wrong plot for geom_path, when using color and group

[szst11]

when I use a table like:

Group1	Group2	Data1	Data2
A	Y	0	0
A	X	1	0
A	Y	2	0
A	X	3	0
A	X	4	0
B	Y	5	1
B	Y	6	1
B	X	7	1
B	X	8	1
B	X	9	1

and plot that with:

chart = (
    ggplot(
        df,
        aes(
            x='Data1',
            y='Data2',
            color="Group1",
            group="Group2",
        ),
    )
    + geom_path()
)

I get a graph, where Data2 for A is not always 0:
(uploading of the png does not work for me, but hopefully the link does)

Marimo notebook

[TyberiusPrime]

Your group and color axis aren't the same. If you add in a geom_point, this becomes visible. What color should the line connecting a red and a blue dot have?

[szst11]

Hi @TyberiusPrime ,
I added a plotly graph to the notebook to visualise the expected behaviour.
new Notebook

It is the goal, that group and color are not the same to get separated sections of a line(which should stay as a line).
There should not be a connection between the A(always 0) and B(always 1) lines.

[TyberiusPrime]

I think what you need is to have one distinct group2 value per line you want drawn.

Plotnine will connect all points within one group.

I have no clue what plotly is doing - possibly drawing one line per group & color combination?

[szst11]

@TyberiusPrime
yes, one line per group & color combination would also be my expectation.

[TyberiusPrime]

I imagine plotnine does the same ggplot does.

Then you'll need to pass in Group1+Group2 for the group aesthetic.

Which sort of incidentally works, since you've string-typed columns.

Maybe @has2k1 can chime in whether there's an explicit 'interaction of these levels' operator one would use here.

[szst11]

I added a hopefully more precise example in that Notebook (now with the plot):
No datapoint of a is in the high value-range.

[has2k1]

There should not be a connection between the A(always 0) and B(always 1) lines.

In plotnine, there should be a connection because the group aesthetic is special. It does not interact with any other aesthetic. All points that are mapped to the same group will be part of the same path. If you do not map to the group, then it is derived internally after doing the interaction of the discrete aesthetic mappings.

import pandas as pd
from plotnine import *

# Dataframe from original query
df1 = pd.DataFrame({
  "Group1": ["A","A", "A", "A", "A", "B", "B", "B", "B", "B"],
  "Group2": ["Y", "X", "Y", "X", "X", "Y", "Y", "X", "X", "X"],
  "Data1": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
  "Data2": [0, 0, 0, 0, 0, 1, 1, 1, 1, 1]
})

# Simpler dataframe from demonstration
df2 = pd.DataFrame({
    "x": range(5),
    "y": range(5),
    "g1": list("abcde"),
    "g2": "R",
    "g3": list("XXXYY")
})

If you map to some other aesthetic, they you get the interaction you "expect".

(
    ggplot(df1, aes('Data1', 'Data2', color="Group1", shape="Group2"))
    + geom_path(size=2)
    + geom_point(size=2, color="black")
)

This is a valid expectation because we can account for all aesthetics. The colour of any segment within the path is determined by the colour at the starting point.

# All points have different colours, but belong to same group
(
    ggplot(df2, aes("x", "y", color="g1", group="g2"))
    + geom_path(size=2)
    + geom_point(size=2)
)

# All points have different colours, but belong to two groups so there is a disjoint
(
    ggplot(df2, aes("x", "y", color="g1", group="g3"))
    + geom_path(size=2)
    + geom_point(size=2)
)

However, it is due to a limitation of the underlying graphics device that points with different colours get a segment that is a single colour. Ideally it should be a linear gradient from one color to the next.