[refactor] mailIssueCommentToParticipants

[6543]

utilize map

[6543]

blocked by #10475 (if this is merged first a backport would be harder ...)

[guillep2k]

I don't know. I carefully considered maps for this when I did it. Every time you access a map entry a hash function is executed and possibly a bucket must be allocated. Maps are slower to populate than slices. IMHO the current implementation is faster because it checks the map only one time per item after it is allocated with the required size. Especially in the 90% of the cases where the list is quite short. I've even did some benchmarks, I believe.

I could be looking at this all wrong, though. 😅

[6543]

I don't know. I carefully considered maps for this when I did it. Every time you access a map entry a hash function is executed and possibly a bucket must be allocated. Maps are slower to populate than slices. IMHO the current implementation is faster because it checks the map only one time per item after it is allocated with the required size. Especially in the 90% of the cases where the list is quite short. I've even did some benchmarks, I believe.

I could be looking at this all wrong, though.

we get rid of https://github.com/go-gitea/gitea/pull/10478/files#diff-959cf0e72f197b979a8047af5f475157L113-L119 witch is an impruvement in my opinion ... but how did you make the benchmarks?

[guillep2k]

we get rid of https://github.com/go-gitea/gitea/pull/10478/files#diff-959cf0e72f197b979a8047af5f475157L113-L119 witch is an impruvement in my opinion ... but how did you make the benchmarks?

Those two loops run through the data exactly one time; you may think they go over the same items several times, but they don't. A for ... range on a map would be slower than that.

As for testing, I don't recall exactly, but there's only one way I'd do it, and it's programming loops with static data and timing them running in the 100,000 of times. The database must be excluded from the benchmarks because it's a common part of and the slowest by several orders of magnitude, of course.

[6543]

so if map dont add speed to it, I'll close this ... ?

[guillep2k]

I don't know, others might have a different opinion than me. Or consider your code better for other reasons (e.g. clarity).

[6543]

this got it's change