Replot

This repo is an attempt for a better API to plot graphs with Matplotlib in Python.

Matplotlib is a wonderful Python modules to plot data series, functions and so on. However, I think the API is quite verbose. This is an attempt at providing a better frontend API on top of matplotlib for easy and fast plotting, especially at prototyping time.

Features

These are the current features. I will extend the module whenever I feel the need to introduce new functions and methods. Please let me know about any bad design in the API, or required feature!

Saner default plots

Matplotlib plots are quite ugly by default, colors are not really suited for optimal black and white print, or ease reading for colorblind people. This module defines a clean default colorscheme to solve it (based on Colorbrewer Q10 palette). It also provides direct access to the Tableau10 and Colorbrewer Q9 palettes.

<dt>Support <code>with</code> statement</dt>
<dd>Ever got tired of having to start any figure with a call to
<code>matplotlib.pyplot.subplots()</code>? This module abstracts it using
<code>with</code> statement. New figures are defined by a
<code>with</code> statement, and are <code>show</code>n automatically (or
<code>save</code>d) upon leaving the <code>with</code> context.

<dt>Plot functions</dt>
<dd>Ever got annoyed by the fact that <code>matplotlib</code> can only
plot point series and not evaluate a function <em>à la</em> Mathematica?
This module let you do things like <code>plot(sin, (-10, 10))</code> to
plot a sine function between -10 and 10, using adaptive sampling.

<dt>Order of call of methods is no longer important</dt>
<dd>When calling a method from <code>matplotlib</code>, it is directly
applied to the figure, and not deferred to the final render call. Then, if
calling <code>matplotlib.pyplot.legend()</code> <strong>before</strong>
having actually <code>plot</code>ted anything, it will fail. This is not
the case with this module, as it abstracts on top of
<code>matplotlib</code> and do the actual render only when the figure is
to be <code>show</code>n. Even after having called the <code>show</code>
method, you can still change everything in your figure!</dd>

<dt>Does not interfere with <code>matplotlib</code></dt>
<dd>You can still use the default <code>matplotlib</code> if you want, as
<code>matplotlib</code> state and parameters are not directly affected by
this module, contrary to what <code>seaborn</code> do when you import it
for instance.</dd>

<dt>Useful aliases</dt>
<dd>You think <code>loc="top left"</code> is easier to remember than
<code>loc="upper left"</code> in a <code>matplotlib.pyplot.legend()</code>
call? No worry, this module aliases it for you! (same for "bottom" with
respect to "lower"). Similarly, you can use <code>xrange</code> or
<code>xlim</code> arguments to specify axes ranges (respectively
<code>yrange</code> / <code>ylim</code>).</dd>

<dt>Automatic legend</dt>
<dd>If any of your plots contains a <code>label</code> keyword, a legend
will be added automatically on your graph (you can still explicitly tell
it not to add a legend by setting the <code>legend</code> attribute to
<code>False</code>).</dd>

<dt>Use <code>LaTeX</code> rendering in <code>matplotlib</code>, if
available.</dt>
<dd>If <code>replot</code> finds <code>LaTeX</code> installed on your
machine, it will overload <code>matplotlib</code> settings to use
<code>LaTeX</code> rendering.</dd>

<dt>Handle subplots more easily</dt>
<dd>Have you ever struggled with <code>matplotlib</code> to define a subplot
grid and arrange your plot? <code>replot</code> lets you describe your
grid visually using ascii art!</dd>

<dt>"Gridify"</dt>
<dd>You have some plots that you would like to arrange into a grid, to
compare them easily, but you do not want to waste time setting up a grid
and placing your plots at the correct place? <code>replot</code> handles
it for you out of the box!</dd>

<dt>Easy plotting in log scale, orthonormal axis etc</dt>
<dd><code>replot</code> defines <code>logplot</code> and
<code>loglogplot</code> shortcuts functions to plot in <em>log</em> scale
or <em>loglog</em> scale. Use <code>orthonormal=True</code> on a
<code>plot</code> command to plot using orthonormal axes.</dd>

Examples

A more up to date doc is still to be written, but you can have a look at the Examples.ipynb Jupyter notebook for examples, which should cover most of the use cases.

License

This Python module is released under MIT license. Feel free to contribute and reuse. For more details, see LICENSE.txt file.

Thanks

• Matplotlib for their really good backend.
• Seaborn and prettyplotlib which gave me the original idea.
• This code from scipy central for a base code for adaptive sampling.
• Palettable for palettes.
• Cubehelix colorscheme for nice black and white printing and desaturation compatibility.