This changes how DokuWiki looks for reference toa media file which is
about to deleted. Instead of doing a full grep through all pages it now
uses the fulltext index first, then does an exact match on the found
pages.

This speeds up the search significantly on larger wikis. However the
fulltext search limits now apply: images with names shorter than 3 charcters
may not be found.

This needs extensive testing!

darcs-hash:20080223205254-7ad00-486de0a4125d51b4e7999827f710d1d9de8bc60d.gz

This patch makes the highlighting of phrases in search snippets and on
the pages itself much better.

Now a regexp gets passed to the ?s

darcs-hash:20080215174653-7ad00-cd2d6f7d408db7b7dd3cb9974c3eb27f3a9baeac.gz

bool page_exists($id, $rev

darcs-hash:20070930021040-d26fc-e3847bfdd20a36154685262eca94211cfd461e83.gz

The use of realpath() to clean up relative file names caused some
trouble in certain setups relying on symlinks or having restricitve
file structure setups.

This patch replaces all realpath() calls with a PHP only replacement
which should solve those problems.

darcs-hash:20070930184250-7ad00-512ff04c95f57fc9eaf104f80372237a3c94286f.gz

darcs-hash:20070804081226-7ad00-a8e7127c7122a96f9817158d87e1a364d8cdbc9f.gz

darcs-hash:20070718104839-7ad00-50348c1834c78e891f049023d2e8894d6bb0a00b.gz

darcs-hash:20070514222527-d5083-53ed619daf07d0a84c52161465d163abf1400529.gz

darcs-hash:20070330215042-19e2d-3528f2412ff044eb45158f349db5bbb5e32d907b.gz

darcs-hash:20070303220143-7ad00-5d592dbebaae371c03102b20ae7e0d9e433b378b.gz

darcs-hash:20070205191848-7ad00-77ad5a398534a7a64884e155c4607350e0f25a7c.gz

darcs-hash:20061124215413-7ad00-f2bd46b7edf70660cc3e0274bd222eafba1edbc6.gz

A modification to the indexer that sorts words based on length. This should make
searching a little bit more efficient. After the patch is applied, your old index
will be automatically converted to the new format (when you visit a page). The
new index format is:

1. Index files are stored in savedir/index
2. Word lists are stored as wlen.idx. This used to be word.idx.
3. Word indexes are stored as ilen.idx. This used to be index.idx.
4. The page list, page.idx, is simply copied to the new location.

Any plugins you have, such as the blog plugin, that read the index files need to
be updated.

darcs-hash:20061112194900-2b9f0-a975498ccf0a1d39c6df73b79bcd028d5e81c389.gz

- add deaccented and romanised page names to index word list
- remove stop words from tokens used in backlink search

darcs-hash:20061105195453-9b6ab-6c4989eb75782af60a3de3bddbc99a83de2b4c80.gz

This patch removes some commented code fragments and alternative
snippet generators

darcs-hash:20061008090624-7ad00-14bfee2ded6c6c8ef43ad02a4c02a5d95ee9daf7.gz

- rework utf8_substr() NOMBSTRING code to always use pcre
- remove work around for utf8_substr() and large strings from ft_snippet()

darcs-hash:20060928165122-9b6ab-0eefc216f07f9d7e7d8eb62ce26605c28ee340fa.gz

This patch primarily updates p_cached_xhtml() and p_cached_instructions() to
allow their caching logic to be surrounded by an event trigger.

p_cached_xhtml() has been rewritten as the more general p_cached_output() to
support other render output formats besides 'xhtml'. All calls to
p_cached_xhtml() have been changed to refer to the new function.

New event:

name:        PARSER_CACHE_USE
data:        cache object (see below)
action:      determine if cache file can be used
preventable: yes
result:      bool, true to use cache file, false otherwise

Cache operations have been generalised in a new class, cache, extended to
cache_parser, cache_renderer & cache_instructions. Details can be found in
inc/cache.php

For handling of above event, key properties are:
- page, if present the wiki page id,
        may not always be present, e.g. when called for locale xhtml files
- file, source file
- mode, renderer mode (e.g. 'xhtml') or 'i' for instructions

Other changes:
- cache class counts cache hits against attempts, results are stored in
  {cache_dir}/cache_stats.txt
- adds metadata dependency to renderer page cache
- replaces purgefile dependency for renderer cache with metadata
  'relation references' (internal link) dependency for wiki pages only

darcs-hash:20060911021418-9b6ab-19601ed194b8c8e45236ab72c3e23d78bf777e6c.gz

darcs-hash:20060901002016-9b6ab-716518138edf541a869510d7c2934b9474547fc3.gz

darcs-hash:20060831092146-9b6ab-b00aa29c982ab18117f476b3d01d5111915c9d4b.gz

ft_snippet()
- make utf8 algorithm default
- add workaround for utf8_substr() limitations, bug #891
- fix some indexes which missed out on conversion to utf8
  character counts
- minor improvements

idx_lookup()
- minor changes to wildcard matching code to improve performance
  (changes based on profiling results)

utf8
- specifically set mb_internal_coding to utf-8 when mb_string
  functions will be used.

darcs-hash:20060831003413-9b6ab-712021eda3c959ffe79d8d3fe91d2c9a8acf2b58.gz

 - correct "opt1" algorithm for multibyte utf8
 - minor improvement to "opt2" for short pages
 - add "utf8" algorithm, this algorithm endeavours
   to work with whole utf8 character as much as
   possible.  The resulting snippet will tend to
   100 characters, rather than the 100 bytes of
   "opt1" and "opt2".

darcs-hash:20060826234333-9b6ab-ae4c60c8855a92b133cb8d5a230098203f610e7b.gz

darcs-hash:20060826095311-9b6ab-9a6f272cc7c7532eb2bad8f7b4404c5a16b71109.gz

This adds code to remove or replace invalid UTF-8 bytes and uses it
in the ft_snippets function.

darcs-hash:20060826082919-7ad00-a94004de159ae93ff5b7270fd3e631ff467233cd.gz

darcs-hash:20060825134730-9b6ab-086ee0647af39c4398cf1726324d8215722a39db.gz

This patch includes two alternative algorithms for ft_snippet(), the code
which prepares the snippets seen on the search page - and the most time
consuming part of the production of that page.

If you have $conf['allowdebug'] on, you can specify the search algorithm to
use by adding &_search

darcs-hash:20060825104046-9b6ab-942d81a43cf0f85bfd235cabf6c35dd4b20e0b71.gz

    - now its possible to restrict the fulltext-search to
      multible namespaces

      Examples:

        searchword @ns1 @ns2 @ns3

        "exact phrase" @ns1 @ns2 @ns3

darcs-hash:20060518204647-484ab-061521a81f13360e33496e5163e3cd263a9c1ad6.gz

    - The fulltext-search can now be restricted to a given
      namespace seperated by an "@"

darcs-hash:20060518161855-484ab-1617b6d2c3593525f4d29a789b0a32ebf414b9ae.gz

This patch cleans up the source code to satisfy the coding guidelines (see
http://wiki.splitbrain.org/wiki:development#coding_style)

It converts files to UNIX lineendings and removes tabs and trailing
whitespace. Not all files were cleaned yet.

darcs-hash:20060217222040-7ad00-bba3d2bee3b5aa7cbb5184258abd50805cd071bf.gz

Now searching for word parts is possible by adding or prepending a *
character to the searchword:

'foo*' searches for words beginning with 'foo' eg. 'foobar'
'*foo' looks for words ending in 'foo' eg. 'barfoo'
'*foo*' gets anything with 'foo' in it eg. 'barfoobaz'

darcs-hash:20051127180723-7ad00-1eb29e812ddaf38d9812697bb1cffffe9a5fb330.gz

This new option accepts a RegExp to filter certain pages from all automatic
listings (RSS, recent changes, search results, index). This is useful to
exclude certain pages like the ones used in the sitebar templates. The
regexp is matched against the full page ID with a leading colon. If it
matches the page is assumed to be a hidden one.

IMPORTANT: this is not related to ACL. A hidden page is still visible to all
users (if not restricted by ACL) when linked or called directly.

darcs-hash:20051103101726-6e07b-8d45912a1b4f6cfc9e3fce147c15f84a58ea7ca2.gz

The new handling of asian chars as single words needs a recent PCRE library
(PHP 4.3.10 is known work). If this support isn't available the regexp
compilation will fail. This patch adds a workaround - this means the search
will not work as expected with asian words on older PHP versions.

darcs-hash:20051009124833-7ad00-1319829be5cb73246e13eb65e4c950d43c6ce5bf.gz

Asian languages do not use spaces to seperate words. The indexer however does
a word based lookup. Splitting for example Japanese texts into real words is
only possible with complicated natural language processing, something
completely out of scope for DokuWiki.

This patch solves the problem by treating all asian characters as single
words. When an asian word (consisting of multiple characters) is searched it
is treated as a phrase search, looking up each charcter by it self first,
then checking for the phrase in found documents.

darcs-hash:20050925175451-7ad00-933b33b51b5f2fa05e736c18b8db58a5fdbf41ce.gz

darcs-hash:20050925102211-7ad00-200edd676ba3956f03ec5bcc5149d4aa4bd15e24.gz

darcs-hash:20050921195118-7ad00-9070166cbaa26e3f27f7b92382346a70f5c479a1.gz

darcs-hash:20050912143027-7ad00-b2f3165d8db7122a453ecc63ad031af4467f691f.gz

darcs-hash:20050912141042-7ad00-5ef43525c9fd7ba44206720c54bb566450f93250.gz

darcs-hash:20050903220229-7ad00-5d95f905eaeb3f6b867aa3ee43c2a8bccc533c00.gz

The new search function was added but is not yet integrated into
DokuWikis interface.

darcs-hash:20050828152821-7ad00-a6e79a9dc5aaf41c547cf42dccdbc3b5bc8d303e.gz