Avoid %links accumulating duplicates. (For TOVA)

This is sorta an optimisation, and sorta a bug fix. In one
test case I have available, it can speed a page build up from 3
minutes to 3 seconds.

The root of the problem is that $links{$page} contains arrays of
links, rather than hashes of links. And when a link is found,
it is just pushed onto the array, without checking for dups.

Now, the array is emptied before scanning a page, so there
should not be a lot of opportunity for lots of duplicate links
to pile up in it. But, in some cases, they can, and if there
are hundreds of duplicate links in the array, then scanning it
for matching links, as match_link and some other code does,
becomes much more expensive than it needs to be.

Perhaps the real right fix would be to change the data structure
to a hash. But, the list of links is never accessed like that,
you always want to iterate through it.

I also looked at deduping the list in saveindex, but that does
a lot of unnecessary work, and doesn't completly solve the problem.

So, finally, I decided to add an add_link function that handles deduping,
and make ikiwiki-transition remove the old dup links.

11 files changed, 70 insertions, 17 deletions

diff --git a/IkiWiki.pm b/IkiWiki.pm
index e260fffea..e6efe1889 100644
--- a/IkiWiki.pm
+++ b/IkiWiki.pm
@@ -21,12 +21,12 @@ our @EXPORT = qw(hook debug error template htmlpage add_depends pagespec_match
                  pagespec_match_list bestlink htmllink readfile writefile
 		 pagetype srcfile pagename displaytime will_render gettext urlto
 		 targetpage add_underlay pagetitle titlepage linkpage
-		 newpagefile inject
+		 newpagefile inject add_link
                  %config %links %pagestate %wikistate %renderedfiles
                  %pagesources %destsources);
 our $VERSION = 3.00; # plugin interface version, next is ikiwiki version
 our $version='unknown'; # VERSION_AUTOREPLACE done by Makefile, DNE
-our $installdir=''; # INSTALLDIR_AUTOREPLACE done by Makefile, DNE
+our $installdir='/usr'; # INSTALLDIR_AUTOREPLACE done by Makefile, DNE
 
 # Optimisation.
 use Memoize;
@@ -1757,6 +1757,14 @@ sub inject {
 	use warnings;
 }
 
+sub add_link ($$) {
+	my $page=shift;
+	my $link=shift;
+
+	push @{$links{$page}}, $link
+		unless grep { $_ eq $link } @{$links{$page}};
+}
+
 sub pagespec_merge ($$) {
 	my $a=shift;
 	my $b=shift;
diff --git a/IkiWiki/Plugin/camelcase.pm b/IkiWiki/Plugin/camelcase.pm
index 74a8397d7..088447d6b 100644
--- a/IkiWiki/Plugin/camelcase.pm
+++ b/IkiWiki/Plugin/camelcase.pm
@@ -61,7 +61,7 @@ sub scan (@) {
         my $content=$params{content};
 
 	while ($content =~ /$link_regexp/g) {
-		push @{$links{$page}}, linkpage($1) unless ignored($1)
+		add_link($page, linkpage($1)) unless ignored($1)
 	}
 }
 
diff --git a/IkiWiki/Plugin/img.pm b/IkiWiki/Plugin/img.pm
index d295b833b..a697fea19 100644
--- a/IkiWiki/Plugin/img.pm
+++ b/IkiWiki/Plugin/img.pm
@@ -43,7 +43,7 @@ sub preprocess (@) {
 		return '';
 	}
 
-	push @{$links{$params{page}}}, $image;
+	add_link($params{page}, $image);
 	# optimisation: detect scan mode, and avoid generating the image
 	if (! defined wantarray) {
 		return;
diff --git a/IkiWiki/Plugin/link.pm b/IkiWiki/Plugin/link.pm
index b79273f96..4c1add985 100644
--- a/IkiWiki/Plugin/link.pm
+++ b/IkiWiki/Plugin/link.pm
@@ -86,7 +86,7 @@ sub scan (@) {
 	my $content=$params{content};
 
 	while ($content =~ /(?<!\\)$link_regexp/g) {
-		push @{$links{$page}}, linkpage($2);
+		add_link($page, linkpage($2));
 	}
 }
 
diff --git a/IkiWiki/Plugin/meta.pm b/IkiWiki/Plugin/meta.pm
index 4a22fed30..cc5455d64 100644
--- a/IkiWiki/Plugin/meta.pm
+++ b/IkiWiki/Plugin/meta.pm
@@ -110,7 +110,7 @@ sub preprocess (@) {
 	}
 	elsif ($key eq 'link' && ! %params) {
 		# hidden WikiLink
-		push @{$links{$page}}, $value;
+		add_link($page, $value);
 		return "";
 	}
 	elsif ($key eq 'author') {
diff --git a/IkiWiki/Plugin/tag.pm b/IkiWiki/Plugin/tag.pm
index 8fe9c6828..cdcfaf536 100644
--- a/IkiWiki/Plugin/tag.pm
+++ b/IkiWiki/Plugin/tag.pm
@@ -73,7 +73,7 @@ sub preprocess_tag (@) {
 		$tag=linkpage($tag);
 		$tags{$page}{$tag}=1;
 		# hidden WikiLink
-		push @{$links{$page}}, tagpage($tag);
+		add_link($page, tagpage($tag));
 	}
 		
 	return "";
@@ -88,14 +88,14 @@ sub preprocess_taglink (@) {
 		if (/(.*)\|(.*)/) {
 			my $tag=linkpage($2);
 			$tags{$params{page}}{$tag}=1;
-			push @{$links{$params{page}}}, tagpage($tag);
+			add_link($params{page}, tagpage($tag));
 			return taglink($params{page}, $params{destpage}, $tag,
 				linktext => pagetitle($1));
 		}
 		else {
 			my $tag=linkpage($_);
 			$tags{$params{page}}{$tag}=1;
-			push @{$links{$params{page}}}, tagpage($tag);
+			add_link($params{page}, tagpage($tag));
 			return taglink($params{page}, $params{destpage}, $tag);
 		}
 	}
diff --git a/debian/NEWS b/debian/NEWS
index 22513cc4a..62e1543b3 100644
--- a/debian/NEWS
+++ b/debian/NEWS
@@ -1,3 +1,12 @@
+ikiwiki (3.12) UNRELEASED; urgency=low
+
+  You may want to run `ikiwiki-transition deduplinks /path/to/srcdir`
+  after upgrading to this version of ikiwiki. This command will
+  optimise your wiki's saved state, removing duplicate information
+  that can slow ikiwiki down.
+
+ -- Joey Hess <joeyh@debian.org>  Wed, 06 May 2009 00:25:06 -0400
+
 ikiwiki (3.01) unstable; urgency=low
 
   If your wiki uses git, and you have a `diffurl` configured in
diff --git a/debian/changelog b/debian/changelog
index 9244ccdb8..1e1d48d78 100644
--- a/debian/changelog
+++ b/debian/changelog
@@ -5,6 +5,10 @@ ikiwiki (3.12) UNRELEASED; urgency=low
     fails on nonexistant directories with some broken perl
     versions.
   * inline: Minor optimisation.
+  * add_link: New function, which plugins should use rather than
+    modifying %links directly, to avoid it accumulating duplicates.
+  * ikiwiki-transition: Add a deduplinks action, that can be used
+    to remove duplicate links and optimise a wiki w/o rebuilding it.
 
  -- Joey Hess <joeyh@debian.org>  Mon, 04 May 2009 19:17:39 -0400
 
diff --git a/doc/ikiwiki-transition.mdwn b/doc/ikiwiki-transition.mdwn
index 18836d5f5..e0b853ecf 100644
--- a/doc/ikiwiki-transition.mdwn
+++ b/doc/ikiwiki-transition.mdwn
@@ -61,6 +61,13 @@ If this is not done explicitly, a user's plaintext password will be
 automatically converted to a hash when a user logs in for the first time
 after upgrade to ikiwiki 2.48.
 
+# deduplinks srcdir
+
+In the past, bugs in ikiwiki have allowed duplicate link information
+to be stored in its indexdb. This mode removes such duplicate information,
+which may speed up wikis afflicted by it. Note that rebuilding the wiki
+will have the same effect.
+
 # AUTHOR
 
 Josh Triplett <josh@freedesktop.org>, Joey Hess <joey@ikiwiki.info>
diff --git a/doc/plugins/write.mdwn b/doc/plugins/write.mdwn
index 0b358b215..28da243d5 100644
--- a/doc/plugins/write.mdwn
+++ b/doc/plugins/write.mdwn
@@ -107,8 +107,8 @@ adding or removing files from it.
 
 This hook is called early in the process of building the wiki, and is used
 as a first pass scan of the page, to collect metadata about the page. It's
-mostly used to scan the page for [[WikiLinks|ikiwiki/WikiLink]], and add them to `%links`.
-Present in IkiWiki 2.40 and later.
+mostly used to scan the page for [[WikiLinks|ikiwiki/WikiLink]], and add
+them to `%links`. Present in IkiWiki 2.40 and later.
 
 The function is passed named parameters "page" and "content". Its return
 value is ignored.
@@ -151,11 +151,11 @@ parameter is set to a true value if the page is being previewed.
 If `hook` is passed an optional "scan" parameter, set to a true value, this
 makes the hook be called during the preliminary scan that ikiwiki makes of
 updated pages, before begining to render pages. This should be done if the
-hook modifies data in `%links`. Note that doing so will make the hook be
-run twice per page build, so avoid doing it for expensive hooks. (As an
-optimisation, if your preprocessor hook is called in a void context, you
-can assume it's being run in scan mode, and avoid doing expensive things at
-that point.)
+hook modifies data in `%links` (typically by calling `add_link`). Note that
+doing so will make the hook be run twice per page build, so avoid doing it
+for expensive hooks. (As an optimisation, if your preprocessor hook is
+called in a void context, you can assume it's being run in scan mode, and
+avoid doing expensive things at that point.)
 
 Note that if the [[htmlscrubber]] is enabled, html in
 preprocessor [[ikiwiki/directive]] output is sanitised, which may limit what
@@ -174,7 +174,8 @@ links. The function is passed named parameters "page", "destpage", and
 and later.
 
 Plugins that implement linkify must also implement a scan hook, that scans
-for the links on the page and adds them to `%links`.
+for the links on the page and adds them to `%links` (typically by calling
+`add_link`).
 
 ### htmlize
 
@@ -753,6 +754,11 @@ Optionally, a third parameter can be passed, to specify the preferred
 filename of the page. For example, `targetpage("foo", "rss", "feed")`
 will yield something like `foo/feed.rss`.
 
+#### `add_link($$)`
+
+This adds a link to `%links`, ensuring that duplicate links are not
+added. Pass it the page that contains the link, and the link text.
+
 ## Miscellaneous
 
 ### Internal use pages
diff --git a/ikiwiki-transition b/ikiwiki-transition
index 599261a09..f17868d73 100755
--- a/ikiwiki-transition
+++ b/ikiwiki-transition
@@ -220,6 +220,21 @@ sub moveprefs {
 	IkiWiki::Setup::dump($setup);
 }
 
+sub deduplinks {
+	my $dir=shift;
+	if (! defined $dir) {
+		usage();		
+	}
+	$config{wikistatedir}=$dir."/.ikiwiki";
+	IkiWiki::loadindex();
+	foreach my $page (keys %links) {
+		my %l;
+		$l{$_}=1 foreach @{$links{$page}};
+		$links{$page}=[keys %l]
+	}
+	IkiWiki::saveindex();
+}
+
 sub usage {
 	print STDERR "Usage: ikiwiki-transition type ...\n";
 	print STDERR "Currently supported transition subcommands:\n";
@@ -229,6 +244,7 @@ sub usage {
 	print STDERR "\tmoveprefs setupfile\n";
 	print STDERR "\thashpassword srcdir\n";
 	print STDERR "\tindexdb srcdir\n";
+	print STDERR "\tdeduplinks srcdir\n";
 	exit 1;
 }
 
@@ -253,6 +269,9 @@ elsif ($mode eq 'setupformat') {
 elsif ($mode eq 'moveprefs') {
 	moveprefs(@ARGV);
 }
+elsif ($mode eq 'deduplinks') {
+	deduplinks(@ARGV);
+}
 else {
 	usage();
 }