botocore.span_pointers.dynamodb 1.449 ms (0.64%)

ddtrace._trace.utils_botocore.span_pointers.telemetry 0.264 ms (0.12%)

ddtrace._trace.utils_botocore.span_pointers.s3 0.628 ms (0.28%)

ddtrace._trace.utils_botocore 0.249 ms (0.11%)

The following import paths have grown:

ddtrace.auto 0.171 ms (0.08%)

ddtrace.bootstrap.sitecustomize 0.171 ms (0.08%)

ddtrace._trace.trace_handlers 0.078 ms (0.03%)

ddtrace.contrib.trace_utils 0.078 ms (0.03%)

ddtrace.contrib.internal.trace_utils 0.078 ms (0.03%)

ddtrace.ext.user 0.078 ms (0.03%)

ddtrace.bootstrap.preload 0.020 ms (0.01%)

ddtrace.internal.products 0.020 ms (0.01%)

ddtrace.settings.dynamic_instrumentation 0.020 ms (0.01%)

The following import paths have shrunk:

ddtrace.auto 2.094 ms (0.92%)

ddtrace.bootstrap.sitecustomize 1.435 ms (0.63%)

ddtrace.bootstrap.preload 1.284 ms (0.57%)

ddtrace.internal.products 1.284 ms (0.57%)

ddtrace.internal.remoteconfig.client 0.641 ms (0.28%)

ddtrace._trace.trace_handlers 0.076 ms (0.03%)

ddtrace.contrib.internal.botocore.constants 0.076 ms (0.03%)

ddtrace.contrib.internal.botocore 0.076 ms (0.03%)

ddtrace.appsec._common_module_patches 0.075 ms (0.03%)

ddtrace.appsec._asm_request_context 0.075 ms (0.03%)

ddtrace 0.659 ms (0.29%)

[purple4reina]

Reduces cold start time in aws lambda by about 2.5%

[pr-commenter]

Benchmarks

Benchmark execution time: 2025-03-24 21:37:52

Comparing candidate commit 5a900df in PR branch rey.abolofia/lazy-pointers with baseline commit 08ee25b in branch main.

Found 0 performance improvements and 0 performance regressions! Performance is the same for 478 metrics, 2 unstable metrics.

[brettlangdon]

I don't love adding the imports into the event handlers, since if they are called at a high rate, then might have a perf impact. I don't think we have benchmarks to validate this though.

Given the other comments about potentially refactoring this to d patch on import/remove integration specific handlers from always being loaded, I think that would alleviate any initial feelings I have over perf.

That all being said, think we can handle things as follow-ups.