1 # Copyright (C) 2020-2021 all contributors <meta@public-inbox.org>
2 # License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt>
4 # *-external commands of lei
5 package PublicInbox::LeiExternal;
8 use PublicInbox::Config;
11 my ($self, $cb, @arg) = @_;
12 my $cfg = $self->_lei_cfg;
14 for my $sec (grep(/\Aexternal\./, @{$cfg->{-section_order}})) {
15 my $loc = substr($sec, length('external.'));
16 $boost{$loc} = $cfg->{"$sec.boost"};
18 return \%boost if !wantarray && !$cb;
20 # highest boost first, but stable for alphabetic tie break
22 my @order = sort { $boost{$b} <=> $boost{$a} } sort keys %boost;
23 if (ref($cb) eq 'CODE') {
24 for my $loc (@order) {
25 $cb->(@arg, $loc, $boost{$loc});
27 } elsif (ref($cb) eq 'HASH') {
30 @order; # scalar or array
33 sub ext_canonicalize {
34 my $location = $_[-1]; # $_[0] may be $lei
35 if ($location !~ m!\Ahttps?://!) {
36 PublicInbox::Config::rel2abs_collapsed($location);
39 my $uri = URI->new($location)->canonical;
40 my $path = $uri->path . '/';
41 $path =~ tr!/!/!s; # squeeze redundant '/'
47 # TODO: we will probably extract glob2re into a separate module for
48 # PublicInbox::Filter::Base and maybe other places
49 my %re_map = ( '*' => '[^/]*?', '?' => '[^/]',
50 '[' => '[', ']' => ']', ',' => ',' );
53 my $re = $_[-1]; # $_[0] may be $lei
57 my $schema_host_port = '';
59 # don't glob URL-looking things that look like IPv6
60 if ($re =~ s!\A([a-z0-9\+]+://\[[a-f0-9\:]+\](?::[0-9]+)?/)!!i) {
61 $schema_host_port = quotemeta $1; # "http://[::1]:1234"
63 my $changes = ($re =~ s!(.)!
64 $re_map{$p eq '\\' ? '' : do {
65 if ($1 eq '[') { ++$in_bracket }
66 elsif ($1 eq ']') { --$in_bracket }
67 elsif ($1 eq ',') { ++$qm } # no change
71 ($p eq '-' && $in_bracket) ? $p : (++$qm, "\Q$p")
73 # bashism (also supported by curl): {a,b,c} => (a|b|c)
74 $changes += ($re =~ s/([^\\]*)\\\{([^,]*,[^\\]*)\\\}/
75 (my $in_braces = $2) =~ tr!,!|!;
78 ($changes - $qm) ? $schema_host_port.$re : undef;
81 # get canonicalized externals list matching $loc
82 # $is_exclude denotes it's for --exclude
83 # otherwise it's for --only/--include is assumed
85 my ($self, $loc, $is_exclude) = @_;
86 return (ext_canonicalize($loc)) if -e $loc;
88 my @cur = externals_each($self);
89 my $do_glob = !$self->{opt}->{globoff}; # glob by default
90 if ($do_glob && (my $re = glob2re($loc))) {
91 @m = grep(m!$re!, @cur);
92 return @m if scalar(@m);
93 } elsif (index($loc, '/') < 0) { # exact basename match:
94 @m = grep(m!/\Q$loc\E/?\z!, @cur);
95 return @m if scalar(@m) == 1;
96 } elsif ($is_exclude) { # URL, maybe:
97 my $canon = ext_canonicalize($loc);
98 @m = grep(m!\A\Q$canon\E\z!, @cur);
99 return @m if scalar(@m) == 1;
101 return (ext_canonicalize($loc));
103 if (scalar(@m) == 0) {
104 die "`$loc' is unknown\n";
106 die("`$loc' is ambiguous:\n", map { "\t$_\n" } @m, "\n");
110 sub canonicalize_excludes {
111 my ($lei, $excludes) = @_;
113 for my $loc (@$excludes) {
114 my @l = get_externals($lei, $loc, 1);
120 # returns an anonymous sub which returns an array of potential results
121 sub complete_url_prepare {
122 my $argv = $_[-1]; # $_[0] may be $lei
123 # Workaround bash word-splitting URLs to ['https', ':', '//' ...]
124 # Maybe there's a better way to go about this in
125 # contrib/completion/lei-completion.bash
127 my $cur = pop(@$argv) // '';
130 if ($cur eq ':' && @x) {
134 while (@x > 2 && $x[0] !~ /\A(?:http|nntp|imap)s?\z/i &&
138 if (@x >= 2) { # qw(https : hostname : 443) or qw(http :)
140 } else { # just filter out the flags and hope for the best
141 $re = join('', grep(!/^-/, @$argv));
143 $re = quotemeta($re);
146 # the "//;" here (for AUTH=ANONYMOUS) interacts badly with
147 # bash tab completion, strip it out for now since our commands
148 # work w/o it. Not sure if there's a better solution...
149 $_[0] =~ s!//;AUTH=ANONYMOUS\@!//!i;
151 # only return the part specified on the CLI
152 # don't duplicate if already 100% completed
153 $_[0] =~ /\A$re(\Q$cur\E.*)/ ? ($cur eq $1 ? () : $1) : ()
155 wantarray ? ($re, $cur, $match_cb) : $match_cb;