Search not working

[Migrated From Jforum.net]

Hi, I would like to install JForum on my intranet. JForum has a nice clean look and fast - great job!! Everything seems to work except I have not been able to get the search function working. I have not installed the CVS version yet. Do you have any suggestions?

I am using Weblogic 8.1
MySQL v4.1.9
Java 1.4.2_05

thanks,
Jon

[originally posted on jforum.net by jcoppert]

[Migrated From Jforum.net]

hhmm.. the search is a bit unstable, but it was supposed to work. Maybe this is another problem related to mysql 4.1 (just a guess). Please see if jforum_search_words and jforum_search_wordmatch have any records.
The code which handles that is src/net/jforum/drivers/generic/SearchModel.java.

There are some people trying to integrate Lucene with jforum, but I still don't have a date, since most of the code is for test purposes so far.

Rafael
[originally posted on jforum.net by Rafael Steil]

[Migrated From Jforum.net]

Thanks for the pointers. I checked the tables and they appear populated correctly. If the search works like it does here I'm happy for now! I haven't had much chance to look at the code yet.....
[originally posted on jforum.net by jcoppert]

[Migrated From Jforum.net]

Hi,

Is there any improvements on this point?
Seems that problem still exits in Tomcat 4.1, MySql 4.1.9 and JForum 2.1

Jon, did you found any solutions?

Thanks in advance,
Gleb

p.s. jforum_search_words and jforum_search_wordmatch populated well :?

?�?��?�?�?�?�?�
[originally posted on jforum.net by gst]

[Migrated From Jforum.net]

I have tested it using mysql 4.1 and it is working for me. If some of your guys are able to debug the application, it will help a lot, since I cannot reproduce the bug.

Rafael
[originally posted on jforum.net by Rafael Steil]

[Migrated From Jforum.net]

Ah, yes, indeed. It does not work well with cyrillic chars. I have no official fix for it so far.

Rafael
[originally posted on jforum.net by Rafael Steil]

[Migrated From Jforum.net]

Looks like it fails with unicode chars but it definatelly works with ascii.

When I'm trying to find some cyrillic text (I added some in my prev message) it returns nothing - even this jforum.net version ;)
[originally posted on jforum.net by gst]

[Migrated From Jforum.net]

mysql 4.1.x tomcat 5.0.x jforum 2.1.3

the same trouble occured.

maybe it need call method getBytes()
i will try it later.
[originally posted on jforum.net by cnticws]

[Migrated From Jforum.net]

Hi guys,

Any progress on making search to work with unicode? (It fails not only with cyrillic, it can't find for example "esṭo" from https://coderanch.com/t/574757 ;) )

Rafael, do you have an idea on where to start looking?

Thanks,
Gleb
[originally posted on jforum.net by gst]

[Migrated From Jforum.net]

Yes. Once someone said that chaning the code

if (word.length() < 3) {
    continue;
}

to

if (word.length() < 1) {
    continue;
}

in src/net/jforum/drivers/generic/SearchModel.java

would fix the bug. You can try it.

Rafael
[originally posted on jforum.net by Rafael Steil]

[Migrated From Jforum.net]

Yep... that's our guy ! :P

Rafael
[originally posted on jforum.net by Rafael Steil]

[Migrated From Jforum.net]

Yes, that someone is me.
see this post for detail
https://coderanch.com/t/574739
[originally posted on jforum.net by andowson]

[Migrated From Jforum.net]

Thanks for the tip, but I believe described solution is not for my case, because I have jforum_search_words populated OK
(and also I want to searh on existing posts ;) )

I'm almost sure that it is something with text encoding.
I tried to debug and outputted to the system console pm.getPost().getText() from PrivateMessageAction and kw from SearchAction.
When using the same word in both cases console output looks different, i.e. text is encoded differently.

UPD. He-he... The difference between search action and other actions like private message is that search form uses GET method!


[originally posted on jforum.net by gst]

[Migrated From Jforum.net]

You can not rely on the console output because console encoding is different for web, specially on Windows.

Maybe it's a problem with the query, which use LIKE / = 'xxx' to check for matches. For me is virtually impossible to test it. I can guide you to the source code, if you want.

Rafael
[originally posted on jforum.net by Rafael Steil]

[Migrated From Jforum.net]

As I understand the problem is that when submitting form data by GET, search string gets encoded with default ISO-8859-1 rather than supplied content charset (which is correct in POST case).

This code change in SearchAction helps:

this.kw = this.addSlashes(this.request.getParameter("search_keywords"));

to

String keywords = new String(this.request.getParameter("search_keywords").getBytes("ISO-8859-1"), SystemGlobals.getValue(ConfigKeys.ENCODING));
this.kw = this.addSlashes(keywords);

My search is working now

Cheers,
Gleb
[originally posted on jforum.net by gst]

[Migrated From Jforum.net]

Btw, Rafael, you mentioned

Probably I'll put some configuration option to these issues.. something like "search.index.all.words" and "seach.use.like".

Is this planned for the next release? ;)
[originally posted on jforum.net by gst]