[yt-dev] Small adjustment for the PhasePlot API

[Nathan Goldbaum]

Hi all,

There's a minor issue with the PhasePlot class that I'd like to fix -
unfortunately fixing it implies an API change.  Before I issue a pull
request I'd like to ask the group what they think.

Right now a PhasePlot can be created in two ways.  Both are via the
PhasePlot constructor.  The first uses what we have in the docs:

ph = PhasePlot(dd, 'density', 'temperature', ['cell_mass', 'cell_volume'])

The second is via the profile keyword argument:

profile = create_profile(dd, ['density', 'temperature'], ['cell_mass',
'cell_volume'])
ph = PhasePlot(source, None, None, None, profile=profile)

As you can see, the second approach is not so nice right now due to the
positional arguments in the PhasePlot initializer (I'm passing None where I
can pass dummy arguments -- I could pass anything in these slots, they are
ignored by the PhasePlot initializer).

This would make a lot more sense if PhasePlot didn't have positional
arguments and you could just do PhasePlot(profile=profile).  Then again,
I'd prefer not to change the API so drastically and add possible headaches
for dealing with the two ways the PhasePlot initializer can be used.

For ProfilePlot, we don't have a profile keyword argument.  Instead there
is a classmethod named from_profiles that creates a ProfilePlot without
explicitly invoking the class initializer.

I'd like to add a from_profiles method to PhasePlot to help make this
nicer.  I'd also like to remove the profile keyword argument, but wanted to
check with the list first before opening a PR that breaks the PhasePlot
API.  I could also leave the profile keyword argument, but for the sake of
simplicity I'd prefer if from_profiles were the only way to create a
PhasePlot from a profile object.

Thanks for your advice!

-Nathan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.spacepope.org/pipermail/yt-dev-spacepope.org/attachments/20140701/a113d70a/attachment.htm>