make links cannonical

[finanalyst]

The problem

In the pod sources there are multiple uses of what appears to be a non-cannonical link, eg. L<Iterable>.

The pod description doc/Language/pod.pod6 and S26 would indicate that this should be a link to the local source Iterable.pod6. Instead the link is automagically converted by the doc system into a html link /type/iterable.

There is no indication that I can find to indicate why a single word link should be converted to point at a file in the arbitrary subdirectory html/type

Solution provided

Consequently, I have written a short program to convert any L<link> where link matches \w+ into L<link|/type/link> and any link of the form L<description|link> to L<description|/type/link>. The program is given below.

The changes in this PR are the result of these alterations.

I have run make html; make xtest and checked that individual links that I manually identified continue to work as before. (make xtest throws up a number of failure from xt/ directory tests, but none of these appear to be a consequence of the changes in this commit.

#! /usr/bin/env perl6
use v6.c;

my  $top-dir = '../perl6-doc/doc/';
#my $url = 'https\:\/\/docs\.perl6\.org';
my @extensions = <pod pod6>;

#| Recursively finds all pod files
my @pods = my sub recurse ($dir) {
     gather for dir($dir) {
         take .Str if  .extension ~~ any( @extensions );
         take slip sort recurse $_ if .d;
     }
}($top-dir); # is the first definition of $dir
my SetHash $tainted-files .= new;
for @pods -> $p {
    my @file = $p.IO.lines;
    my $tainted = False;
    for @file {
        next unless m/ 'L<' \w+ '>' /;
        s:g[ 'L<'~'>' (\w+) ] = "L<$0|\/type\/$0>";
        s:g[ 'L<' ~ '>' [ (.+) '|' (\w+)] ] = "L<$0|/type/$1>";
        $tainted = True;
    }
    $p.IO.spurt(@file.join("\n")) if $tainted;
    say "Changed: $p" if $tainted;
    $tainted-files{ ~$p }++ if $tainted;
}
'tainted-files'.IO.spurt: $tainted-files.keys.join("\n")

[JJ]

If there's some error, must not be a big one. It probably makes sense, and it will make also sense to eliminate the htmlify.p6 part that does this...