api: conditionally load entrypoints for OTEL_PYTHON_CONTEXT by codeboten · Pull Request #5144 · open-telemetry/opentelemetry-python

codeboten · 2026-04-24T20:43:11Z

Description

This prevents unnecessarily importing the entry points library, reducing the memory footprint of loading the API by about a MB (27%):

Scenario	Memory before	Memory after	delta
Import API only	5.43 MB	4.49 MB	-940 KB
API + get_tracer (no-op)	5.05 MB	4.10 MB	-950 KB

Type of change

Please delete options that are not relevant.

Bug fix (non-breaking change which fixes an issue)

How Has This Been Tested?

Please describe the tests that you ran to verify your changes. Provide instructions so we can reproduce. Please also list any relevant details for your test configuration

Ran unit tests locallt

Does This PR Require a Contrib Repo Change?

No.

Checklist:

Followed the style guidelines of this project
Changelogs have been updated
Unit tests have been added
Documentation has been updated

This prevents unnecessarily importing the entry points library, reducing the memory footprint of loading the API by about a MB (27%): |Scenario | Memory before | Memory after | delta | | - | - | - | - | |Import API only | 5.43 MB | 4.49 MB | -940 KB | |API + get_tracer (no-op) | 5.05 MB | 4.10 MB | -950 KB | Signed-off-by: Alex Boten <223565+codeboten@users.noreply.github.com>

MikeGoldsmith

Looks good, thanks @codeboten. Please add a changelog entry 👍🏻

Signed-off-by: Alex Boten <223565+codeboten@users.noreply.github.com>

codeboten · 2026-04-27T16:49:16Z

added changelog, ptal

Signed-off-by: Alex Boten <223565+codeboten@users.noreply.github.com>

codeboten

dont think the check-links failure is related to this change

MikeGoldsmith

Nice, thanks @codeboten 🚀

codeboten · 2026-05-11T16:44:17Z

@open-telemetry/python-maintainers is there anything else needed on this pr to move it forward?

herin049 · 2026-05-11T17:17:34Z

@open-telemetry/python-maintainers is there anything else needed on this pr to move it forward?

@codeboten There are a large number of PRs in the backlog, so unfortunately it's just a bit of a waiting game.

codeboten · 2026-05-11T19:07:57Z

thanks @herin049, just making sure there isn't something i should be doing here :)

emdneto · 2026-05-12T22:35:22Z

Thanks for the PR!

Just a heads-up: we no longer update CHANGELOG.md directly. The changelog is now generated from changelog fragments using Towncrier.

Please add the appropriate changelog fragment for this change instead of editing CHANGELOG.md manually. You can find the instructions and expected format in CONTRIBUTING.md.

xrmx · 2026-05-13T15:18:31Z

We have a WIP PR that will remove importlib_metadata dependency so maybe we can revise this after that to check if importlib.metadata has the same footprint?

Signed-off-by: Alex Boten <223565+codeboten@users.noreply.github.com>

MikeGoldsmith · 2026-05-14T15:26:18Z

We have a WIP PR that will remove importlib_metadata dependency so maybe we can revise this after that to check if importlib.metadata has the same footprint?

Is there any harm is accepting this and then updating to the other PR to update the import library if we want to accept it? This PR predates the importlib PR by 3 weeks and been approved for 2 weeks.

This is the PR:

Remove importlib-metadata dependency #5203

xrmx · 2026-05-14T15:43:44Z

We have a WIP PR that will remove importlib_metadata dependency so maybe we can revise this after that to check if importlib.metadata has the same footprint?

Is there any harm is accepting this and then updating to the other PR to update the import library if we want to accept it? This PR predates the importlib PR by 3 weeks and been approved for 2 weeks.

Sorry I'm assuming that moving the import of importlib_metadata into the user of that code takes part in improving memory usage and startup time. Then I was asking to double check that's the case with the stdlib version as well. If that's not the case then we can put back it at the top level?

codeboten · 2026-05-14T15:57:24Z

I've run some comparisons on top of the changes in the importlib-metadata PR and here's the results:

Scenario	Without branch	With branch	Savings
1. Python baseline (no OTel)	—	—	—
2. Import opentelemetry API	5.85 MB	5.58 MB	279 KB
3. API + get_tracer (no-op)	5.47 MB	5.20 MB	280 KB
4. SDK TracerProvider (no exporter)	8.38 MB	8.11 MB	279 KB
5. SDK TracerProvider + OTLP	17.93 MB	17.66 MB	278 KB
6. SDK MeterProvider (no exporter)	9.18 MB	8.90 MB	278 KB
7. SDK LoggerProvider (no exporter)	8.26 MB	7.98 MB	278 KB
8. Full SDK	9.87 MB	9.60 MB	278 KB

aabmass

@xrmx OK to merge this? It looks like it's still a noticeable improvement

Signed-off-by: Alex Boten <223565+codeboten@users.noreply.github.com>

aabmass · 2026-05-20T03:01:32Z

From public API Check

opentelemetry-api/src/opentelemetry/propagate/__init__.py:0: propagators: Public object was removed
opentelemetry-api/src/opentelemetry/propagate/__init__.py:0: environ_propagators: Public object was removed
opentelemetry-api/src/opentelemetry/propagate/__init__.py:0: propagator: Public object was removed

I wonder if anyone in the wild was using these accidentally exposed imports. We could probably do a quick GH search to check if this inadvertently breaks anyone.

Copilot

Pull request overview

This PR reduces the baseline import cost of opentelemetry-api by deferring entry_points usage (and related metadata machinery) until it’s actually needed, especially around context selection via OTEL_PYTHON_CONTEXT.

Changes:

Lazily import entry_points in provider loading utilities to avoid importing entry point machinery during a plain API import.
Update propagator initialization to avoid entry_points when OTEL_PROPAGATORS is unset (use direct default propagator classes instead).
Update runtime context initialization to return ContextVarsRuntimeContext directly when OTEL_PYTHON_CONTEXT is not set, and only consult entry points when it is set.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

File	Description
opentelemetry-api/src/opentelemetry/util/_providers.py	Moves `entry_points` import inside `_load_provider` to avoid import-time overhead unless env-driven provider loading occurs.
opentelemetry-api/src/opentelemetry/propagate/init.py	Introduces `_load_propagators()` and avoids entry point loading for the default propagators when `OTEL_PROPAGATORS` is unset.
opentelemetry-api/src/opentelemetry/context/init.py	Avoids `entry_points` unless `OTEL_PYTHON_CONTEXT` is explicitly configured; falls back directly to `ContextVarsRuntimeContext`.
.changelog/5144.changed	Adds a changelog entry for the conditional entry point import behavior change.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

+    # pylint: disable=import-outside-toplevel,no-name-in-module
+    from opentelemetry.util._importlib_metadata import (  # noqa: PLC0415
+        entry_points,
+    )
+
+    propagators: list[textmap.TextMapPropagator] = []
+    for propagator in configured.split(","):
+        propagator = propagator.strip()
+        if propagator.lower() == "none":
+            logger.debug(
+                "OTEL_PROPAGATORS environment variable contains none, removing all propagators"
+            )
+            return composite.CompositePropagator([])


+    configured_context = os.environ.get(OTEL_PYTHON_CONTEXT)
+    if not configured_context:
+        return ContextVarsRuntimeContext()

-    # FIXME use a better implementation of a configuration manager
-    # to avoid having to get configuration values straight from
-    # environment variables
-    default_context = "contextvars_context"
-
-    configured_context = environ.get(OTEL_PYTHON_CONTEXT, default_context)  # type: str
+    # pylint: disable=import-outside-toplevel,no-name-in-module
+    from opentelemetry.util._importlib_metadata import (  # noqa: PLC0415
+        entry_points,
+    )

    try:
-        return next(  # type: ignore
-            iter(  # type: ignore
-                entry_points(  # type: ignore
-                    group="opentelemetry_context",
-                    name=configured_context,
+        return next(
+            iter(
+                entry_points(
+                    group="opentelemetry_context", name=configured_context
                )
            )
        ).load()()


xrmx · 2026-05-20T08:53:44Z

@xrmx OK to merge this? It looks like it's still a noticeable improvement

I'm not opposed to merge this since we have a bunch of approvals already and there's some memory savings. I haven't profiled this myself so I don't know what part of the changes is responsible of the savings and I'm not sure all the delayed imports in this PRs (e.g. the propagators) makes a difference or not.

Out of scope for to this PR but if we are going to think about memory optimizations we should probably track something in benchmarks so we can see if there's regressions.
Speaking of optimizations, auto-instrumentation sdk config is probalby the biggest user of entry points so if they are a pain point there we can have the biggest wins.

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

aabmass · 2026-05-20T20:22:27Z

Agreed no lazy imports. IMO we should just wait for 3.15 since it's been solved in python.

Speaking of optimizations, auto-instrumentation sdk config is probalby the biggest user of entry points so if they are a pain point there we can have the biggest wins.

I believe this PR is trying to optimizing the opposite case, where entry_points can be totally avoided easily. (sorry if I'm recapping what you already knew)

Signed-off-by: Alex Boten <223565+codeboten@users.noreply.github.com>

codeboten · 2026-05-20T22:11:18Z

From public API Check
opentelemetry-api/src/opentelemetry/propagate/__init__.py:0: propagators: Public object was removed
opentelemetry-api/src/opentelemetry/propagate/__init__.py:0: environ_propagators: Public object was removed
opentelemetry-api/src/opentelemetry/propagate/__init__.py:0: propagator: Public object was removed
I wonder if anyone in the wild was using these accidentally exposed imports. We could probably do a quick GH search to check if this inadvertently breaks anyone.

I ended up putting those back, didn't want to break a public api

aabmass · 2026-05-21T02:38:40Z

Thank you @codeboten! I turned on auto merge which should start once the Copilot comments are resolved

codeboten requested a review from a team as a code owner April 24, 2026 20:43

github-project-automation Bot added this to Python PR digest Apr 24, 2026

codeboten marked this pull request as draft April 24, 2026 20:43

MikeGoldsmith reviewed Apr 27, 2026

View reviewed changes

add ignore rule to import lint, update changelog

4024ef7

Signed-off-by: Alex Boten <223565+codeboten@users.noreply.github.com>

codeboten marked this pull request as ready for review April 27, 2026 16:49

codeboten added 2 commits April 27, 2026 11:49

fix precommit

2861d76

Signed-off-by: Alex Boten <223565+codeboten@users.noreply.github.com>

clean up changelog

460af4d

Signed-off-by: Alex Boten <223565+codeboten@users.noreply.github.com>

codeboten commented Apr 28, 2026

View reviewed changes

Merge branch 'main' into codeboten/load-context-api

7d2e1c6

MikeGoldsmith approved these changes Apr 29, 2026

View reviewed changes

github-project-automation Bot moved this to Approved PRs in Python PR digest Apr 29, 2026

codeboten and others added 7 commits April 30, 2026 08:01

Merge branch 'main' into codeboten/load-context-api

c0411a2

Merge branch 'main' into codeboten/load-context-api

6099eee

Merge branch 'main' into codeboten/load-context-api

91c3996

Merge branch 'main' into codeboten/load-context-api

c50a2fd

Merge branch 'main' into codeboten/load-context-api

6750d11

Merge branch 'main' into codeboten/load-context-api

76c5580

Merge branch 'main' into codeboten/load-context-api

bd6ae52

herin049 approved these changes May 11, 2026

View reviewed changes

Merge branch 'main' into codeboten/load-context-api

d0de9e8

Merge branch 'main' into codeboten/load-context-api

c1bf486

Merge branch 'main' into codeboten/load-context-api

4744de1

codeboten added 2 commits May 14, 2026 08:17

use changelog fragment

9152c35

Signed-off-by: Alex Boten <223565+codeboten@users.noreply.github.com>

Merge branch 'main' into codeboten/load-context-api

2a66974

Merge branch 'main' into codeboten/load-context-api

6bb9251

aabmass approved these changes May 18, 2026

View reviewed changes

Comment thread opentelemetry-api/src/opentelemetry/context/__init__.py

codeboten added 3 commits May 19, 2026 07:22

Merge branch 'main' into codeboten/load-context-api

9d8e230

apply the same pattern to propagators and providers

aba0170

Signed-off-by: Alex Boten <223565+codeboten@users.noreply.github.com>

precommit

7affb49

Signed-off-by: Alex Boten <223565+codeboten@users.noreply.github.com>

aabmass approved these changes May 20, 2026

View reviewed changes

Comment thread opentelemetry-api/src/opentelemetry/util/_providers.py

aabmass requested a review from Copilot May 20, 2026 03:01

Copilot started reviewing on behalf of aabmass May 20, 2026 03:02 View session

Copilot AI reviewed May 20, 2026

View reviewed changes

Potential fix for pull request finding

ab54873

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

codeboten added 3 commits May 20, 2026 14:59

put public symbols back

8879126

Signed-off-by: Alex Boten <223565+codeboten@users.noreply.github.com>

rename vars

861cf10

Signed-off-by: Alex Boten <223565+codeboten@users.noreply.github.com>

Merge branch 'main' into codeboten/load-context-api

bbabf36

aabmass enabled auto-merge May 21, 2026 02:37

Conversation

codeboten commented Apr 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Type of change

How Has This Been Tested?

Does This PR Require a Contrib Repo Change?

Checklist:

Uh oh!

MikeGoldsmith left a comment

Choose a reason for hiding this comment

Uh oh!

codeboten commented Apr 27, 2026

Uh oh!

codeboten left a comment

Choose a reason for hiding this comment

Uh oh!

MikeGoldsmith left a comment

Choose a reason for hiding this comment

Uh oh!

codeboten commented May 11, 2026

Uh oh!

herin049 commented May 11, 2026

Uh oh!

codeboten commented May 11, 2026

Uh oh!

emdneto commented May 12, 2026

Uh oh!

xrmx commented May 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

MikeGoldsmith commented May 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

xrmx commented May 14, 2026

Uh oh!

codeboten commented May 14, 2026

Uh oh!

aabmass left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

aabmass commented May 20, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

xrmx commented May 20, 2026

Uh oh!

aabmass commented May 20, 2026

Uh oh!

codeboten commented May 20, 2026

Uh oh!

aabmass commented May 21, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

codeboten commented Apr 24, 2026 •

edited

Loading

xrmx commented May 13, 2026 •

edited

Loading

MikeGoldsmith commented May 14, 2026 •

edited

Loading