[dokuwiki] darcs changes 2006-08-31
- From: andi@xxxxxxxxxxxxxx (Andreas Gohr)
- To: dokuwiki@xxxxxxxxxxxxx
- Date: Thu, 31 Aug 2006 04:00:02 +0200 (CEST)
Good Morning!
This are the darcs changes for DokuWiki committed
yesterday. Please test them and report bugs.
---------------------------------------------------------------------
Wed Aug 30 23:04:59 CEST 2006 Andreas Gohr <andi[at]splitbrain.org>
* conflict resolution for memory cache init
Wed Aug 30 20:27:53 CEST 2006 Ben Coburn <btcoburn[at]silicodon.net>
* scalable changelog redesign
This patch provides a rewritten changelog system that is designed to run
efficiently on both small and large wikis. The patch includes a plugin to
convert changelogs from the current format. The conversion is
non-destructive and happens automatically. For more information on the new
changelog format see "http://wiki.splitbrain.org/wiki:changelog";.
Structure
In short the changelog is now stored in per-page changelog files, with a
recent changes cache. The recent changes cache is kept in
"/data/meta/_dokuwiki.changes" and trimmed daily. The per-page changelogs
are kept in "/data/meta/<ns>/<page_id>.changes" files. To preserve
revision information for revisions stored in the attic, the "*.changes"
files are not removed when their page is deleted. This allows the full
life-cycle of page creation, deletion, and reversion to be tracked.
Format
The changelog line format now uses a general "line type" field in place of
the special "minor" change syntax. There is also an extra field that can
be used to store arbitrary data associated with special line types. The
reverted line type (R) is a good example. There the extra field holds the
revision date used as the source for reverting the page. See the wiki for
the complete syntax description.
Code Notes
The changelog functions have been rewritten to load the whole file only if
it is small. For larger files, the function loads only the relevant
chunk(s). Parsed changelog lines are cached in memory to speed future
function calls.
getRevisionInfo
A binary search is used to locate the chunk expected to contain the
requested revision. The whole chunk is parsed, and adjacent lines are
optimistically cached to speed consecutive calls.
getRevisions
Reads the changelog file backwards (newest first) in chunks until the
requested number of lines have been read. Parsed changelog lines are
cached for subsequent calls to getRevisionInfo. Because revisions are read
from the changelog they are no longer guaranteed to exist in the attic.
(Note: Even with lines of arbitrary length getRevisionInfo and
getRevisions never split changelog lines while reading. This is done by
sliding the "file pointer" forward to the end of a line after each blind
seek.)
isMinor
Removed. To detect a minor edit check the type as follows:
$parsed_logline['type']==='e';
(The old method signature is incompatible with the new changelog format
that stores an explicit logline type.)
parseChangelogLine
Added. Parses a line in the new changelog format into an associative
array. See the wiki for changelog syntax details.
array_dichotomic_search and hasTimeStamp
Removed because the new changelog system does not require them.
saveMetadata
Refactored and removed. This code was rolled into addLogEntry because the
meta data now caches the current log line, and because it was only being
called from saveWikiText (which calls addLogEntry).
html_revisions
Now paginates revisions with the same "page size" as recent changes.
INFO (global var)
The page metadata is now stored in the $INFO global as $INFO['meta']. The
parsed changelog line for the current page revision is stored in the
metadata as 'last_change', so to get the current revision
use $INFO['meta']['last_change'].
runTrimRecentChanges
Used in the indexer to limit the size of the recent changes changelog
cache. The cache is trimmed to the latest $conf['changes_days'] days, or
one "page" of changes on inactive wikis. The 'changes_days' setting is a
new configuration parameter added for this purpose. It takes an integer
value of days because the changelog cache is trimmed (at most) once every
24 hours. This function also triggers a private action event to import the
old changelog as needed.
importoldchangelog (plugin)
Used to import and convert the old changelog. Per-page changelog files are
created for all page-ids in the changelog. Recent changes are added to the
recent changes cache. The changelog importer also reconstructs changelog
information for revisions in the attic that are missing changelog lines.
These reconstructed changes use the default "reverted" summary string
surrounded with parentheses as their summary. For example "(old revision
restored)". Changes reconstructed from the attic will also be missing the
author's IP address, user name, and summary which were lost with the
missing changelog line. The importoldchangelog plugin will run
automatically without any need for user intervention. (Note: Very large
wikis running on servers with php in "safe mode" may run out of time
because the execution time limit can not be increased by the code.)
do=check (dokuwiki act)
The "check" action will now report if an old changelog exists and is being
or has been imported. If the import ran out of time, "check" should issue
a warning that the import did not finish. To try the import again remove
the "/data/meta/_dokuwiki.changes" and "/data/meta/_dokuwiki.changes_*"
files. (It is also possible to do the import on another server that is not
in "safe mode" and copy the "/data/meta" directory to the "safe mode"
server.)
NOTE TO WIKI USERS: Do not delete the "/data/meta" directory. It now
contains changelog information that can not be regenerated from the
other wiki data.
---------------------------------------------------------------------
Single patches can be downloaded from
http://dev.splitbrain.org/darcs/index.cgi/dokuwiki/?c=patches
Bye,
your darcs changlog mailer
--
DokuWiki mailing list - more info at
http://wiki.splitbrain.org/wiki:mailinglist
Other related posts:
