Indexing is slow

Are these 5,000 nodes generated by the devel module?

I'm testing this module on my development server, populating nodes using the devel module. I've not tried the perl indexer, but Xapian took nearly 3 minutes to index only 500 nodes, I think that qualifies as slow.

Reading through the code, I see that you open the database, write to it, and close/flush it for each and every node that is indexed -- wouldn't it make more sense to open it once, write a whole bunch of nodes, then close/flush it?

Ah, I see now that this is what you're talking about re the static in your comment, and I see the comment in your code now. Okay, I'll help track down why the index is unusable when you use a static.

Passing the whole db_query result into the reindex function, and then looping there through all the nodes is extremely fast and generates a valid index. I'm still trying to understand why using a static does not work.

The issue had nothing to do with using a static. Instead, it's the way the PHP bindings work. After you are done making changes to a writable Xapian database instance, you have to set it to NULL otherwise it won't be flushed to disk.

The attached patch adds a new "flush" parameter to the _xapian_init_database() function, and updates all instances of opening a writable database to call the init function when finished, causing all changes to be flushed to disk.

It now takes ~35 seconds to index 5,000 devel generated nodes on my test server.