Re: [Matplotlib-users] How to provide colorbars for scatterplots created this way?

[Paul H. <pmh...@gm...>]

On Thu, Dec 20, 2012 at 1:05 PM, Kynn Jones <ky...@gm...> wrote:

> I create PNG files of scatterplots with code that, in essence, goes as in
> the sketch below:
>
>
> cmap = (matplotlib.color.LinearSegmentedColormap.
>         from_list('blueWhiteRed', ['blue', 'white', 'red']))
>
> fig = matplotlib.figure.Figure(figsize=(4, 4), dpi=72)
> ax = fig.gca()
>
> for marker in 'o s ^ *'.split():
>
>     X, Y, COLOR = zip(*((record.x, record.y, record.level)
>                         for record in data if record.marker == marker))
>
>     ax.scatter(X, Y, marker=marker,
>                c=COLOR, vmin=0, vmax=1, cmap=cmap,
>                **otherkwargs)
>
> # various settings of ticks, labels, etc. omitted
>
> canvas = matplotlib.backends.backend_agg.FigureCanvasAgg(fig)
> fig.set_canvas(canvas)
>
> # IMPORTANT: the generated figure is *not* displayed on the screen, but
> # rather it is output to disk as a PNG file:
> canvas.print_png('/path/to/output/fig.png')
>
>
>
> My question is this:
>
> What do I need to add to the code above to get a vertical colorbar
> (representing the colormap incmap) along the plot's right edge?
>
> I word the question in this way because I am not sufficiently facile with
> Matplotlib to deviate too far from the working code above.
>
> In particular, my code *has* to be able to produce PNG files
> *non-interactively*, so the last line in the code sketch above is really
> essential.
>
> Thanks in advance!
>
> kj
>
>
Can you provide some more information and a self-contained example? What is
your record object? Is it a pandas dataframe? Are the limits of
record.level consistent with vmax and vmin kwargs fed in the call to
ax.scatter?

Typically you can just do:

import matplotlib.pyplot as plt
# blah blah
fig, ax = plt.subplots()
s = ax.scatter()...
cb = plt.colorbar(s)
cb.set_label('Cbar Label Here')

Also, I don't think you need to mess with the backend stuff. Just do
fig.savefig('figname.png'). If you need separate markers for each set, make
a single call to scatter for each data group, and use numpy to figure out
what the appropriate vmax a vmin limits are for the colorbar.

-paul