Recreate span on every run of a decorated function

[anton-ryzhov]

Description

start_as_current_span could be used as decorator. But it works incorrectly in this case — a span is created only once on decoration time and it is reused for every run of function.

With that change it creates new span on every run.

Fixes # (issue)

Type of change

Please delete options that are not relevant.

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

How Has This Been Tested?

Unit test provided

Does This PR Require a Contrib Repo Change?

• Yes. - Link to PR:
• No.

Checklist:

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

[anton-ryzhov]

Any comments?

[toumorokoshi]

This is a really interesting use case! I see the value of being able to cheaply wrap functions like this.

My one concern is the function isn't intuitively named to indicate it's a decorator. It may be good to add examples to the docs/faq page.

[anton-ryzhov]

I thought it was implied to use it as a decorator, that's rather expected to me. Saves one indentation level.

named to indicate it's a decorator.

How decorators should be named? Is there any special convention on it? I don't know any.

[ocelotl]

I thought it was implied to use it as a decorator, that's rather expected to me. Saves one indentation level.

named to indicate it's a decorator.

How decorators should be named? Is there any special convention on it? I don't know any.

There are no naming conventions for decorators. Mentioning in the documentation that this function can be used also as a decorator would be valuable, since every example so far does not use this pattern.

[aabmass]

Does this same decorator pattern work with start_span()? I would expect them to behave the same

[anton-ryzhov]

It doesn't use @contextmanager, so no. And it should not

[owais]

This is a really interesting use case! I see the value of being able to cheaply wrap functions like this.

My one concern is the function isn't intuitively named to indicate it's a decorator. It may be good to add examples to the docs/faq page.

Agree with this. A better name would be nice. Ignoring all other Otel conventions, something like @otel.trace(name='') would have been most intuitive name for me but we probably want to stay closer to otel conventions and almost everywhere else we use span in functions that start or create spans. So, perhaps something like @with_span(...) or @record_span(...)?

[anton-ryzhov]

Some pipelines are broken after branch update. I don't know why but pretty sure not because of this PR.

Naming question — is it part of this PR? Could you please merge that bugfix first?

[anton-ryzhov]

Lets merge it before CI will disapprove it again?

[codeboten]

@anton-ryzhov looks like we missed it again if you can update the branch

[anton-ryzhov]

@codeboten updated

[anton-ryzhov]

Any chance that this will be merged?

With current setup CHANGELOG file will always have conflicts, so only chance to merge PR is to update it right before merging. But the PR is missed again and again.

[codeboten]

@anton-ryzhov if you can resolve the conflict, @lzchen and I can hold off on merging any other changes until this one passes CI and gets merged

[anton-ryzhov]

Updated

[codeboten]

Thanks @anton-ryzhov! Merged

[anton-ryzhov]

Thank you