CanD - Easily create complex layouts in Matplotlib

Do you plot in Matplotlib, build your diagrams in Inkscape or Illustrator, and then lay out your figures in InDesign or Powerpoint? Do you wish you could easily generate scientific figures exclusively in a single Python script? CanD gives you the best of all possible worlds.

CanD features

CanD (CANvas Designer) is a system for laying out axes and plot elements in a figure within Python and Matplotlib. It provides:

• A framework for precisely specifying the position of subplots. Axes may be individually positioned, positioned in grids, or a combination.
• Improved font management. Alternative to the matplotlib font manager. Any system font may be easily used. Fonts are synchronized across all text present in the figure, including mathematical equations. Currently this is the only reliable cross-platform method for synchronizing fonts in matplotlib to our knowledge.
• An affine algebra system for specifying positions. Points and vectors from different coordinate systems may be used interchangeably. This allows for unprecedented control over element positioning.
• A unified interface for plot elements. This allows lines, arrows, polygons, and text to be drawn with a single line of code. For example, it is easy to draw an arrow from a data point in a scatterplot to a plot in a different subpanel.
• Easy positioning of raster (.png) images. Currently, importing an image directly into matplotlib is difficult and results in aliasing artifacts. CanD allows .png images to be positioned just like any other plot element.

CanD does not provide:

• Methods for plotting data. CanD gives you a reference to the axis object, with which you may use any of matplotlib's standard plotting functions. Seaborn is a popular library which creates elegant and attractive plots with little effort, and is mostly compatible with CanD.
• Zero-thought layouts. CanD will not find a good layout for you. If you want automatically-positioned subplots, CanD is not the right choice for you. CanD will, however, make it easy to implement a layout you have in your mind.

CanD is currently pre-alpha software. Examples and documentation will be available eventually. Draft documentation on readthedocs is currently available.

Example

from cand import Canvas, Vector, Point
import matplotlib.pyplot as plt
import seaborn as sns
import numpy as np

# Create a canvas 10 cm x 15 cm.  Use Lucida as the font.
c = Canvas(15, 10, "cm")
c.set_font("Lucida Std")

# Add an axis from the point (2,2) to (14,8) in centimeters.  Name it "sinewave", 
# and plot a sine wave on it.
c.add_axis("sinewave", Point(2, 2, "cm"), Point(14, 8, "cm"))
ax = c.ax("sinewave")
ax.plot(np.linspace(0, 2*np.pi, 500), np.sin(np.linspace(0, 2*np.pi, 500)))
sns.despine(ax=ax)

# Create an inset axis which is 1 x 1 inch, positioned on the right side of the 
# plot, 75% of the way across and 60% up on the "sinewave" axis.  Name it "sineinset".
insetsize = Vector(1, 1, "inches")
insetloc = Point(.75, .6, "axis_sinewave")
c.add_axis("sineinset", insetloc, insetloc+insetsize)
c.ax("sineinset").plot(np.linspace(0, np.pi), np.sin(np.linspace(0, np.pi)))

# Add a arrow pointing to the peak of one sine wave to the other, i.e. (pi/2,1) 
# in data coordinates (i.e. units of the "sinewave" or "sineinset" axis).
c.add_arrow(Point(np.pi/2, 1, "sinewave"), Point(np.pi/2, 1, "sineinset"))

# Add a watermark of the CanD logo in the lower right corner.
c.add_image("cand-logo.png", Point(1, 0, "figure"), width=Vector(2, 0, "cm"), 
            ha="right", va="bottom")

# Save as a .pdf file, and then show the result
c.save("demo_plot.pdf")
c.show()

Why CanD?

Matplotlib is a powerful plotting library, but as a result of this power comes complexity. Publication-quality figures require complex layouts with tight positioning and alignment of multiple display elements across different axes. They also often require mixing raster and vector images, and tight control over font size and style. These are in practice quite difficult in matplotlib, not only because of the amount of code required for these tasks, but also due to the difficulty of keeping the different parts of the figure in sync with respect to alignment and style.

Due to this difficulty, users will often export plots from matplotlib and manually position them within PowerPoint, Inkscape, InDesign, or other software. Without extra care, font sizes and line thicknesses may not be consistent, and figures can easily get out of sync with one another.

Installation

CanD depends on:

• numpy
• matplotlib
• seaborn
• paranoid-scientist (version 0.2.1 or greater)
• PyMuPDF
• Pillow

Install with:

pip install CanD

or by downloading the package from github and running:

python setup.py install