Re: [gentoo-dev] [PATCH] eapi7-ver.eclass: 'Early adopter' version of EAPI 7 version manip

[Michał Górny]

W dniu pią, 08.09.2017 o godzinie 13∶19 +0200, użytkownik Michał Górny
napisał:
> EAPI 7 is introducing new version manipulation and comparison functions
> that aim to replace versionator.eclass. This eclass provides an 'early
> adopter' versions of those routines.
>
> It serves two goals:
>
> a. getting wider review and some real-life testing before
> the specification is set in stone, and
>
> b. making it possible to adapt ebuilds to the new routines early,
> reducing the future work of EAPI 7 porting.
>
> For more details on the new logic, please see the eclass documentation.
> Long story short, we are introducing three functions:
>
> 1. ver_cut -- to get substrings of the version string,
>
> 2. ver_rs -- to replace version separators via indices,
>
> 3. ver_test -- to compare two version numbers.
>
> The third function is not implemented in the eclass. It's meant to reuse
> the algorithms from the package manager, and the final implementation
> will most likely reuse the code from the package manager (e.g. via IPC).
>

Merged now, with some documentation fixes and additional benchmark to
compare it with versionator.eclass. Here are the approximate results:

cutting
real 5.64 5.62 5.55 5.57 5.56 => 5.59 avg
user 6.25 6.23 6.24 6.10 6.09 => 6.18 avg
cutting_versionator
real 105.00 104.10 104.40 104.30 104.10 => 104.38 avg
user 71.30 71.40 71.20 70.70 71.80 => 71.28 avg
replacing
real 4.76 4.75 4.79 4.77 4.75 => 4.76 avg
user 5.16 5.48 5.60 5.62 5.11 => 5.39 avg
replacing_versionator
real 68.50 68.60 68.40 68.20 68.40 => 68.42 avg
user 51.60 51.40 51.10 51.60 52.30 => 51.60 avg

Please note that versionator tests were run with 10 times less
repetitions and the results were multiplied by 10 to avoid it taking
forever.

-- 
Best regards,
Michał Górny