Working with Matplotlib in Virtual environments

When running Matplotlib in a virtual environment you may discover a few issues. Matplotlib itself has no issue with virtual environments. However, some of the external GUI frameworks that Matplotlib uses for interactive figures may be tricky to install in a virtual environment. Everything below assumes some familiarity with the Matplotlib backends as found in What is a backend?.

If you only use the IPython and Jupyter Notebook’s inline and notebook backends, or non-interactive backends, you should not have any issues and can ignore everything below.

Likewise, the Tk framework (TkAgg backend) does not require any external dependencies and is normally always available. On certain Linux distributions, a package named python-tk (or similar) needs to be installed.

Otherwise, the situation (at the time of writing) is as follows:

GUI framework	pip-installable?	conda or conda-forge-installable?
PyQt5	on Python>=3.5	yes
PyQt4	PySide: on Windows and OSX	yes
PyGObject	no	on Linux
PyGTK	no	no
wxPython	yes [1]	yes

[1]	OSX and Windows wheels available on PyPI. Linux wheels available but not on PyPI, see https://wxpython.org/pages/downloads/.

In other cases, you need to install the package in the global (system) site-packages, and somehow make it available from within the virtual environment. This can be achieved by any of the following methods (in all cases, the system-wide Python and the virtualenv Python must be of the same version):

• Using virtualenv’s --system-site-packages option when creating an environment adds all system-wide packages to the virtual environment. However, this breaks the isolation between the virtual environment and the system install. Among other issues it results in hard to debug problems with system packages shadowing the environment packages. If you use virtualenvwrapper, this can be toggled with the toggleglobalsitepackages command.
• vext allows controlled access from within the virtualenv to specific system-wide packages without the overall shadowing issue. A specific package needs to be installed for each framework, e.g. vext.pyqt5, etc. It is recommended to use vext>=0.7.0 as earlier versions misconfigure the logging system.

If you are using Matplotlib on OSX, you may also want to consider the OSX framework FAQ.