Hi,

We are using Hibernate Search 3.0.2 with Hibernate 3.2.6.

We have a class bridge on a class that is meant to gather 3 fields into a single lucene field.

We also have a specific workflow steps that basically crawls a lot of data and create/update existing instance of that class. During that process, none of those 3 fields are updated. For some reasons, a transaction is created to update every object during crawling

What we noticed is that a lot of time is spent in the lucene reindex process.

I have a couple of questions:

* Is there a way to control the document update using a class brige? (i.e. realize it is an update and none of the three fields have been updated so nothing should be done)
* Is there a way to disable automatic indexing per session so that the indexing can be done at the end of the crawling process?

Any other insight?

Thanks,
Stéphane

I am not sure I understand completely what you are saying. Your crawler creates a transaction for each object? That does not sound really efficient. For batch indexing I recommend the FullTextSession.index() operation, but of course don't create a single transaction for each update.


No, once the class bridge is called, Hibernate Search has already determined that there was an insert/update even for the entity. It cannot know that the properties relevant for the class bridge have not changed.


No. One option might be to have the crawler as separate application in which you would turn off autoindexing. But maybe you could explain first a little bit more about how this crawler works. Some code examples might be handy as well ;-)

--Hardy