1089301 - Investigating cheating on matching address headers to improve search/filter speed

Status: VERIFIED FIXED

(MailNews Core :: Search, defect)

Description

[Joshua Cranmer [:jcranmer]]

Parsing addressing headers is relatively expensive. The logic we use in the search code is basically "matches any name or any email"--if we're using the Contains/DoesntContain match properties, we don't even need to parse the address headers and we could use a mere decoded header instead (potentially faster if bug 1089229 is fixed and extended to addressing headers). Even the other matches could be filtered with a prepass if we expect them to fire relatively infrequently.

Comment 1

[Joshua Cranmer [:jcranmer]]

Attachment: Oh hey, this is fast

This requires the patches over in bug 1089298 to apply. Since quick search only does a Contains match, this makes it REALLY fast.

Comment 2

[Kent James (:rkent)]

Comment on attachment 8559279 [details] [diff] [review]
Oh hey, this is fast

Review of attachment 8559279 [details] [diff] [review]:
-----------------------------------------------------------------

This need for patch is a good example of the types of issues we are going to continue to face if we try to convert most things to javascript. You have this beautiful javascript code that does precise mime processing - but then you have to hack around it because it is slower than we want, and we end up introducing approximations like this to get the speed up.

Still, I think that we need to do this. But we will need to be on the lookout for unforeseen issues with this search approximation.

Comment 3

[Joshua Cranmer [:jcranmer]]

https://hg.mozilla.org/comm-central/rev/f8b8e522d2dc

Comment 4

[Wayne Mery (:wsmwk)]

thanks. speed is much better