NAME

    Sort::HashKeys - Get a sorted-by-key list from a hash

SYNOPSIS

        use Sort::HashKeys;
    
        my %hash;
    
        @sorted_1 = map { ($_, $hash{$_}) } sort keys %hash;
        @sorted_2 = Sort::HashKeys::sort(%hash);
        # Same outcome, but the second is faster

DESCRIPTION

        [13:37:51]  <a3f>   Is there a better way to get a sorted list out of a hash
                            than map { ($_, $versions{$_}) } reverse sort keys @versions
                            or iterating manually?
        [13:39:06]  <a3f>   oh I could provide a compare function to sort and chunk the
                            list two by two..
        [13:40:15]  <haarg> i'd probably go with the map{} reverse sort keys
        [13:41:04]  <a3f>   I don't like it that it repeats the lookup for all keys.
                            Of course wouldn't matter in practice but still…
        [13:43:40]  <haarg> whatever other solution you find will be slower
        [13:49:05]  <a3f>   put it into a list, pass it to XS, run qsort(3) over it
                            with double the element size. and compare taking only the
                            first part into account. return it back?

BENCHMARK

    See benchmark.pl
    <https://github.com/athreef/Sort-HashKeys/blob/master/benchmark.pl> in
    this distribution. Test was run on a Haswell 2.6 GHz i5 CPU (4278U) for
    a minute each on a copy of a randomly generated hash of 1000 keys. Keys
    were alphanumeric with length between 1 and 6 and values of integers
    between 1 and 1000.

        # perl-5.24.0 w/ BSD libc
                                                 Rate
        map {($_,$h{$_})} sort keys %h           1589/s       --      -28%     -32%     -33%
        S::HK::sort(%h) w/ qsort (threaded)      2199/s       38%      --       -6%      -7%
        S::HK::sort(%h) w/ qsort_r (threaded)    2338/s       47%       6%      --       -2%
        S::HK::sort(%h) w/ qsort (non-threaded)  2375/s       49%       8%       2%      --

    49% faster on non-threaded Perl and up to 47% faster on threaded Perl.

    The Slowdown with qsort on threaded build is because the interpreter
    state is fetched from thread local storage on every comparison. On GNU
    and BSD systems, this is avoided by using non-standard qsort_r, which
    allows passing an extra parameter to the comparison function.

METHODS AND ARGUMENTS

    sort(@)

      Sorts a hash-like list (key1 => val1, key2 => val2>) by means of your
      libc's qsort <http://en.cppreference.com/w/c/algorithm/qsort> in
      ascending order according to Perl's internal sv_cmp function.

      Unlike Perl's built-in sort. qsort(3) is not required to be a stable
      sort and providing a custom comparison function is not yet supported.

      Lists with odd number of elements are padded by an undef.

    reverse_sort(@)

      Sorts in descending order.

GIT REPOSITORY

    http://github.com/athreef/Sort-HashKeys

AUTHOR

    Ahmad Fatoum <athreef@cpan.org>, http://a3f.at

COPYRIGHT AND LICENSE

    Copyright (C) 2017 Ahmad Fatoum

    This library is free software; you can redistribute it and/or modify it
    under the same terms as Perl itself.