how to set the default color cycle

[syrte]

I went through the ultraplot document, but failed to find a clear explanation on how to set the default color cycle to a custom color sequence.

The configuration page mentioned rc['cycle'].
The Color cycles page mentioned how to define a cycle with uplt. Cycle, but this cycle can not be passed to rc actually.
These are all relevant info that I found.

I don't know what's the recommended practice with ultraplot.
I found a workaround using a mixture of uplt and mpl.

from matplotlib import pyplot as plt
import matplotlib as mpl
import ultraplot as uplt
import numpy as np


ctab = ["#5790fc", "#f89c20", "#e42536", "#964a8b", "#9c9ca1", "#7a21dd"]
uplt.rc['axes.prop_cycle'] = mpl.cycler(color=ctab)
# uplt.rc['cycle'] = mpl.cycler(color=ctab)  # does not work

for i in range(6):
    plt.plot(np.arange(10) + i)

[cvanelteren]

The rc file is getting on my radar the last couple of days. It currently is not checked for internal validity (that is it lacks tests). I see that the cycle is listed but if what you are saying it correct it is not correctly passing this on. axes.prop_cycle is how it is named under matplotlib. To construct the cycler please us uplt.Cycle

[cvanelteren]

The correct code would be

import ultraplot as uplt, numpy as np
import numpy as np


ctab = ["#5790fc", "#f89c20", "#e42536", "#964a8b", "#9c9ca1", "#7a21dd"]
uplt.rc["cycle"] = uplt.Cycle(color=ctab)
fig, ax = uplt.subplots()
for i in range(6):
    ax.plot(np.arange(10) + i)
uplt.show(block=True)

which currently errors for cycle but works fine for axes.prop_cycle.

[cvanelteren]

So I looked at this again and it seems to automagically work now. Can you confirm this?

[cvanelteren]

Ok I found the issue. The issue is that the docs state that rc["cycle"] maps to "axes.prop_cycle", however, in the code it reads it as a colormap not necessarily a cycler object. I will fix this such that both are possible.

[cvanelteren]

This requires a bit more tooling than I thought. The issue is that the docs are implying that we can set any cycler -- whereas the code only allows colormaps and in particular it allows only to set strings for this property. There is also something funky going on with the colormap names being copied with a different name -- which I have been aware of for a while. A fix for this would also require to fix that. For now I will keep this open and suggest to use "axes.prop_cycle" directly. Apologies for the inconvenience :(

[syrte]

I confirm that uplt.rc["axes.prop_cycle"] = uplt.Cycle(color=ctab) works.
A naive question: if rc["cycle"] is simply supposed to map to "axes.prop_cycle" (though not exactly implemented this way yet), why do we need rc["cycle"] in the first place? Because the item name is shorter?

[cvanelteren]

I think that Luke did it for two reasons; one, like you mentioned the name is shorter, but second the name is more intuitive. The implementation is a bit odd however. For example you can use it like this currently:

uplt.rc["cycle"] = "qual1"

This works perfectly fine. I believe even passing a colormap object works, just not the actual object its name implies its designed for (a cycler).

[cvanelteren]

Fixed on main!