Look for "-with-pcre-regex..." within your phpinfo page. Many times PHP is configured with this flag, that messes up PCRE unicode support.

This was the case with my server. I compiled php without the flag and everything worked fine.

The flag "--enable-utf8" helps also

.

Deleted

Yes, my research led me to recompile PHP.

After speaking to the people who manage our server they say that custom PHP builds are not supported with their hosting software. Is there any other way to get the search to work without recompiling php or will I need to use a different server?

Maybe one of the C5 guys has a workaround? Good luck.

This is a shot in the dark, since I've never been on a server that's had this problem, but...

We're using the Zend Framework's lucene search indexer implementation, and I think if you use one other than the unicode you won't need this.

Check out concrete/libraries/indexed_search.php

Anywhere you see



change it to



You will also probably have to modify blocks/search/controller.php as well.

I changed the code in the two files and it got rid of the Utf8Num Analyer needs PCRE Unicode error message but now I receive the message "Unable to complete search: failed to open stream: No such file or directory". I get the same message when I try to reindex the search engine in the dashboard.

this doesn't change a lot.

I debugged a few things and here are a few hints:

- indexed_search: for some pages I had two characters for an umlaut. I could fix that with utf8_decode($this->getBodyContentFromPage($c))

- query string: the keyword I'm searching for needs to be encoded as well firefox passes it like "%F6ffentliche" this needs utf8_encode($q) in the controller

I my case, everything looks like proper utf8 text now but it still doesn't work! Lucene really bugs me, will try to find the problem after Christmas..

Hey Andrew,

I'm a bit late for Christmas but I fixed almost all the problems with lucene..

Looks like the right indexer isn't enough. It always truncated my text if there was a "character" like "ü".
(instead of "Gründung" the index contained only "Gr")

I was able to fix this problem with the patch I've attached.
Please note that you probably also have to set entities to raw in the tinymce configuration (entities: "raw") otherwise the index contains html entities like ü and not the proper unicode character ü!

It should work, but be careful with the charset.. ISO-8859-1 is not UTF-8 and as soon as someone uses my patch, this might cause some problems....



UPDATE: According to the documentation you could also use: setlocale(LC_ALL, 'de_DE.iso-8859-1') but I'm not a big fan of setlocale since it doesn't work on all the server outthere... It also works doesn't work per thread but rather per process which might change the locale even if you don't call setlocale just because another scripts within the same process called it...