--- rxvt-unicode/src/perl/matcher	2021/07/14 12:39:57	1.36
+++ rxvt-unicode/src/perl/matcher	2022/12/09 05:06:46	1.40
@@ -21,7 +21,7 @@
 mouse button specified in the C<matcher.button> resource (default 2, or
 middle), the program specified in the C<matcher.launcher> resource
 (default, the C<url-launcher> resource, C<sensible-browser>) will be started
-with the matched text as first argument.  The default configuration is
+with the matched text as first argument. The default configuration is
 suitable for matching URLs and launching a web browser, like the
 former "mark-urls" extension.
 
@@ -31,7 +31,7 @@
 The launcher can also be overridden on a per-pattern basis.
 
 It is possible to activate the most recently seen match or a list of matches
-from the keyboard.  Simply bind a keysym to "matcher:last" or
+from the keyboard. Simply bind a keysym to "matcher:last" or
 "matcher:list" as seen in the example below.
 
 The C<matcher:select> action enables a mode in which it is possible to
@@ -39,7 +39,7 @@
 or copy them to the clipboard. While the mode is active, normal terminal
 input/output is suspended and the following bindings are recognized:
 
-=over 4
+=over
 
 =item C<Up>
 
@@ -88,14 +88,44 @@
     URxvt.matcher.pattern.2:  \\B(/\\S+?):(\\d+)(?=:|$)
     URxvt.matcher.launcher.2: gvim +$2 $1
 
+=head2 Regex encoding/wide character matching
+
+Urxvt stores all text as unicode, in a special encoding that uses
+one character/code point per column. For various reasons, the regular
+expressions are matched directly against this encoding, which means there are a few things
+you need to keep in mind:
+
+=over
+
+=item X resources/command line arguments are locale-encoded
+
+The regexes taken from the command line or resources will be converted
+from locale encoding to unicode. This can change the number of code points
+per character.
+
+=item Wide characters are column-padded with C<$urxvt::NOCHAR>
+
+Wide characters (such as kanji and sometimes tabs) are padded with
+a special character value (C<$urxvt::NOCHAR>). That means that
+constructs such as C<\w> or C<.> will only match part of a character, as
+C<$urxvt::NOCHAR> is not matched by C<\w> and both only match the first
+"column" of a wide character.
+
+That means you have to incorporate C<$urxvt::NOCHAR> into parts of regexes
+that may match wide characters. For example, to match C<\w+> you might
+want to use C<[\w$urxvt::NOCHAR]+> instead, and to match a single character
+(C<.>) you might want to use C<.$urxvt::NOCHAR*> instead.
+
+=back
+
 =cut
 
 my $url =
    qr{
       (?:https?://|ftp://|news://|mailto:|file://|\bwww\.)
-      [\w\-\@;\/?:&=%\$.+!*\x27,~#]*
+      [\w\-\@;\/?:&=%\$.+!*\x27,~#$urxvt::NOCHAR]*
       (
-         \([\w\-\@;\/?:&=%\$.+!*\x27,~#]*\)| # Allow a pair of matched parentheses
+         \([\w\-\@;\/?:&=%\$.+!*\x27,~#$urxvt::NOCHAR]*\)| # Allow a pair of matched parentheses
          [\w\-\@;\/?:&=%\$+*~]  # exclude some trailing characters (heuristic)
       )+
    }x;
@@ -195,16 +225,19 @@
    my ($self) = shift;
    my $row = $self->nrow - 1;
    my @exec;
+
    while ($row >= $self->top_row) {
       my $line = $self->line ($row);
-      @exec = $self->command_for($row);
-      last if(@exec);
+      @exec = $self->command_for ($row);
+      last if @exec;
 
       $row = $line->beg - 1;
    }
-   if(@exec) {
+
+   if (@exec) {
       return $self->exec_async (@exec);
    }
+
    ()
 }
 
@@ -253,9 +286,7 @@
 
    my @defaults = ($url);
    my @matchers;
-   for (my $idx = 0; defined (my $res = $self->my_resource ("pattern.$idx") || $defaults[$idx]); $idx++) {
-      $res = $self->locale_decode ($res);
-      utf8::encode $res;
+   for (my $idx = 0; defined (my $res = $self->locale_decode ($self->my_resource ("pattern.$idx")) || $defaults[$idx]); $idx++) {
       my $launcher = $self->my_resource ("launcher.$idx");
       $launcher =~ s/\$&|\$\{&\}/\${0}/g if $launcher;
       my $rend = $self->parse_rend($self->my_resource ("rend.$idx"));
@@ -315,15 +346,18 @@
          my @end = @+;
          my @exec;
 
-         if (!defined($off) || ($-[0] <= $off && $+[0] >= $off)) {
+         if (!(defined $off) || ($-[0] <= $off && $+[0] >= $off)) {
             if ($launcher !~ /\$/) {
                @exec = ($launcher, $match);
             } else {
                # It'd be nice to just access a list like ($&,$1,$2...),
                # but alas, m//g behaves differently in list context.
-               @exec = map { s/\$(\d+)|\$\{(\d+)\}/
-                  substr $text, $begin[$1 || $2], $end[$1 || $2] - $begin[$1 || $2]
-                  /egx; $_ } split /\s+/, $launcher;
+               @exec = map {
+                     s{\$(\d+)|\$\{(\d+)\}}{
+                        substr $text, $begin[$1 || $2], $end[$1 || $2] - $begin[$1 || $2]
+                     }egx;
+                     $_
+                  } split /\s+/, $launcher;
             }
 
             push @matches, [ $line->coord_of ($begin[0]), $line->coord_of ($end[0]), $match, @exec ];
@@ -331,7 +365,7 @@
       }
    }
 
-   @matches;
+   @matches
 }
 
 sub command_for {
@@ -348,8 +382,11 @@
 
 sub on_button_press {
    my ($self, $event) = @_;
-   if($self->valid_button($event)
-      && (my @exec = $self->command_for($event->{row},$event->{col}))) {
+
+   if (
+      $self->valid_button ($event)
+      && (my @exec = $self->command_for ($event->{row}, $event->{col}))
+   ) {
       $self->{row} = $event->{row};
       $self->{col} = $event->{col};
       $self->{cmd} = \@exec;
@@ -372,12 +409,13 @@
 
    return if !defined $row;
 
-   if($row == $event->{row} && abs($col-$event->{col}) < 2
-      && join("\x00", @$cmd) eq join("\x00", $self->command_for($row,$col))) {
-      if($self->valid_button($event)) {
-
+   if (
+      $row == $event->{row}
+      && (abs $col-$event->{col}) < 2
+      && (join "\x00", @$cmd) eq (join "\x00", $self->command_for ($row, $col))
+   ) {
+      if ($self->valid_button ($event)) {
          $self->exec_async (@$cmd);
-
       }
    }