Code optimization: The relabel operation is used very frequently, and strconv.FormatInt() with better performance should be used.

[aiwhj]

Code optimization: The relabel operation is used very frequently, and strconv.FormatInt() with better performance should be used.
Fixed #13668

[machine424]

even if it seems trivial to you, adding some benchmarks may help justify this.

[aknuds1]

Agree with @machine424.

[tesla59]

The difference is very minimal

goos: darwin
goarch: arm64
pkg: github.com/prometheus/prometheus/model/relabel
                     │   fmt.txt   │             strconv.txt             │
                     │   sec/op    │   sec/op     vs base                │
Relabel/example-8      1.211µ ± 0%   1.210µ ± 0%       ~ (p=0.350 n=100)
Relabel/kubernetes-8   5.432µ ± 0%   5.358µ ± 0%  -1.35% (n=100)
geomean                2.564µ        2.546µ       -0.70%

                     │   fmt.txt    │             strconv.txt              │
                     │     B/op     │     B/op      vs base                │
Relabel/example-8        606.0 ± 0%     606.0 ± 0%   0.00% (p=0.010 n=100)
Relabel/kubernetes-8   1.417Ki ± 0%   1.417Ki ± 0%  -0.03% (p=0.000 n=100)
geomean                  937.7          937.6       -0.02%

                     │  fmt.txt   │             strconv.txt              │
                     │ allocs/op  │ allocs/op   vs base                  │
Relabel/example-8      19.00 ± 0%   19.00 ± 0%       ~ (p=1.000 n=100) ¹
Relabel/kubernetes-8   36.00 ± 0%   36.00 ± 0%       ~ (p=1.000 n=100) ¹
geomean                26.15        26.15       +0.00%
¹ all samples are equal

Number of test samples: 100
Run on MacBook Air M2 with 16GB of RAM and MacOS 14.4.1

[tesla59]

Although I think using strconv would be idiomatic because it is more readable

[machine424]

Maybe we should enforce this using a linter or sth.
There is golangci/golangci-lint#3714
and #13220 tried to do that.

[alexandear]

I created PR #14091, in which I made the same changes and enabled the perfsprint linter to prevent similar situations in the future.

[tesla59]

Maybe we should enforce this using a linter or sth. There is golangci/golangci-lint#3714 and #13220 tried to do that.

As per your suggestions, i have created the pr #14092 which enables the linter and fixed all issues generated by it

[machine424]

Thanks, @tesla59, for your responsiveness and efforts. However, I’m afraid @alexandear re-opened his PR first. He’s the author of the closed PR I linked to, where we discussed the reopening. You could help us review it if you want.
(To avoid such conflicts in the future, I’d suggest expressing your intention to work on an issue before creating a PR ;))

[tesla59]

Thanks, @tesla59, for your responsiveness and efforts. However, I’m afraid @alexandear re-opened his PR first. He’s the author of the closed PR I linked to, where we discussed the reopening. You could help us review it if you want.
(To avoid such conflicts in the future, I’d suggest expressing your intention to work on an issue before creating a PR ;))

Sure @machine424 I'll make sure to comment before working to avoid such conflicts

[bboreham]

Thanks!