[SciPy-Dev] signal.freqs/freqz plot keyword

[Ralf Gommers]

On Sat, Jul 17, 2010 at 10:09 PM, Ryan May <rmay31 at gmail.com> wrote:

> On Jul 17, 2010, at 4:38, Ralf Gommers <ralf.gommers at googlemail.com>
> wrote:
>
> > freqs/freqz have a keyword plot=None, which if True does this:
> > if not plot is None:
> >     plot(w, h)
> > while plot is not even defined.
>
> If it's not none, it should be a callable passed in, which let's you seine
> your own plotting function. Actually using matplotlib's plot produces
> unexpected results,  you end up plotting the real part of the complex
> transfer function, not the magnitude. This actually bit me recently and
> should probably be documented.
>

Ah, that makes sense. Would indeed have been clearer with documentation.
I'll add it.

>
> > keyword should be removed or the implementation should be something like:
> >     if plot:
> >         try:
> >             import matplotlib.pyplot as plt
> >             plt.plot(w, h)
> >             plt.show()
> >         except ImportError:
> >             warnings.warn("`plot` is True, but can't import
> matplotlib.pyplot.")
>
> No, you can use any plotting library right now.
>
>
> > Removal makes more sense I think, since scipy does not depend on
> matplotlib. And why provide a plot keyword in these functions and not in
> many others where it would also make sense?
>
> It's mimmicking MATLAB functionality, which will plot the transfer function
> if you call freqz and don't save the results.
>
> Looks like it's been part of the signature since the functions were added,
so never mind the removing. But I still find it very inconsistent and
odd-looking.

Cheers,
Ralf
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/scipy-dev/attachments/20100718/8e7ab1cc/attachment.html>