Add more data to Server-Timing header

[develop7]

Introduces the performance metrics' collection and rendering in Server-Timing header

fixes #2771

• metrics
  • query
  • jwt
  • plan
  • render
• tests
• CHANGELOG

[develop7]

dead code error at https://github.com/PostgREST/postgrest/actions/runs/6418699702/job/17427023445?pr=2983#step:4:119, the matchHeaderPresent gets unused after my PR; I'd say it's a handy matcher, I'd leave it, but don't know how

[steve-chavez]

@develop7 Interesting! So doing:

# PGRST_DB_PLAN_ENABLED="true" postgrest-with-postgresql-15 postgrest-run

$ curl 'localhost:3000/projects?select=*,clients(*)'  -I
Server-Timing: jwt;dur=16.0, render;dur=8.8, plan;dur=16956.1, query;dur=763.5

$ curl 'localhost:3000/projects?select=*,clients(*)'  -I
Server-Timing: jwt;dur=13.3, render;dur=7.5, plan;dur=118.7, query;dur=1156.1

$ curl 'localhost:3000/projects?select=*,clients(*)'  -I
Server-Timing: jwt;dur=34.3, render;dur=11.9, plan;dur=283.7, query;dur=1121.3

$ curl 'localhost:3000/projects?select=*,clients(*)'  -I
Server-Timing: jwt;dur=15.4, render;dur=5.0, plan;dur=119.0, query;dur=478.7

$ curl 'localhost:3000/projects?select=*,clients(*)'  -I
Server-Timing: jwt;dur=30.4, render;dur=9.8, plan;dur=274.1, query;dur=1165.8

Shows that:

• The plan takes a lot the first time, this must be because of PostgREST takes a lot of time to make first embedding request with big schema #2450
• The plan time varies and is not negligible.
• The bulk of the time goes to the query though.
• render time is low

dead code error at https://github.com/PostgREST/postgrest/actions/runs/6418699702/job/17427023445?pr=2983#step:4:119, the matchHeaderPresent gets unused after my PR; I'd say it's a handy matcher, I'd leave it, but don't know how

Yeah, maybe just comment it for now?

[wolfgangwalther]

dead code error at https://github.com/PostgREST/postgrest/actions/runs/6418699702/job/17427023445?pr=2983#step:4:119, the matchHeaderPresent gets unused after my PR; I'd say it's a handy matcher, I'd leave it, but don't know how

Yeah, maybe just comment it for now?

Please don't. The dead code checker is there for exactly this reason: To remove dead code. If you comment it out... the code is still dead. The point of removing dead code is to reduce maintenance effort. And commenting out does not help with it.

[develop7]

Please don't. The dead code checker is there for exactly this reason: To remove dead code. If you comment it out... the code is still dead. The point of removing dead code is to reduce maintenance effort. And commenting out does not help with it.

Good point, although the above matcher still can be handy later, can it? Ideally, it belongs if not hspec-wai itself, then hspec-wai-extra for certain. This SpecHelper module serves as a project-specific lightweight library, rather than stores business logic, so I'm not sure deleting helpers from here would decrease maintenance efforts significantly. On the other hand, the scenario when one day we delete it and next month we need it again, seems quite feasible to me.

[wolfgangwalther]

Good point, although the above matcher still can be handy later, can it? Ideally, it belongs if not hspec-wai itself, then hspec-wai-extra for certain. This SpecHelper module serves as a project-specific lightweight library, rather than stores business logic, so I'm not sure deleting helpers from here would decrease maintenance efforts significantly. On the other hand, the scenario when one day we delete it and next month we need it again, seems quite feasible to me.

Of course it can be handy later. But nothing stops you from adding it later, again.

The point about maintainability is: You can't make sure to have dead code (commented out code is even worse) to stay in sync with the remaining application - because it's dead. Are you sure the helper will work exactly the same when commenting it in again in a year or two? Maybe yes, maybe no. In this case the chance of a yes might be quite high.. but this is much more general than that: You can never be sure.

And it reduces readability, too. If I have to skip over commented out code everywhere... that's just much harder for me to parse visually.

Commenting out code should be a no-go, once you check for dead code...

[develop7]

@wolfgangwalther agree. Deleted it.

[develop7]

I also wonder if we should adopt prometheus support… oh, there's #1526 about that.

[steve-chavez]

@develop7 Great work! The new modules and test helpers are really clean too 👍.

Don't forget to add an entry to the CHANGELOG!

[develop7]

@steve-chavez changelog done, rebasing done. Hit it!