[FIXED] How do I get matplotlib to actually update to the latest version?

Issue

I'm trying to change the font color in the legend of a barplot by setting the labelcolor parameter, and it looks like I'll need Matplotlib 3.3 or later to do so (currently I'm getting TypeError: __init__() got an unexpected keyword argument 'labelcolor' when I try). I currently have 3.2.2 installed.

I tried running conda install -c conda-forge matplotlib in both my base environment and a custom environment, and it just "updated" to 3.2.2 again. I'd prefer to avoid trying a pip install so I don't risk screwing up other packages.

Thanks in advance.

Solution

The command

conda install -c conda-forge matplotlib

translates to the imperative

With the conda-forge channel prioritized, ensure that some version of the package matplotlib is installed in the current environment.

In contrast, according to OP, we want the imperative statement

Ensure that at least version 3.3 of the matplotlib package is installed in the current environment.

which translates to the command

conda install matplotlib[version='>=3.3']

This does not guarantee that Conda can satisfy this command (e.g., it could conflict with previous specifications), only that this is the literal translation.

Channel Specification

Note that including -c conda-forge will prioritize that channel, but does not explicit specify that Conda must use that channel to source the package. This is because Conda will also take into account the channel_priority configuration value, and will behave differently depending on whether 'strict' or 'flexible' is set.

However, Conda's MatchSpec is sufficiently expressive to explicitly demand a particular package is sourced from a given channel. For example, to demand that at least version 3.3 is installed from Conda Forge, would take the form

conda install conda-forge::matplotlib[version='>=3.3']

Answered By - merv