On Mon, 15 Jun 2009 15:43:13 -0400 Matt <hellzf...@gmail.com> wrote: > I'm going to use the multipocessing library from here forward so I can > take advantage of multiple cores and clusters. Either one should work > for my use, since in my non-demonstration code each thread spends most > of it's time waiting for a separate non-Python subprocess (created > with subprocess.Popen) to finish anyway. (I guess Python would see > this as IO-blocking) Therefore, if you can fix my toy example with > threading, that's fine. > > DB.py, followed by a KeyboardInterrupt yields the output in a.out. I > want roughly the output in desired.out. > > What do I need to do to modify this code to get my desired output and > corresponding functionality? It would be a shame if this wasn't > possible in any pure-Python way.
I don't know how complex task you have, but solving trivial IO blocks with threads or subprocesses look either like ugly hack or an overkill to me. Why not just use I/O without blocking? It's not 80s or 90s anymore, where you had to create subprocess to handle every non-synchronous task, and since the main burden will be pushed into non-py subprocesses already, why not implement controller as a nice, clean and simple single-threaded event loop? Consider this recipe: http://code.activestate.com/recipes/576759/ And if the task before you is complex indeed, involving more than just two to five child processes with a simple "while True: ..." loop, consider using twisted framework - it'll allow you to do incredible stuff with any number of sockets with just few lines of code in a clean, abstracted way. Latter would also mean that you can always replace os pipes with network sockets just by changing transport name, distributing your app to any number of machines. -- Mike Kazantsev // fraggod.net
signature.asc
Description: PGP signature
-- http://mail.python.org/mailman/listinfo/python-list
