Make it possible to extract fields from the context in the slog adapter

[recht]

Since slog has context logging built in it would be nice to be able to utilize that with a zap handler, for example to log tracing context.

[CLAassistant]

All committers have signed the CLA.

[r-hang]

This makes sense to me! I think the name WithFieldsFromContext is a bit more descriptive than WithContextExtractor.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (87577d8) 98.28% compared to head (a241cd5) 98.43%.
Report is 6 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1372      +/-   ##
==========================================
+ Coverage   98.28%   98.43%   +0.14%     
==========================================
  Files          53       53              
  Lines        3493     3505      +12     
==========================================
+ Hits         3433     3450      +17     
+ Misses         50       46       -4     
+ Partials       10        9       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[recht]

This makes sense to me! I think the name WithFieldsFromContext is a bit more descriptive than WithContextExtractor.

I can change it to that, but do you then have any suggestion for the field/type name?

[r-hang]

My apologies, I've thought about it some more and looked at more code and propose a modification to

WithContextFieldExtractors(extractors ...ContextFieldExtractor), where a ContextFieldExtractor takes a context.Context and returns a single zapcore.Field.

I think this will make composing and sharing logic between multiple extractors a bit easier than if the option constructor took in a single extractor function that returned multiple fields. What do you think?

[recht]

Sorry for the late reply - we could do that instead, but it will lead to some amount of duplicated work. My specific use case is to add tracing info to logs, and for that I have something like this:

zapslog.WithContextExtractor(func(ctx context.Context) []zapcore.Field {
		span := trace.SpanFromContext(ctx)
		sctx := span.SpanContext()
		if sctx.IsValid() {
			return []zapcore.Field{
				zap.Stringer("trace_id", sctx.TraceID()),
				zap.Stringer("span_id", sctx.SpanID()),
				zap.Int8("trace_flags", int8(sctx.TraceFlags())),
			}
		}
		return nil
	})

Splitting that into 3 extractors instead then requires me to get the span context 3 times, check if it's valid, and then emit single fields.

How about instead allowing multiple extractors, but allow them to still return a slice of fields?

[r-hang]

Sorry for the delay, I will follow up here by Weds!

[r-hang]

Makes sense to me.

Providing options for multiple extractors that can provide multiple fields seems adaptable to both cases where someone might want to re-use context extractors or provide a single one for all of their fields.

Will approve after CI passes.

CI