Which problem is this PR solving?

We have a large mono repo and our app users have the need for the following usage:

  new HttpInstrumentation({
    startOutgoingSpanHook(request) {
      const headers = request.headers ?? {};
      if ('x-otel-path-template' in headers) {
        return {
          [SEMATTRS_HTTP_ROUTE]: headers['x-otel-path-template'], // A path template, e.g. `/users/{id}` or `/blog/{blogId}/posts/{postId}`
        };
      }
      return {};
    },
  }),

Since our users are tracking SLOs for their services for different endpoints, they need histograms observations to include the path tempate. This path template comes from mappersmith and we take care of setting it ourselves, as can be seen above. Yes, it is a bit hacky, this path template is sent via a header, but that's our problem. 😬 It's good enough for us for now at least.

But we need to patch opentelemetry to read it and put it in the histogram observation, hence the PR.

Short description of the changes

Ensure that 'http.route' makes it to metric attributes, if it is available from the span attribute.

Type of change

Please delete options that are not relevant.

• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)

I hesitate to check breaking. If someone today is also setting SEMATTRS_HTTP_ROUTE in startOutgoingSpanHook and NOT expecting it to result in a new label in prometheus metrics, I guess that's a unexpected. But not breaking.

How Has This Been Tested?

We have had this code running in production for 1 month.

• Added tests, see PR diff

Checklist:

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