Colorbar levels issue with small range values

[mickaellalande]

I'm having trouble with colorbar recently, here is an example of it (link with #117? ; I didn't update recently I hope this haven't been solved already, proplot version: 0.4.2):

A first example that works (I purposely multiplied the temperature by 0.01 for having small values for next examples):

import xarray as xr
import proplot as plot

da = xr.tutorial.open_dataset('air_temperature').isel(time=0).air

f, axs = plot.subplots(proj='cyl', axwidth=3)

cmap='ColdHot'
levels=plot.arange(-10,10, 1)

m = axs.pcolormesh(((da-273.15)*0.01), cmap=cmap, levels=levels, colorbar='r')

axs.format(labels=True, coast=True, borders=True,
    latlim=(da.lat.min(), da.lat.max()), lonlim=(da.lon.min(), da.lon.max())
)

Then with

levels=plot.arange(-2, 2, 0.1)

It still works but already the 0 is not good...

And if I go for a smaller range:

levels=plot.arange(-1, 1, 0.1)

this time I have an error:

[...]
~/miniconda3/lib/python3.7/site-packages/proplot/axistools.py in __call__(self, x, pos)
    468                 # Hard limit of MAX_DIGITS sigfigs
    469                 string = ('{:.%df}' % min(
--> 470                     abs(np.log10(x) // 1), MAX_DIGITS)).format(x)
    471                 continue
    472             break

ValueError: cannot convert float NaN to integer

Knowing that it is the levels I need to use for another application, so I'm stuck for now.

[lukelbd]

This can be temporarily fixed by applying a SimpleFormatter with explicitly limited decimal precision (see #117), e.g. as follows:

ax.colorbar(..., formatter='simple')

Note that by default, this formatter still trims trailing zeros (this can also be turned off with the zerotrim argument).

I'm pretty busy this week but will try to fix this by next weekend; this issue has high priority. Thanks for the report.

[mickaellalande]

ax.colorbar(..., formatter=('simple', 3))

Ok, great thanks for the trick! It works fine so far like this 😄

[lukelbd]

This issue should be fixed for >=99% of cases by 9b7f89f, specifically the lines:

https://github.com/lukelbd/proplot/blob/9b7f89fdad3a8e127570aeccbe138c9c089d1fff/proplot/ticker.py#L467-L491