autodoc: Change dictionary sort order #22441
Conversation
To review this pull request, I built blead through
Grepping the two generated files (as described below), I couldn't find any
If I grepped for ... then took a ... which appears to meet your objective.
Using the search procedure described above, I found in blead: ... while in the branch I found: So there was no change in |
khwilliamson
force-pushed
the
api_sort_order
branch
from
c719da9 to
32eaf9d
Compare
|
Sorry for my glib description. I forgot that this retained the existing sort order of numbers where the come after letters. I changed the commit message to give real examples of the things that did change |
autodoc.pl
Outdated
| # Convert all digit sequences to same length with leading zeros, so for | ||
| # example, 8 will compare less than 16 (using a fill length value that | ||
| # should be longer than any sequence in the input). | ||
| # Convert all digit sequences to be the same length with leading zeros, so | ||
| # that, for example '8' will sort before '16' (using a fill length value | ||
| # that should be longer than any sequence in the input). | ||
| $a =~ s/(\d+)/sprintf "%06d", $1/ge; | ||
| $b =~ s/(\d+)/sprintf "%06d", $1/ge; | ||
|
|
||
| # Translate any underscores and digits so they compare after all Unicode | ||
| # characters | ||
| $a =~ tr[_0-9]/\x{110000}-\x{11000A}/; | ||
| $b =~ tr[_0-9]/\x{110000}-\x{11000A}/; | ||
| # Translate any underscores so they sort lowest. This causes 'word1_word2' | ||
| # to sort before 'word1word2' for all words. | ||
| # And translate any digits so they come after anything else. This causes | ||
| # digits to sort highest) | ||
| $a =~ tr[_0-9]/\0\x{110000}-\x{110009}/; | ||
| $b =~ tr[_0-9]/\0\x{110000}-\x{110009}/; | ||
|
|
||
| # Then move leading underscores to the end, translating them to above | ||
| # everything else. This causes '_word_' to compare just after 'word_' | ||
| $a .= "\x{11000A}" x length $1 if $a =~ s/ ^ (\0+) //x; | ||
| $b .= "\x{11000A}" x length $1 if $b =~ s/ ^ (\0+) //x; | ||
|
|
||
| use feature 'state'; | ||
| # Modify \w, \W to reflect the changes. | ||
| state $ud = '\x{110000}-\x{11000A}'; # xlated underscore, digits | ||
| state $w = "\\w$ud"; # new \w string | ||
| use feature 'state'; | ||
| state $w = "\\w\0\x{110000}-\x{11000A}"; # new \w string | ||
| state $mod_w = qr/[$w]/; | ||
| state $mod_W = qr/[^$w]/; | ||
|
|
||
| # Only \w for initial comparison | ||
| my $a_only_word = uc($a =~ s/$mod_W//gr); | ||
| my $b_only_word = uc($b =~ s/$mod_W//gr); | ||
|
|
||
| # And not initial nor interior underscores nor digits (by squeezing them | ||
| # out) | ||
| my $a_stripped = $a_only_word =~ s/ (*atomic:[$ud]+) (*pla: $mod_w ) //grxx; | ||
| my $b_stripped = $b_only_word =~ s/ (*atomic:[$ud]+) (*pla: $mod_w ) //grxx; | ||
| # Strip out \W. | ||
| my $a_stripped = $a =~ s/$mod_W//gr; | ||
| my $b_stripped = $b =~ s/$mod_W//gr; |
There was a problem hiding this comment.
All this duplicated code between $a and $b could go into a function.
This makes this more in line with Data::Dumper sorting. upper/lower case continues to not matter, and numbers continue to come after letters, so that ckWARN2() comes after plain ckWARN(). It changes non-leading underscores to come before letters, so that ck_warner comes before ckWARN. And it changes so leading underscores come after non-leading, so that aMY_CXT and aMY_CXT_ come before _aMY_CXT.
khwilliamson
force-pushed
the
api_sort_order
branch
from
32eaf9d to
15b7330
Compare
This makes this more in line with Data::Dumper sorting.
upper/lower case continues to not matter, and numbers continue to come after letters, so that
ckWARN2()comes after plainckWARN().It changes non-leading underscores to come before letters, so that
ck_warnercomes beforeckWARN.And it changes so leading underscores come after non-leading, so that
aMY_CXTandaMY_CXT_come before_aMY_CXT