Left to decide: how and where to set fmt and res. We could use pymat_set_variable to set variables like _plot_fmt and _plot_res, and use those in make_figs, or we could specify the arguments as part of the calls to pymat_eval and pymat_feval. What do you think @arokem?

Thanks for looking into this.

Just to see that I understand the different proposals: do you propose to set global variables on the matlab side and use these to determine the format and resolution in make_figs? Should we consider having a _pymat_config struct array (or other object) to address this kind of issue more generally?

I like the idea of a global config struct, shall I give it a shot?

Yes - please do. The main thing that worries me (for now) is name-space
collisions with this variable. I might come up with other things later on.

On Sun, Feb 8, 2015 at 2:53 PM, Steven Silvester notifications@github.com
wrote:

I like the idea of a global config struct, shall I give it a shot?

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

We could put it in its own namespace. 

—
Sent from Mailbox

On Mon, Feb 9, 2015 at 3:53 PM, Ariel Rokem notifications@github.com
wrote:

Yes - please do. The main thing that worries me (for now) is name-space
collisions with this variable. I might come up with other things later on.
On Sun, Feb 8, 2015 at 2:53 PM, Steven Silvester notifications@github.com
wrote:

I like the idea of a global config struct, shall I give it a shot?

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

---

Reply to this email directly or view it on GitHub:
#128 (comment)

If you mean package folders, I'm pretty sure Octave doesn't support those, unfortunately.

Another alternative: we could leave the current make_figs alone and use defaultfigurepapersize, defaultfigureunits, and defaultfigurepaperunits to define a height and width (the default resolution is 150dpi).

Heck I can do that right now with the current matlab_kernel.

matlab_kernel 0.3 demonstrates this behaviour.

Do you have an example notebook you could put up on nbviewer?

On Tue, Feb 10, 2015 at 6:54 PM, Steven Silvester notifications@github.com
wrote:

matlab_kernel 0.3 demonstrates this behaviour.

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

In action: http://nbviewer.ipython.org/github/Calysto/matlab_kernel/blob/master/matlab_kernel.ipynb

That's fantastic! I particularly enjoy the %python magic :-)

Does this mean that we should close this PR, or is it still needed?

Good call, I added a %%python magic example. I'd say close this one in favour of #134.