[IPython-dev] Environment in load balanced map_async?

[MinRK]

This is the single most common confusion users face when getting started
with IPython.parallel: namespace management, which I have discussed on
Stack Overflow a few
<http://stackoverflow.com/questions/10857250/python-name-space-issues-with-ipython-parallel/10859394#10859394>
 times <http://stackoverflow.com/a/12307741/938949>, and in the
IPython.parallel
tutorial
<http://nbviewer.ipython.org/github/minrk/IPython-parallel-tutorial/blob/master/tutorial/Remote%20Execution.ipynb#working-with-the-engine-namespace>
.

The gist is that in your code, kernels is just a reference to
__main__.kernels on the engine, which hasn’t been defined. You would solve
the same problem with view['kernels'] = kernels, or (depending on what
`kernels` is, this doesn’t always work for serialization reasons)
view.map_async(kernels.do_function_direct,
[1,2,3]).

-MinRK
​

On Fri, Sep 12, 2014 at 2:17 PM, Doug Blank <doug.blank at gmail.com> wrote:

> Can you imagine why this would fail:
>
> view_load_balanced.map_async(
>             lambda arg: kernels.do_function_direct(arg),
>             [1, 2, 3])
>
> but that this ugly, terrible hack would work:
>
> view.execute("""
> import os
> os.kernels = kernels
> """)
> view_load_balanced.map_async(
>             lambda arg: os.kernels.do_function_direct(arg),
>             [1, 2, 3])
>
> Why would I not be able to find a variable in the environment, but I can
> put it in a module, and find it there?
>
> -Doug
>
> _______________________________________________
> IPython-dev mailing list
> IPython-dev at scipy.org
> http://mail.scipy.org/mailman/listinfo/ipython-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/ipython-dev/attachments/20140912/c5428b37/attachment.html>