René's URL Explorer Experiment

Title: Cycler by tacaswell · Pull Request #4258 · matplotlib/matplotlib · GitHub

Open Graph Title: Cycler by tacaswell · Pull Request #4258 · matplotlib/matplotlib

X Title: Cycler by tacaswell · Pull Request #4258 · matplotlib/matplotlib

Description: a = SingleCycler('c', ['r', 'purple', 'lime']) b = SingleCycler('lw', [1, 5]) c = SingleCycler('markevery', [10, 50]) m = SingleCycler('marker', 'o') # note length 1 fig, ax_lst = plt.subplots(2, 2) def cycle_demo(ax, cycler): c_len = len(cycler) th = np.linspace(0, 2*np.pi, 1024) y = np.sin(th) for j, p in zip(range(2*c_len), cycler): ax.plot(th, j + y, **p) for ax, cc in zip(ax_lst.ravel(), [a, a + b, a*b, (c*m)*a]): cycle_demo(ax, cc) As discussed in several places the problem of doing arbitrary style cycles over arbitrary parameters and inputs is a bit vexing. Inspired by Holoviews and sensible composable objects, I have implemented Cyclers which give an easy API to construct these cycles. There are two types of composition + which is an 'inner' product implemented by zipping the two cycles together and * which is an 'outer' product which is implemented with itertools.product. There are two implementation (one of which will be deleted), but am a bit torn on which one I like better. The Cycler class is a stand alone class which handles both 'single' cycles and 'compound' cycles via a classmethod constructor and some implicit magic. The other is a set of three classes {_base_cycler, SingleCycler, and CompoundCycler} which handle respectively the composition logic, 'single' cycles and compound cycles. I am leaning towards the 3 class implementation because it has less implicit magic and users don't have to use a class method to get useful objects. In both cases, the objects have len, __iter__ (which is infinite), finite_iter (which in __iter__, but finite), and keys (which says what properties the cycle is over). The composition fails if the same property appears on both sides. These are useful on their own, but I think replacing the color cycle on Axes with one of these would be a major usability gain. I am not wedded to any of the class names. Documentation, examples, and testing will come later if people think this is a good idea. Should probably also add an in-place version of + and * as well. attn @WeatherGod @efiring @mdboom @pelson @danielballan

Open Graph Description: a = SingleCycler('c', ['r', 'purple', 'lime']) b = SingleCycler('lw', [1, 5]) c = SingleCycler('markevery', [10, 50]) m = SingleCycler('marker...

X Description: a = SingleCycler('c', ['r', 'purple', 'lime']) b = SingleCycler('lw', [1, 5]) c = SingleCycler('markevery&...

Opengraph URL: https://github.com/matplotlib/matplotlib/pull/4258

X: @github

direct link

Domain: github.com

Links:

Viewport: width=device-width