Set color for each bar

[lwq-star]

Hello, when I was drawing a bar chart, I noticed that ultraplot cannot set a color for each individual bar. Could you add this feature?

import ultraplot as uplt
import numpy as np
import pandas as pd

fig, ax = uplt.subplots(journal='nat1')

x = np.arange(5)
y = [10, 15, 7, 12, 9]
data = pd.DataFrame(y, index=x)

cycle = uplt.Cycle('RdYlGn', 5)
obj = ax.bar(data, cycle=cycle, edgecolor='black')

[cvanelteren]

I think what you want is this:

import numpy as np
import pandas as pd
import ultraplot as uplt

x = np.arange(5)
y = [10, 15, 7, 12, 9]
data = pd.DataFrame(y, index=x)

cycle = uplt.Cycle("RdYlGn", 5)

fig, ax = uplt.subplots(journal="nat1")
for idx, row in data.iterrows():
    ax.bar(idx, row.iloc[0], cycle = cycle, edgecolor = "black")

By default the "unit" you are selecting is put a bar at every x. The first color will be red for thos bars. If you call it again it will draw the next color. In your case you want $n$ different bars for each color.
Matplotlib behaves in a similar way IIRC.

[lwq-star]

When I tried to add labels for each bar, it strangely only drew a single bar. Does this mean I have to add the labels manually myself?

import numpy as np
import pandas as pd
import ultraplot as uplt

x = np.arange(5)
y = [10, 15, 7, 12, 9]
data = pd.DataFrame(y, index=x)

cycle = uplt.Cycle("RdYlGn", 5)

fig, ax = uplt.subplots(journal="nat1")
for idx, row in data.iterrows():
    ax.bar(idx, row.iloc[0], cycle=cycle, edgecolor="black", bar_labels=True)

[cvanelteren]

Ah that is strange. That should not happen. The bars are still drawn, .e.g:

fig, ax = uplt.subplots(journal="nat1")
for idx, row in data.iterrows():
    ax.bar(idx, row.iloc[0], cycle=cycle, edgecolor="black", bar_labels=True)
ax.set_xlim(-1, 5)

Let me see what is going on.

[cvanelteren]

Found the issue.

[cvanelteren]

I can also check if we can add the feature you describe to dispatch the colors for the bars

[cvanelteren]

The issue is that we do allow for this

ax.bar(
    x,
    y,
    colors=[cycle.get_next()["color"] for _ in x],
    edgecolor="black",
    bar_labels=True,
)

which is confusing

[cvanelteren]

@beckermr should we allow for expanding a cycler if we plot something that only has 1 column or leave it as is? The above snippets is a bit confusing that we allow for that but not for the cycler.

I think allow for expansion is a bit of an issue too. For example, say I do plot with a loop then the bars will be colored and expanded using the cycler which is not intended.

[beckermr]

I don't understand the question. Can you provide more detail?

[cvanelteren]

Currently when you call bar it will use the cycle per call; if you want the colors for each bar to be set you have to go through colors and specifically obtain the colors per bar; we could allow for cycle to be set on colors and then achieve the effect that the user wants.

[beckermr]

Ah I see. The whole approach of ultraplot, at least as far as I know, is to have the same semantics on various aspect of plotting apply across all plot commands. So if we allowed passing a cycle to colors, what would that mean for other plotting commands? For example, what does this imply for plot or scatter?

[cvanelteren]

Yeah for plotting the semantics would be strange. For scatter too. I think we should keep it as is.