What is the best plotting library for Python?

Question

What Python plotting library do you recommend?

It should be noted the following considerations:

• is it cross-OS?
• speed
• how clean it's interface?
• "prettyness" of the resulting plots
• License
• etc?

In the Python Wiki there is a list, but there is little comparison between them.

---

Interesting SO links (not quite duplicates, but usefull to someone comming to this question):

• barchart o plot 3d in python
• python plotting libraries
• wavelet plot with python libraries

Answer 1

If you've ever used MATLAB, check out MatPlotLib. The code syntax & plots generated are very similar to MATLAB. You can easily make publication-quality plots with minimal effort.

Answer 2

I've used matplotlib with excellent results.

Answer 3

My coworker successfully used matplotlib for his entire master thesis (psychology). He's recommended this library to me as well, though I only used it briefly.

Answer 4

I used pygooglechart, as it produced pretty charts with no extra work in prettyfication, while the others usually created butt ugly charts unless you put down an effort in choosing the right fonts and colors, etc.

Answer 5

I have been using matplotlib for my plotting, and have been extremely pleased thus far. In addition to creating plots that just plain look nice, there are several features I find compelling:

• The gallery gives a listing of many different plots that can be generated, along with the code that generated the plot. Whenever I'm trying something new, I'll skim through there to see if an example already exists.
• The basemap package may not be useful for everyone, but just about everything we do involves some sort of geo-referencing. It's great to be able to switch map projections, regions, etc. with minimal code.
• It plays well with other packages, producing output in may different formats. I've used it with Django as well to render custom plots directly into the HTTP response.

I've found most of the documentation very helpful, and it seems to be under active development. Combine it with IPython and you have yourself an excellent environment for both interactive data analysis and also publication-quality graphics.

As mentioned, it includes a pylab mode where the plotting commands are similar to MATLAB. I've found this helpful in getting started, but being able to explore a little deeper and deal with the OO interface has been useful as well.

Edit: Since you asked about cross-platform, I've used this package personally on an IBM Power6 (running AIX), a Cray XT5, and Linux and Windows machines. Although I haven't used it personally, OS X is also supported.

Answer 6

It depends on what you mean by "plotting" (some people could write entire dissertations on the differences between plots and charts). If you want really nice looking "charts", which can function as scientifically useful "plots" in most cases, you should take a look at ChartDirector. It has wrappers for Python/C/C++/Java/Ruby/Perl/etc., supports internationalization, is fast, and has a pretty reasonable API (runs on at least Solaris, Windows, Linux, and MacOS X).

One note would be that it is non-free ($99), but imo presents the best looking graphics (full anti-aliasing, etc.) of any available package. There is a free trial available so you can get a feel for it before handing over any cash.

Answer 7

If you are interested in Interactive plots with linked zooming/panning there is the Chaco library. A part of the Enthought Tool Suite

Answer 8

I've recently wondered about this as well, trying to compare Matplotlib with PyQwt for my own uses. Here's the summary, you may find it useful.

Answer 9

I generally use matplotlib, mostly for reasons already mentioned by other answers. But I've been using CairoPlot lately when I need something quick and pretty because the defaults work out nicely. With matplotlib, I have to tweak things a little bit every time.

Answer 10

CairoPlot produces very pretty plots (it uses gradients and pleasing colours by default), but it's not a full featured as matplotlib. So if you're looking for features, matplotlib is probably the way to go, but if you're looking for pretty charts (for UIs, etc.), I'd take a look at CairoPlot as see if it has what you need.

Answer 11

I'll join everyone else in recommending matplotlib but the best feature in my mind is the support for LaTeX. The syntax is really natural as long as you remember to use raw strings:

xlabel(r"$\frac{x^2}{y^4}$")

I think you need to set text.markup = "tex" in your matplotlibrc configuration file for this to work.

If you want every piece of text in the graph to be LaTeX'd including the axis tick labels you can set text.usetex = True in matplotlibrc. This is great for making your graph have the same look as the rest of your paper when using LaTeX.

Answer 12

I found MatPlotLib to be an enormous headache to install and use cross-platform. Not being able to generate 3D plots (even static/non-interactive ones) was the last straw, so I quickly dropped MatPlotLib. My go-to for plotting from Python has since become gnuplot, which has a very nice, simple Python wrapper called Gnuplot.py. If you can do it in gnuplot, you can pretty much do it in Python with Gnuplot.py.

Answer 13

For 3D plotting, in terms of power in speed, Mayavi2 offers probably the best tradeoff as it is very powerful and versatile on big dataset.

It is also however a big dependency.

Answer 14

gplot is a very simple Gnuplot in Python. check it at pyla.codeplex.com

Answer 15

also http://root.cern.ch/drupal/image is quite good (mainly used in particle physics, but aplicable everywhere) Also it ships with LOTS of advanced statistical stuff from simple fitting to MVA. OpenGL support included. CHeers, MadMUd