[IPython-user] [IPython-dev] Development plans update

Brian Granger ellisonbg.net@gmail....
Tue Feb 5 09:55:47 CST 2008


> %bg is a different thing, async jobs would still happen in order (they
> are put on queue).
>
> I was thinking of something like this:
>
> - a = Foo()
>
> This is synchronous
>
> - &a= Foo()
>
> This returns immediately and is executed asynchronously
>
> - Automatic determination:
>
> %autoasync
>
> a = Foo()

I think we definitely want to explore something like this.  But, as I
mentioned above, magic commands will be parsed and handled by the
backend.  But, the frontend is where all the decisions about
synchronous/asynch behavior must be made.  I guess  I see two options:

1) Create a new syntax for frontend parsed/handled magic commands.

2) Just handle these things though the actual gui.  For instance there
could be a toggle button for blocking/non-blocking modes.

Fun to think about though.

> This blocks for a few seconds, then sees that it's a "long-lasting
> command" and gives the prompt again (entering asynchronous mode until
> the queue is empty again)
>
> All of this is easy to implement using a single Queue structure,
> though, so we don't need to design it up-front.
>
> > in deciding how a double TAB should complete (i.e. variables depending
> > on jobs in 'background' should never be tried to be completed).
>
> Yeah, it's easy to scan for assignment (every variable on left side of
> single = is considered "incomplete" until executed, and we can
> maintain a dict of incomplete variable).

True.

> --
> Ville M. Vainio - vivainio.googlepages.com
> blog=360.yahoo.com/villevainio - g[mail | talk]='vivainio'
>
> _______________________________________________
> IPython-user mailing list
> IPython-user@scipy.org
> http://lists.ipython.scipy.org/mailman/listinfo/ipython-user
>


More information about the IPython-user mailing list