Don't use viper.Get* in tight loops

[bep]

Profiling indicate that using viper.GetBool inside the url funcs (MakePath etc.) have a surprising cost.

Just removing one viper.Get temporary from MakePathSanitized:

benchmark           old ns/op       new ns/op       delta
BenchmarkHugo-4     20347856514     17916615991     -11.95%

benchmark           old allocs     new allocs     delta
BenchmarkHugo-4     66611825       60243953       -9.56%

benchmark           old bytes       new bytes       delta
BenchmarkHugo-4     10282345656     10122546104     -1.55%

[bep]

I have "suggested" this for the 0.17 release -- but it is probably a more natural fit for the next ... My PR isn't entirely finished, but I will wrap it up during the day.

I'm not sure why Viper allocates memory on a simple GetBool (I will look into that), but we should anyhow avoid doing lookups in tight loops (I didn't consider the URL funcs to be in "tight loops").

benchmark           old ns/op       new ns/op       delta
BenchmarkHugo-4     32771543767     20956368681     -36.05%

benchmark           old allocs     new allocs     delta
BenchmarkHugo-4     66611886       51747681       -22.31%

benchmark           old bytes       new bytes      delta
BenchmarkHugo-4     10272277160     9906356584     -3.56%

/cc @spf13

[bep]

Also see spf13/viper#242

[bep]

With the fix in cast, the win is less, but still significant:

benchmark           old ns/op       new ns/op       delta
BenchmarkHugo-4     24116936011     19791183884     -17.94%

benchmark           old allocs     new allocs     delta
BenchmarkHugo-4     60902955       50003607       -17.90%

benchmark           old bytes       new bytes      delta
BenchmarkHugo-4     10121519632     9857226960     -2.61%

[moorereason]

I like your PathSpec approach. I was thinking along the same lines when I first saw this issue, but with your fix in cast, I'd say we hold this change for 0.18.

[bep]

I have another simple fix for Viper coming that will improve this even further.

[bep]

This is the delta after the latest Viper update:

benchmark           old ns/op       new ns/op       delta
BenchmarkHugo-4     14465408023     13147573109     -9.11%

benchmark           old allocs     new allocs     delta
BenchmarkHugo-4     55147163       48210421       -12.58%

benchmark           old bytes      new bytes      delta
BenchmarkHugo-4     9998583512     9815322448     -1.83%

Still worth doing, but not super-important anymore.

[github-actions]

This issue has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.