CMIS CONTAINS() not finding documents as expected

[rmcveigh]

I'm using 3.4 Enterprise.  Using the Share UI, I can search for the string "test" and find content nodes as expected.  I can use the advanced search scoped to a custom content type and still find instances of that type.  However, executing a CMIS search isn't finding the documents.  I've tried

select * from cmis:document where CONTAINS('test')

and

select * from my:type where CONTAINS('test')

and a bunch of other permutations and I can't figure out yet what the UI is doing differently to return the documents I expect where I haven't been able to mimic that behavior using a CMIS call.  What is different between the UI? 

I've also tried CONTAINS('TEXT:test') (not sure if that syntax is right) and no luck there either.  I get this exception when I try:

Caused by: org.alfresco.cmis.CMISQueryException: 02140674 Unknown column/property TEXT

I probably misread, but the docs for CMIS query lead me to believe this was possible.  I was reading here: http://wiki.alfresco.com/wiki/CMIS_Query_Language#Differences_between_cmis-strict_and_cmis-alfresco and following the link to the Alfresco FTS docs.

-Ryan

[rmcveigh]

I should clarify my goal:  I wish to search for a given string across all indexed properties of a custom content type.  So while this works:  select * from my:type where contains('cmis:name:test')

This isn't realistic unless I want to OR all my custom properties in - which I assumed wasn't necessary.

-Ryan

[andy]

Hi

CONTAINS() uses TEXT as the default field (it may well not be exposed to use direct).
TEXT is any property of type d:content.

The share UI expands the query using something like the macro below:

"%(cm:name cm:title cm:description ia:whatEvent ia:descriptionEvent lnk:title lnk:description TEXT)


ALL can be used from share to search over all properties.
I would have to check if this is exposed in CMIS.


CONTAINS('ALL:test')

It is possible that ALL like TEXT is not exposed in your Alfrecso version…..
so what version are you using and can you post the rest of the stack trace.

Andy

[rmcveigh]

Hi Andy, thanks for the reply.  I'm running 3.4 Enterprise.  I tried this query: 

SELECT * FROM cmis:document WHERE CONTAINS('ALL:test')‍

When I do, I get a failure and this is in the log:

Caused by: org.alfresco.cmis.CMISQueryException: 02160067 Unknown column/property ALL
        at org.alfresco.cmis.search.CmisFunctionEvaluationContext.checkFieldApplies(CmisFunctionEvaluationContext.java:366)
        at org.alfresco.repo.search.impl.parsers.FTSQueryParser.buildFieldReference(FTSQueryParser.java:981)
        at org.alfresco.repo.search.impl.parsers.FTSQueryParser.buildFTSTest(FTSQueryParser.java:248)
        at org.alfresco.repo.search.impl.parsers.FTSQueryParser.buildFTSConnective(FTSQueryParser.java:169)
        at org.alfresco.repo.search.impl.parsers.FTSQueryParser.buildFTSConnective(FTSQueryParser.java:156)
        at org.alfresco.repo.search.impl.parsers.FTSQueryParser.buildFTS(FTSQueryParser.java:110)
        at org.alfresco.cmis.search.CMISQueryParser.buildPredicate(CMISQueryParser.java:457)
        at org.alfresco.cmis.search.CMISQueryParser.buildTest(CMISQueryParser.java:283)
        at org.alfresco.cmis.search.CMISQueryParser.buildNegation(CMISQueryParser.java:263)
        at org.alfresco.cmis.search.CMISQueryParser.buildConjunction(CMISQueryParser.java:232)
        at org.alfresco.cmis.search.CMISQueryParser.buildDisjunction(CMISQueryParser.java:205)
        at org.alfresco.cmis.search.CMISQueryParser.parse(CMISQueryParser.java:160)
        at org.alfresco.cmis.search.CMISQueryServiceImpl.query(CMISQueryServiceImpl.java:126)
        at org.alfresco.repo.cmis.rest.CMISScript.query(CMISScript.java:652)
‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍

[rmcveigh]

Andy,

I've experimented some more and I'm able to do some interesting things to perhaps meet my requirement, but I'd like to believe there's a better way.  My requirement is to execute a "full text" search across all indexed fields for a given content type (e.g. cmis:name, myType:field1, myType:field2, cm:content, cm:title, etc).

I can execute a query like: 

select d.*, o.* from cmis:document as d join cm:titled as o on d.cmis:objectid = o.cmis:objectid where CONTAINS(d,'cmis:name:test OR cmis:createdBy:admin') OR CONTAINS(o,'cm:title:test')‍

This implies I can query on all indexed fields of my type by explicitly OR'ing them in the various scoped CONTAINS predicates.  This seems heavy handed.  Is there a better way?

Another approach which also works is:

select d.*, o.* from cmis:document as d join cm:titled as o on d.cmis:objectid = o.cmis:objectid where o.cm:title like '%25test%25' OR CONTAINS(d,'cmis:name:test')
‍‍

When would I prefer to use the LIKE predicate over CONTAINS or vice versa?

-Ryan

[andy]

Hi

Sticking to the spec you are allowed one CONTAINS().

For now I would drop the alias in the CONTAINS and "OR" all the terms together in one single contains statement (with no alias)

It I remember correctly, the extended Alfrecso FTS embedded in CMIS should understand embedded "d.cm:content:woof" style. However, as we do not do any actual join and the CMIS contains does not actually check the property binding tot he type (as the default for CONTAINS() is non strict mode).

I will fix up the special pseudo fields like ALL, d:content to work in CMIS contains.

Andy

[rmcveigh]

Thanks Andy,

I had been unable to use a single CONTAINS statement with both cmis properties and cm:title etc, I had to OR as I described above.  Is there a JIRA issue I can track for the updates you've described?

-Ryan

[andy]

Hi

The issue is here https://issues.alfresco.com/jira/browse/ALF-7827

Andy

[thestorm]

whats the variable for fulltext content then? (I wanna query for fulltext index)

[andy]

Hi

The additional fields are exposed in 3.4.3 and the related bug has been resolved.

Until you have this fix: CMIS contains searches all properties of type d:content (if you specify no field) or the specified properties you give.

CONTAINS('woof OR cmis:name:woof')

should search all content and the name property. TEXT:woof will be the same as woof on its own when the additional fields are available.

Andy