[darcs-users] hashed-storage index

[Petr Rockai]

Eric Kow <kowey at darcs.net> writes:
> Will loading/reading a 7.2M file for darcs operations create any
> noticeable overhead?

Well, as I said, parsing the file is an actual CPU bottleneck in darcs
whatsnew. That's why it's not plain text, for one. Nevertheless, it's much
faster than what we had before (if I recall correctly, darcs 2.2 will run out
of memory after a few minutes trying to do darcs whatsnew on the 80k-file
repository, unless mtimes happen to match, in which case it'll merely take a
few seconds -- whereas the new code will take less than one).

Well, what do you mean with "noticeable overhead" anyway? Of course, it would
be better if darcs wouldn't need to do anything, but it would be sort of
pointless to use it, then. </irony> ... Ah, and "reading" the file is done with
mmap, so that's a zero-copy operation.

Yours,
   Petr.

PS: Git index for the same repository is 5.5M, which I expect to be due to git
using shorter (sha1 as opposed to sha256) hashes.

-- 
Peter Rockai | me()mornfall!net | prockai()redhat!com
 http://blog.mornfall.net | http://web.mornfall.net

"In My Egotistical Opinion, most people's C programs should be
 indented six feet downward and covered with dirt."
     -- Blair P. Houghton on the subject of C program indentation