Looks really cool! I'll let Ariel look at this before merging.

Note that a bulk of the code was pulled from #63.

Open question: should run_func be made to look more like run_code, returning stdout and any created figures? Also, should the dynamic functions return just the response or the full dictionary? For example m.plot([1,2,3]) would make the plot but we'd never see it. If we got the path to the created figure, we could do something with it.

Here's what I propose:
Use pymat_eval for both, and either run evalin or a version of pymat_feval here.

Just to see that I understand: this would grab the figures and stdout in
the magic even if you call a function that creates the figure? I like that.

On Tue, Feb 17, 2015 at 7:17 AM, Steven Silvester notifications@github.com
wrote:

Here's what I propose:
Use pymat_eval for both, and either run evalin or a version of pymat_feval
here
https://github.com/arokem/python-matlab-bridge/blob/master/pymatbridge/matlab/util/pymat_eval.m#L38
.

—
Reply to this email directly or view it on GitHub
#145 (comment)
.

To be clear, the matlab magic already uses run_code exclusively, so this would affect run_func and the dynamically created functions.

I think this is good to go now. I will make the changes to the Matlab end in another PR.

Gotcha. Looks good!