Wildcards in Lucene Searches

[dbevacqua]

Hi

This follows on from an earlier thread which wasn't resolved.

I'm having difficulty getting wildcard searches to work. For example:

@cm\:name:heal*‍

I've traced through the relevant code and the problem seems to be in org.alfresco.repo.search.impl.lucene.QueryParser, around line 720:

  final public Query Term(String field) throws ParseException {
  Token term, boost=null, fuzzySlop=null, goop1, goop2;
  boolean prefix = false;
  boolean wildcard = false;
  boolean fuzzy = false;
  boolean rangein = false;
  Query q;
    switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
    case TERM:
    case PREFIXTERM:
    case WILDTERM:
    case NUMBER:
      switch ((jj_ntk==-1)?jj_ntk():jj_ntk) {
      case TERM:
        term = jj_consume_token(TERM);
        break;
      case PREFIXTERM:
        term = jj_consume_token(PREFIXTERM);
                             prefix=true;
        break;
      case WILDTERM:
        term = jj_consume_token(WILDTERM);
                           wildcard=true;‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍

in jj_ntk() when the token is 'heal*', the kind is (incorrectly) identified as 19 (PREFIXTERM), not WILDTERM.

Or am I missing something?

This is a real problem for us. I am prepared to admit it is something I am doing but I'd appreciate some guidance on the matter if anyone has any ideas.

Thanks

Dominic

[dbevacqua]

I should add that the result is that this line

         q = getFieldQuery(field, analyzer, term.image.substring(1, term.image.length()-1), s);‍

is not called when there is a wildcard, but is called when there isn't.

[dbevacqua]

ok please ignore the above, and if you are near enough and appropriately armed, shoot me.

Complete red herring - having just RTFM for PrefixQuery I see that it has correctly identified it. I should never really have doubted JavaCC…

Still no results returned though

[dbevacqua]

Hi

Having looked through the source for LuceneTest and Lucene2Test, I can't see any tests of the wildcard searches. Could somebody point me at one please?

Thanks

Dominic

[dbevacqua]

ok finally got a wildcard search to work by using the full QName of the property:

(escaping removed for readability)

@{ourNamespace}name:health*
20 results

@{ourNamespace}name:health
19 results

@{ourNamespace}name:healthy
1 result

Hope this sheds some light and reassures you that I am not totally insane.

Any idea why the short version is only working when wildcards not given? Something peculiar about our namespace name or the local name?

Dominic

[kevinr]

Yes the full form uri must be used for a lucene search - sorry we should have spotted this earlier when you posted it. I'm not sure when the short form works if at all I would suggest you stick with the long form - that's what we use when generating searches in the web-client.

For instance, this is the query I generate when the user wants to search for content with the name of "myfile.txt":

 TYPE:"{http://www.alfresco.org/model/content/1.0}content" AND @\{http\://www.alfresco.org/model/content/1.0\}name:myfile.txt
‍‍‍

I've not tried to use the short form myself.

Wildcarding is also working correctly. E.g. I generate this search when the user wants to find all files starting with the name "myfile":

 TYPE:"{http://www.alfresco.org/model/content/1.0}content" AND @\{http\://www.alfresco.org/model/content/1.0\}name:myfile*
‍‍‍

Hope this helps,

Kevin

[dbevacqua]

Hi Kevin

Thanks for your reply. The short form works when no wildcard is specified. I notice that the NamespaceResolver bein used only contains a mapping between one of the three prefixes in my content model and the namespace uri. There are a number of other null entries in the prefixes. DOn't know if this is of any help.

I should point out that examples on this wiki page:

http://wiki.alfresco.com/wiki/Search#Finding_nodes_by_text_property_values

use short names.

Anyway I shall use the full name from now on.

Thanks

Dominic

[kevinr]

Interesting that short names seem to be the problem - i've emailed our Lucene expert who wrote our integration and when he gets back from holiday hopefully he can shed some light on the issue.

Thanks,

Kevin

[jtorres]

Thanks a lot, It has been a great help. I was really mad trying to find what was wrong.

[andy]

Hi

The short names issue with wildcard queries (and fuzzy queries and prefix queries) is fixed in 2.0.

Regards

Andy