@@ -150,17 +153,19 @@
/* * Generates the DBM file.
*/
-apr_status_t generate_index(request_rec *r, apr_file_t * f);
+apr_status_t mbox_generate_index(request_rec *r, apr_file_t * f,
+ mbox_indexer_t *indexer, const char* list,
+ const char* domain);
Is the include from above present because of this prototype change? What does the Lucene index have to do with the generation of the mod_mbox index? Why can't the indexing for the search be done separately? Why tie it into the generation of master index? At a quick glance, Lucene should be able to separately generate its own indexes after mod_mbox has created its index.
(I'd think it should be kept separate: especially considering how specific it is to Lucene.)
I don't think the simple API I made is specific to Lucene or Lucene4c -- in fact the original code behind the indexer/search API was written with CLucene, the C++ implementation of Lucene.
I viewed doing the full text indexing as part of the index generation to be a good thing, as all of the information that is required for a full text search is available and in memory at this point.
The other option is yes, to make them completely separate, which does have some advantages, but it means without more significant refactoring, you will need to read every mbox file twice.
-Paul
