perf: avoid allocations when filtering nodes

[dignifiedquire]

After @whyrusleeping demonstrated us how to find slow code, I went and made this a bit faster.

Test used to profile is in go-mfs

pprof before

pprof after

[whyrusleeping]

nice improvement.

Probably want to use the 'idiomatic' slice filter though: https://github.com/golang/go/wiki/SliceTricks#filtering-without-allocating

[dignifiedquire]

I found mixed reports about the actual speed of that. If doing the inplace version is what we want I‘ll compare the two and see from there. cc @Stebalien

…

On 28. Aug 2018, 21:50 +0200, Whyrusleeping ***@***.***>, wrote: nice improvement. Probably want to use the 'idiomatic' slice filter though: https://github.com/golang/go/wiki/SliceTricks#filtering-without-allocating — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub, or mute the thread.

[Stebalien]

In-place is definitely what we want. The slice trick may not be quite as fast but it's much easier to read and should be fast enough for us.

[dignifiedquire]

changed, speed improvements in that test are similar, so looks like go does optimize this nicely