Re: [RFC \ WISH] Add -o option to git-rev-list

[Andreas Ericsson <ae@op5.se>]

Marco Costalba wrote:
> On 12/11/06, Andreas Ericsson <ae@op5.se> wrote:
>> Marco Costalba wrote:
>> > On 12/10/06, Linus Torvalds <torvalds@osdl.org> wrote:
>> >>
>> >> Why don't you use the pipe and standard read()?
>> >>
>> >> Even if you use "popen()" and get a "FILE *" back, you can still do
>> >>
>> >>         int fd = fileno(file);
>> >>
>> >> and use the raw IO capabilities.
>> >>
>> >> The thing is, temporary files can actually be faster under Linux just
>> >> because the Linux page-cache simply kicks ass. But it's not going 
>> to be
>> >> _that_ big of a difference, and you need all that crazy "wait for
>> >> rev-list
>> >> to finish" and the "clean up temp-file on errors" etc crap, so 
>> there's no
>> >> way it's a better solution.
>> >>
>> >
>> > Two things.
>> >
>> > - memory use: the next natural step with files is, instead of loading
>> > the file content in memory and *keep it there*, we could load one
>> > chunk at a time, index the chunk and discard. At the end we keep in
>> > memory only indexing info to quickly get to the data when needed, but
>> > the big part of data stay on the file.
>> >
>>
>> memory usage vs speed tradeoff. Since qgit is a pure user-app, I think
>> it's safe to opt for the memory hungry option. If people run it on too
>> lowbie hardware they'll just have to make do with other ways of viewing
>> the DAG or shutting down some other programs.
>>
>> > - This is probably my ignorance, but experimenting with popen() I
>> > found I could not know *when* git-rev-list ends because both feof()
>> > and ferror() give 0 after a fread() with git-rev-list already defunct.
>> > Not having a reference to the process (it is hidden behind popen() ),
>> > I had to check for 0 bytes read after a successful read (to avoid
>> > racing in case I ask the pipe before the first data it's ready) to
>> > know that job is finished and call pclose().
>> >
>>
>> (coding in MUA, so highly untested)
>>
> 
> Thanks Andreas, I will do some tests with your code. But at first
> sight I fail to see (I'm not an expert on this tough ;-)  ) where is
> the difference from using popen() and fileno() to get the file
> descriptors.
> 

read() vs fread(), so no libc buffers. When I did comparisons with this 
(a long time ago, I don't have the test-program around) in style of

	read(out[0], buf, sizeof(buf));
	write(fileno(stdout), buf, sizeof(buf));

with a command line like this;

	cat any-file | test-program > /dev/null

I saw a static ~10ms increase in execution time compared to

	cat any-file > /dev/null

regardless of the size of "any-file", so I assume this overhead comes 
from the extra fork(), which you'll never get rid of unless you use 
libgit.a.

-- 
Andreas Ericsson                   andreas.ericsson@op5.se
OP5 AB                             www.op5.se