chore(instrumentation-pg): Inline incubating constants from @opentelemetry/semantic-conventions #2599
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #2599 +/- ##
==========================================
+ Coverage 90.78% 90.79% +0.01%
==========================================
Files 169 170 +1
Lines 8055 8064 +9
Branches 1643 1643
==========================================
+ Hits 7313 7322 +9
Misses 742 742
|
|
The reason for this versioning being pinned, it's because this package is using the I see other packages that use the |
|
Ah! Apologies, I didn't go digging deep enough. In that case, can I just change this PR to pin to 1.28.0 so that it will be deduped by package managers (at least until 1.29.0 comes out)? |
Hi 🙂 I'd say it's a mistake. When using the incubating entrypoint, we should always pin for now until we have a decision on open-telemetry/opentelemetry-js#5182. Sorry for the inconvenience. |
|
Hi @nwalters512 , looks like we will go with a different approach of copying those specific values from the incubation, see more here: open-telemetry/opentelemetry-js#5182 (comment) I would be happy to review if that is something you want to work on. |
|
@maryliag I'd be happy to open some PRs for this! I'll work on |
|
thank you @nwalters512! |
nwalters512
changed the title
There was a problem hiding this comment.
@maryliag I repurposed this PR to include the inlining of the relevant constants.
| @@ -65,7 +65,7 @@ import { | |||
| METRIC_DB_CLIENT_OPERATION_DURATION, | |||
| ATTR_DB_NAMESPACE, | |||
| ATTR_DB_OPERATION_NAME, | |||
| } from '@opentelemetry/semantic-conventions/incubating'; | |||
| } from './semantic-conventions'; | |||
There was a problem hiding this comment.
open-telemetry/opentelemetry-js#5182 (comment) suggested using semconv.ts; I went with semantic-conventions.ts since it directly matches the package name. Happy to use the other suggestion if that's strongly preferred.
| /** | ||
| * These constants are considered experimental exports of `@opentelemetry/semantic-conventions`. | ||
| * They're being inlined until they're officially exported by `@opentelemetry/semantic-conventions`. | ||
| */ |
There was a problem hiding this comment.
This little comment will appear at the top of this file in other packages; let's workshop it if needed before it proliferates.
| /** | ||
| * The number of connections that are currently in state described by the `state` attribute | ||
| * | ||
| * @experimental This metric is experimental and is subject to breaking changes in minor releases of `@opentelemetry/semantic-conventions`. |
There was a problem hiding this comment.
I copied these all in full, including the "experimental" warning. Happy to edit this to remove it desired.
|
Sorry about the failing tests, I'll get those fixed up. |
|
Thank you for working on this! @trentm would you mind confirming if that is what you had in mind? |
Which problem is this PR solving?
No other instrumentation packages pin@opentelemetry/semantic-conventions, and from what I can tell, there's no good reason that@opentelemetry/instumentation-pgshould be pinning it. When the dependency is pinned, npm can't dedupe it to a higher version, which results in multiple versions of this package being installed.As decided in open-telemetry/opentelemetry-js#5182 (comment),
incubatingconstants from@opentelemetry/semantic-conventionsshould be inlined instead of pinning@opentelemetry/semantic-conventionsto a particular version.Short description of the changes
In@opentelemetry/instrumentation-pg,@opentelemetry/semantic-conventionsnow uses a^semver range to allow newer minor versions to be installed.This PR updates
@opentelemetry/instrumentation-pgto include inlined copies of the experimental constants it needs. This, in turn, allows us to unpin@opentelemetry/semantic-conventions.