viztricks

Recipes and helper functions for plotting with Python + matplotlib

Installation

To install, clone this repository and run:

python setup.py install

This package depends on numpy and matplotlib. Some functions require optional dependencies:

• voronoi_filled requires scipy
• pca_ellipse requires scikit-learn
• animated GIF support for FigureSaver requires the convert program from imagemagick on the system PATH.

To run the test suite:

./run_tests.sh

Testing requires python packages nose and nose-cov.

Tricks

viztricks is a collection of helper functions which break down into some rough categories:

Convenience

• plot(X, ...): handles most of the common cases when plotting a set of n points represented by a (n,d)-shaped numpy array.
• plot_trajectories(T, ...): similar to plot, but takes a list of numpy arrays to plot as disjoint trajectories.
• imagesc(X, ...): recreates the behavior of Matlab's imagesc function.
• axes_grid(n): constructs a grid of n subplots, laid out for space efficiency.
• vector_field(pts, dirs, ...): handles most of the common cases when working with quiver or quiver3d.

Shims

• violinplot: supplies violin plots, which were added in matplotlib's 1.4 release.

Extensions

• gradient_line(x, y, ...): plots a simple 2d line, but colors each segment according to a colormap based on point order.
• irregular_contour(x, y, z, ...): mimics plt.contourf, but allows the (x,y) data points to be irregularly spaced.
• voronoi_filled(pts, colors): plots a Voronoi diagram with each point's voronoi cell colored.
• pca_ellipse(X, ...): computes and plots an ellipse representing the first two principal components of X. Useful for overlaying on scatter plots.
• embedded_images(X, images, ...): selects a non-overlapping subset of images at locations X and displays them in the figure.

Utilities

• FigureSaver is a context manager for converting code that plots and shows multiple figures into a no-frills animation creator.