meta: make NameEquals alloc-free

[quasilyte]

Benchstat results:

name                 old time/op    new time/op    delta
NameEquals/1part-8     76.2ns ± 2%    15.3ns ± 2%   -79.94%  (p=0.008 n=5+5)
NameEquals/3parts-8     174ns ± 1%      30ns ± 1%   -82.74%  (p=0.008 n=5+5)

name                 old alloc/op   new alloc/op   delta
NameEquals/1part-8      16.0B ± 0%      0.0B       -100.00%  (p=0.008 n=5+5)
NameEquals/3parts-8     48.0B ± 0%      0.0B       -100.00%  (p=0.008 n=5+5)

name                 old allocs/op  new allocs/op  delta
NameEquals/1part-8       1.00 ± 0%      0.00       -100.00%  (p=0.008 n=5+5)
NameEquals/3parts-8      1.00 ± 0%      0.00       -100.00%  (p=0.008 n=5+5)

Refs #364

Signed-off-by: Iskander Sharipov quasilyte@gmail.com

[YuriyNasretdinov]

This code is slightly incorrect. I think we also need to check two things:

1. That there is a character \ after part.Value ends if it's not the final part (essentially that the prefix is part.Value + "\\").
2. If it is the final part we need to check that there are no more characters in the string.

Otherwise the following would be equal:

a\b = a\bb because we only check the prefixes
a b\ = a\b because we don't check where is \ symbol

[quasilyte]

Good catch, thanks!

I believe if we handle the last part separately with "==" instead of "hasPrefix", all issues go away.

I've added test cases that check whether the mentioned scenarios lead to expected results.

1 a\b = a\bb because we only check the prefixes

Now we check the last part with ==, should be fixed

2. a b\ = a\b because we don't check where is \ symbol

This will fail in old code as well, since the second part will be checked as
HasPrefix("b", ""), which is false. In updated code it shouldn't be an issue as well, as "b"!="".

1. That there is a character \ after part.Value ends if it's not the final part (essentially that the prefix is part.Value + "\").

I can't make up an example that will pass NameEquals with something except \ as a separator. It will cause either different numbers of parts (we have an early check for it) or difference during the next part comparison. If you have an example, I can add it to tests and then it will be easier to fix it.