feat: add more generalized Stats

[acpana]

Signed-off-by: Alex Pana [email protected]

This patch reworks EvaluationMeta field as a more general struct.

The big win here is seeing Stats for non violating constraints
without manipulating the Result struct, which is not possible with the
implementation in #263. This was evident when we tried to integrate in gatekeeper.

reviewer notes

current papercuts

You will notice we modify the driver interface for Query() to give back a QueryResponse type. This breaks compatibility for drivers. We need to modify the Query() signature regardless to give back stateless, per call metrics. Adding the QueryResponse struct helps future proof for further additions.

UX in action

Here's what consuming this, say in g8r, would look like w fmt.Println:

scope: template, statsFor: K8sGoodRego, source: rego, statsName: templateRunTimeNanos, value: 205970, desc: the number of milliseconds it took to evaluate all constraints for a template 
scope: template, statsFor: K8sGoodRego, source: rego, statsName: constraintCount, value: 1, desc: the number of constraints that were evaluated at the same time for the given constraint kind 

These can be logged or sent to a metrics aggregator depending on needs.

call outs

• note that since we are changing the Result struct, some tests in gatekeeper are failing and that's ok

[acpana]

Adding some context;

[codecov-commenter]

Codecov Report

Patch coverage has no change and project coverage change: +1.24 🎉

Comparison is base (fc3dce7) 52.67% compared to head (f31e4ae) 53.92%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #283      +/-   ##
==========================================
+ Coverage   52.67%   53.92%   +1.24%     
==========================================
  Files          67       67              
  Lines        4592     4671      +79     
==========================================
+ Hits         2419     2519     +100     
+ Misses       1902     1885      -17     
+ Partials      271      267       -4     

Flag	Coverage Δ
unittests	53.92% <ø> (+1.24%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

see 5 files with indirect coverage changes

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[maxsmythe]

I like the aggregation model that's starting to take shape here.

[maxsmythe]

LGTM, great job!

[sozercan]

minor comments, otherwise LGTM

[ritazh]

do we want to bubble this error up?

[acpana]

I could be persuaded to bubble the error up. I know that Max made a comment to bubble up the error from the driver.GetDescriptionForStat.

For the client UX, I thought it was better for the emission (logging) to not have to handle the error and always get in hand a string, the description requested or an "unknown description" string.

[maxsmythe]

IMO we should go all-in one direction or the other (either bubble up the error all the way or don't raise an error).

For now, I'm okay with leaving this as-is, we can revamp as we learn more about how this is used.

[ritazh]

minor comment
LGTM