Ken Hornstein writes:
> >What I use now is "mark -list -sequence unseen", which returns compressed
> >lists of messages (i.e. "1-5,7-8" instead of "1 2 3 4 5 7 8"). Parsing
> >this to intersect it with my pick output is relatively fast, though it
> >is of course ineligant compared to getting *just* the list you want. It
> >is also surprisingly slow (like 1/7th of a second to get this vs. other
> >MH programs which run 10x or more faster). I don't really understand why
> >it is so slow, since it is a near character-for-character copy of the
> >one line that the .mh_sequences file has in it. If it wasn't so very
> >slow compared to the other MH programs, I probably would not have even
> >brought it up for now.
>
> You know, I just took a look at it; it should not be slow, actually,
> unless you're running into things like lock contention for the sequence
> file. It does the same things every other MH program does (it calls
> folder_read()) but then it does very little after that. Could you
> do a system call trace and try to figure out what's taking so long?
OK, I hadn't really tried to debug it before, but after this comment
I ran some tests. What I found is that it is not just mark, but
it seems related to the large folder I was testing it on, and some
other MH programs get slow when run under exactly the same circumstances.
Specifically, I was testing it on a very large folder of approx 100K
messages. Both the "mark ..." and a "show N" invokation take about
the more than 1/10th of a second on average, even for extremely short
outputs.
So:
time show 99948 > /dev/null
real 0m0.120s
user 0m0.040s
sys 0m0.068s
time mark -list -sequence unseen > /dev/null
real 0m0.123s
user 0m0.060s
sys 0m0.060s
The actual output of mark is (just so you know it is very tiny)
unseen: 99898-99907 99912-99947 99949-99959
The show command is for a minimally small email.
Using "cat" on the same file that takes show more than 1/10th of a
second is comparatively absurdly fast:
time cat Mail/inbox-old/99948 > /dev/null
real 0m0.002s
user 0m0.000s
sys 0m0.000s
"mark" run on a small folder is quite fast.
But it does seem Weird that it would take so long for these small
operations on a large folder. Is the "folder_read()" troublesome here?
--
Erich Stefan Boleyn <[email protected]> http://www.uruk.org/
"Reality is truly stranger than fiction; Probably why fiction is so popular"
_______________________________________________
Nmh-workers mailing list
[email protected]
https://lists.nongnu.org/mailman/listinfo/nmh-workers
