Hello Lazarus-List,

Usually I have a lot of problems to find something valuable using the search engine of the wiki page at http://wiki.lazarus.freepascal.org and usually I get much better results using google to report the results from the wiki. So I had written (well, in fact modified) a firefox search plugin to be used to search the wiki. It is useful to me and maybe to other people and as it is quite small it is attached to this message.

To install close your FireFox and drop the file in equivalent path as this one:

"C:\Program files\Mozilla FireFox\searchplugins\"

Is there any possibility to get a copy of the current wiki database in order to test some indexing engines that may be added as another search procedure for the wiki ?

If you mean a dump of the current mysql db, then no, that is not possible, because it contains users, email addresses, passwords too.

Vincent

Vincent Snijders wrote

If you mean a dump of the current mysql db, then no, that is not possible, because it contains users, email addresses, passwords too.

Isn't it possible to make a dump without user information?

thanks,
dmitry

I don't know.

With user info:
mysqldump -u username --password=secretpassword wikidbname

What should it be without user information?

Vincent

mysqldump -u username --password=secretpassword wikidbname <tableA> <tableB> ...<tableN>

That way only tables you wish to dump will be dumped :)

I think btw (if I'm not mistaken from other db) that you can also use select statement to retrive only information you want/need in the dump.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20100115/16f3b9c9/attachment.html>

Hello Lazarus-List,

Friday, January 15, 2010, 7:00:34 PM, you wrote:

VS> If you mean a dump of the current mysql db, then no, that is not VS> possible, because it contains users, email addresses, passwords too.

Well, no, only text information (not needed the InnoDB table) and the needed fields to catch the URL to point to that page. An export of that tables, even partially to test is enoght.

If not possible I'll try to populate a wiki from other sources and test the search engines. I was thinking in the current wiki db to get a strait forward possible integration.

Vincent Snijders wrote

What should it be without user information?

Hmm... can the sensitive data be ...modified... on making the dump. i.e. user_password and user_email in users table?!

thanks,
dmitry

Hello Lazarus-List,

you wrote

> Well, no, only text information (not needed the InnoDB table) and the > needed fields to catch the URL to point to that page. An export of > that tables, even partially to test is enoght.

VS> Maybe you can use:
VS> http://wiki.lazarus.freepascal.org/Special:Export

That's fine for most purposes, but not for this case as a "search engine" that searchs across 20-100 pages is not a great challenge. The problem is when there are too many pages with the same terms here and there, which happends in the Laz/fpc wiki where things like "component" appears almost everywhere so a search like "create new component" will spot the rigth article at position 100-200 while a page like the FAQ appears almost always, whichever you search, at first position, maybe because its page creation date is one of the oldest :-? and contains almost any searchable term.

As the dump seems to be a not possible thing, no problem at all, I'll try to perform the experiment with other texts, maybe I can found something like a dump from any other database with many articles, or maybe a bunch of large PDFs talking about the same topic and if it works more or less as expected (I'm not sure about this) integrate it in an empty wiki and populate 10 articles to verify if it works.

Thank you anyway to everybody.

Hello Lazarus-List,

you wrote

> Well, no, only text information (not needed the InnoDB table) and the > needed fields to catch the URL to point to that page. An export of > that tables, even partially to test is enoght.

VS> Maybe you can use:
VS> http://wiki.lazarus.freepascal.org/Special:Export

After some test I had found that Sphinx indexer also have a plugin for MediaWiki. I don't know which version of MediaWiki is running in the Lazarus site but it should work with something newer than 1.10 version (included).

Also it can be installed as a test page so the actual search mode is not replaced and both engines are available.

Would be great if somebody can install it at least for test as I'm quite sure it will report much better results than the current one.

http://www.mediawiki.org/wiki/Extension:SphinxSearch

Complete installation guide:
http://www.mediawiki.org/wiki/Extension:SphinxSearch#Installation_Instructions

Hello Lazarus-List,

Saturday, January 16, 2010, 6:59:34 PM, you wrote:

VS> That is bad news, we are :
VS> http://wiki.lazarus.freepascal.org/Special:Version

1.9.3 has been reported to works also:

1.8.? -- Fails -
1.9.3 - Works - (Gri6507)

Maybe its time to also update :)