perf: speed up otel trace generation #26854
Conversation
| @@ -192,13 +223,28 @@ class Span { | |||
| this.enter(); | |||
| } | |||
|
|
|||
| get traceId() { | |||
There was a problem hiding this comment.
these properties don't exist in otel. if buffers are fastest we could just do this.traceId = buffer without using a private. if/when we do perf work in upstream otel, we could do similar, with strings only being created in places where the api expects it.
There was a problem hiding this comment.
We could also remove these fields entirely then and only have the spanContext() API?
There was a problem hiding this comment.
well, the original idea was to avoid needing to use spanContext, since the extra object allocation is theoretically unneeded when flushing a span.
There was a problem hiding this comment.
Yeah, but if you look at how Deno.tracing.Span spans are flushed now, they don't access either spanContext or traceId (only the internal #traceId).
Speeds up trace generation by about 30%.