splicing colormaps + LinLogNorm about pygsti

[jarthurgross]

The error is being thrown because the way norm is being called in this example causes it to set its vmin and vmax to the same value (11). This causes it to return NaN since there are divisions by vmax - vmin when called on data, which is what the error is all about.

In color_boxplot, norm is called on a set of data rather than a single point, so unless the data are identical a different vmin and vmax are set and you don't get this error. In your example, you can also manually set vmin and vmax when you define norm

linlog_trans = 11
vmin = 0
vmax = 20
norm = pygsti.report.plotting.LinLogNorm(vmin=vmin, vmax=vmax, trans=linlog_trans)

and you won't experience any errors.

For our purposes, the defaults of splice_cmap are set up so you shouldn't have to mess with them. LinLogNorm is designed to linearly map all values below trans to the interval [0, 0.5], and logarithmically map all values above trans to the interval [0.5, 1]. This means that when you have vmin and vmax set sensibly, norm(linlog_trans) will always be 0.5. splice_cmaps is set to default to assigning the different cmaps to equal intervals, so when joining two cmaps the transition from one to the other will occur at 0.5, matching the behavior of LinLogNorm.

[Travis-S]

This makes sense, and for now, we should rely on this default behavior. In the future, if we mess around more with the colormaps, we may need to revisit this.

Closing the issue for now.