<<< Date Index >>>     <<< Thread Index >>>

Re: Add search tuning section to manual



* Kyle Wheeler <kyle-mutt-dev@xxxxxxxxxxxxxx>:
On Monday, November 20 at 10:12 AM, quoth Brendan Cully:
Yes, ~b will use the local cache, and be reasonably fast if you've got your mailbox precached (it's not a bad way to force mutt to cache your entire mailbox, actually).

As noted/wished on #mutt, I'll also add a few lines that since patterns apply to scoring too, the same performance improvements count for scoring...

Also, since servers expect to do searches, they often maintain databases to speed them up, which makes them often much _faster_ than local searches. I've definitely found that to be the case doing =b on large Cyrus-served folders.

Cyrus does this... not all do. Cyrus is actually rather advanced among IMAP servers. As a counter-example, BincIMAP does the equivalent of running 'grep' every time. I *believe* that UW-IMAPd and Courier don't use databases, but I don't know about Dovecot. Do you know differently?

IMAP searches (server side) are exact string matches only, right? Then I'll think about how to include this. A database won't help much with true regex search.

Also, this makes the pattern usage appear in a totally different light since a regex search will always be against the local cache when using regex (if the cache is present, if not messages are always fetched)...

On a related note: I wondered whether it's a bug that mutt uses case-sensitive search when using = patterns (since it automatically selects case-insensitive for regex). Bug or feature?

  bye, Rocco
--
:wq!