Hi,

I am executing a simple query on two fields (title and description) and expect 24 results for the query.
While the resultSize that is fetched is correct, the list returned contains only two results.
Here is the piece of code that I use.

String[] q = new String[] {
advSearchCriteria.getSearchString(),
advSearchCriteria.getSearchString()};

Query luceneQuery = MultiFieldQueryParser.parse(q, new String[] {
"i_title", "i_description" }, new StandardAnalyzer());

org.hibernate.search.FullTextQuery fullTextQuery = fullTextSession
.createFullTextQuery(luceneQuery, FaculteClass.class);
fullTextQuery.setFirstResult(0);

// execute search
List list = fullTextQuery.list(); //list contains only two objects.
resultCount = fullTextQuery.getResultSize(); //this value is 24
// parse result
return parseClassResult(list);

Any help will be greatly appreciated
Thanks,
-rahul

You can try to use getResultList method of FullTextQuery class.

Are you setting somewhere setMaxResult()? This could influence the size of the returned list. getResultSize() in returns the total number of hits independent of 'pagination'.

--Hardy

Nope, aint using setMaxResults anywhere...actually, tried to explicitly set that to around 25 as well but still the same result.

you don't mess with setCriteriaQuery, do you? And you're using the latest version?
After those questions, try and create a minimal test case and post it to JIRA

nope, no messing with setCriteriaQuery and am using release 3.0.0 ga.

ah I know :)
You have emptied your database without emptying your index (for a test maybe).

Hi Emmanuel, Hardy,

I think the problem mentioned above is due to the following piece of code in org.hibernate.search.engine.QueryLoader.load(EntityInfo... entityInfos) :

criteria.add( disjunction );
criteria.list(); //load all objects

//mandatory to keep the same ordering
List result = new ArrayList(entityInfos.length);
for (EntityInfo entityInfo : entityInfos) {
Object element = session.load( entityInfo.clazz, entityInfo.id );
if ( Hibernate.isInitialized( element ) ) {
//all existing elements should have been loaded by the query,
//the other ones are missing ones
result.add( element );
}
}
return result;

The session.load does not load the object and if its not initialized then its not added to the final result list. I was able to find this out by debugging the Hibernate Search source code.

Also this behavior is not consistent. It varies with the no. of records being fetched. For example here are few use cases:

1. In my DB I have 5 records with "Class_<no>" in the 'title' field. The <no> is is an integer(Class_1, Class_2, etc.). It sometimes returns me 1, sometimes 2 records when the search criteria is "class*". The resultListCount returned by fullTextQuery.getResultSize() is 5 which is as expected.

2. I have 20 records with "Indoor Photography_<no>" in the title field. On firing a search with either "indoor*" or "photo*", the result list is empty although the resultListCount returned by fullTextQuery.getResultSize() is 20 which is as expected.

Awaiting to hear something from you guys.

--Regards,
Rakesh Shete

This also sounds to me as if your index is not in sync with the actual objects you have in the db. If you get a list of EntityInfo objects, but load.session() keeps returning null it somehow indicates that the object does not exist in the database.

How do you index your data? Do you rely on the automated indexing or do you trigger a manual indexing of the data. Try using a RAMDirectoryProvider. This would prevent having a stale index. That's just an idea to narrow down you problem.

The only case where I had a similar problem was when setting up JUnit tests for Hibernate Search within a Spring environment using Spring's AbstractJpaTests.

AbstractJpaTests creates a transaction for each test, but in each test I had to first index and commit the transaction before being able to test searches. Not sure if the following lines of codes are toom uch out of context, but the idea is:



Hope this helps.

Hi Hardy,

The indexing is done only once for my application through an load-on-startup servlet and only once. Here is the code for the same:

public void indexClasses() {
SessionFactory sessionFactory = getSessionFactory();
Session session = sessionFactory.openSession();
FullTextSession fullTextSession = Search.createFullTextSession(session);
fullTextSession.setFlushMode(FlushMode.MANUAL);
fullTextSession.setCacheMode(CacheMode.IGNORE);
Transaction transaction = fullTextSession.beginTransaction();
// Scrollable results will avoid loading too many objects in memory

ScrollableResults results = fullTextSession.createCriteria(
FaculteClass.class).scroll(ScrollMode.FORWARD_ONLY);
int index = 0;
while (results.next()) {
index++;
fullTextSession.index(results.get(0)); // index each element
}
transaction.commit();
}

Also I have again debugged through the QueryLoader.load. I do not get null objects for session.load() but proxies which are uninitialized and thats why they are not added to the final result list. I also checked the 'id' value that is passed to the session.load() method. A record for every 'id' exists.

So I do not believe that it is problem of the indexes being out of sync.

Do we have any problem using session.get()?

I understand that session.load() must be used to improve performance and allow lazy initialization.

I think if this is a concern then we could provide an API where we provide the option lazy loading the objects that would be part of the result. Thoughts?

--Regards,
Rakesh Shete

Can anyone help me here?

I am stuck at the moment

The code in HSearch is I believe correct.
I load the object list using the criteria API,
Then get the object using load(). The objects should by then be loaded unless they are not present in the DB. This behavior is what is expected in most cases. In other words, I do not use load to optimize things, but just to have the correct behavior (this is what is explained in the comments BTW).

Try and check to see the reason wht the criteria.list() does not load all the objects, you should find the problem.

I am facing similar problem, the difference is that fullTextQuery.list() returns the correct number of rows, but fullTextQuery.getResultSize() does not.

I am using criteria with the search query (I have to narrow down my result set). The code follows:



The list() method works perfectly (as before mentioned), it seems as if the getResultSize() does not take the criteria into account when returning row count.

Am I doing anything wrong here? The Hibernate Search manual says

so in my opinion it should work. Should I post a bug in JIRA?

Criteria used in fullTextQuery.setCriteriaQuery(criteria);
must not be constrained. it should only be used to change the fetching strategy.

I believe it's written both in the Javadoc and the reference documentation

Constrained Criteria work when I fetch the results with list(), it just does not work with getResultSize().

I understand it can be hard to change this behaviour in Hibernate Search right now but is there any reasonable workaround for this problem? I am searching through a really large number of entities and using pagination. I need to display search result count (the constrained one) on each result page.

Fetching ALL search results by list() and calling size() on the big List seems to be a performance bottleneck.

What would you choose if you were facing such problem?