From: loial on
Ok, thats great. Thanks for the very elegant solution(s)

On 6 Aug, 13:44, Nobody <nob...(a)> wrote:
> On Fri, 06 Aug 2010 02:06:29 -0700, loial wrote:
> > I need to read a large amount of data that is being returned in
> > standard output by a shell script I am calling.
> > (I think the script should really be writing to a file but I have no
> > control over that)
> If the script is writing to stdout, you get to decide whether its stdout
> is a pipe, file, tty, etc.
> > Currently I have the following code. It seeems to work, however I
> > suspect this may not work with large amounts of standard output.
> > process=subprocess.Popen(['myscript', 'param1'],
> > shell=False,stdout=subprocess.PIPE,stderr=subprocess.PIPE)
> > cmdoutput=process.communicate()
> It's certainly not the best way to read large amounts of output.
> Unfortunately, better solutions get complicated when you need to read more
> than one of stdout and stderr, or if you also need to write to stdin.
> If you only need stdout, you can just read from process.stdout in a loop.
> You can leave stderr going to wherever the script's stderr goes (e.g. the
> terminal), or redirect it to a file.
> If you really do need both stdout and stderr, then you either need to
> enable non-blocking I/O, or use a separate thread for each stream, or
> redirect at least one of them to a file.
> FWIW, Popen.communicate() uses non-blocking I/O on Unix and separate
> threads on Windows (the standard library doesn't include a mechanism to
> enable non-blocking I/O on Windows).
> > What is the best way to read a large amount of data from standard
> > output and write to a file?
> For this case, the best way is to just redirect stdout to a file, rather
> than passing it through the script, i.e.:
>         outfile = open('outputfile', 'w')
>         process =, stdout = outfile)
>         outfile.close()