tag 569900 + fixed-upstream pending thanks On Sun, Feb 14, 2010 at 01:29:25PM -0800, Ian Zimmerman wrote: > It seems there's no way to skip files that have not been modified > since omindex last indexed them. This looks like a huge missed > opportunity to speed things up.
This has been implemented upstream: http://trac.xapian.org/ticket/342 The fix makes use of a new backend feature (tracking upper and lower bounds on each value slot). On trunk, we have an upper bound on the lastmod value for any document in the database, so we can just compare with that single threshold to spot most modified documents. A backported patch would have to check the timestamp of every file against its current lastmod value in the database, which requires fetching that data. This would make the "most documents have changed" case slower. Since 1.2.0 is due out in the next few weeks, I don't intend to try to backport this patch, but instead I'm going to focus on helping get 1.2.0 released and packaged for Debian. > In general, it is disconcerting that > omindex can only be given a whole directory to process, and only one > at a time. The Unix Way (TM), and some existing utilities like cpio > and pax, suggest another interface: accept the list of files to be > added or updated on standard input, either newline or NUL separated. Please don't raise unrelated issues in a single ticket! This is: http://trac.xapian.org/ticket/376 Cheers, Olly -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20100215003714.gp8...@survex.com
