使用 Altair 在堆积面积图中堆积文本

Question

I wanted to know if it is possible to have text marks in the corresponding areas of a stacked area chart.

I used median aggregate to get single X and Y axis values otherwise it shows text all through the edge of the chart. However, this aggregate is not foolproof, as if the chart is a little convoluted, then the X axis position may not be the best possible region for the text to be displayed into.

This is as far as I have got -

X=[1,2,3,4,5,6,7,8,9,1,2,3,4,5,6,7,8,9,1,2,3,4,5,6,7,8,9,1,2,3,4,5,6,7,8,9]
V=[1,1,1,2,4,8,6,4,2,1,2,3,4,5,6,7,6,5,1,1,1,1,4,8,4,2,1,1,1,3,4,5,6,6,5,4]
key=['a', 'b', 'c', 'd']
K = [y for x in key for y in (x)*9]

demo = pd.DataFrame({'X': X, 'V': V, 'K': K})
a = alt.Chart(demo).mark_area().encode(
    x='X:O',
    y='V:Q',
    color='K:N'
)
t = alt.Chart(demo).mark_text().encode(
    x='median(X):O',
    y='median(V):Q',
    text=alt.Text('K:N',)
)
a+t

Issue

• The text is not in its proper region.
• The order of the text is also wrong.

It's not that I don't understand why I have these issues, I do actually(the Y position is not aggregating as "stacked" on top of each other), but I do not know how to solve it or if it is even doable as of now.

Answer 1

I would just build a separate dataframe for the text and use that as source. It is much easier and customizable than doing all sorts of transformations in Altair if such a thing is even possible in this case.

import pandas as pd
import altair as alt

X=[1,2,3,4,5,6,7,8,9,1,2,3,4,5,6,7,8,9,1,2,3,4,5,6,7,8,9,1,2,3,4,5,6,7,8,9]
V=[1,1,1,2,4,8,6,4,2,1,2,3,4,5,6,7,6,5,1,1,1,1,4,8,4,2,1,1,1,3,4,5,6,6,5,4]
key=['a', 'b', 'c', 'd']
K = [y for x in key for y in (x)*9]

demo = pd.DataFrame({'X': X, 'V': V, 'K': K})

# find X position where the sum of V's of K's is the maximum (this is at X=6)
idxmax = demo.groupby(["X"]).sum().idxmax()[0]
# find the cumulative sum of V's at position idxmax and
# take away some offset (4) so the labels go down a bit
# iloc[::-1] reverses the order because we want cumulative to start from the bottom (from 'd')
ypos = demo.groupby(["X", "K"]).sum().loc[idxmax].iloc[::-1].cumsum()["V"] - 4
# crate a new dataframe for the text, X column=idmax, Y column=cumulative ypos, K=key
demotext = pd.DataFrame([[idxmax, y, k] for y,k in zip(ypos.tolist(), key[::-1])],
                        columns=["X", "Y", "K"])


a = (alt.Chart(demo).mark_area()
        .encode(
                x='X:O',
                y='V:Q',
                color='K:N')
    )
t = (alt.Chart(demotext).mark_text()
        .encode(
                x='X:O',
                y='Y:Q',
                text='K:N'
))

a+t

Output

Answer 2

I have come to realize that it may not be possible to do it programmatically or maybe its not worth it because of the complexities involved like peaks that are not aligned. Modifying the data given above slightly highlights this issue well -

import pandas as pd
import altair as alt

X=[1,2,3,4,5,6,7,8,9,1,2,3,4,5,6,7,8,9,1,2,3,4,5,6,7,8,9,1,2,3,4,5,6,7,8,9]
V=[1,1,1,2,4,8,6,4,2,1,2,3,4,5,6,7,6,5,1,1,1,1,4,8,4,2,1,1,5,9,5,3,1,1,1,1]
key=['a', 'b', 'c', 'd']
K = [y for x in key for y in (x)*9]

demo = pd.DataFrame({'X': X, 'V': V, 'K': K})

# get the x and y positions for max values of the graph
demo.groupby('K').max()

# crate a new dataframe for the text, X column=idmax, Y column=cumulative ypos, K=key
demotext = pd.DataFrame([[idxmax, y, k] for y,k in zip(ypos.tolist(), key[::-1])],
                        columns=["X", "Y", "K"])


a = (alt.Chart(demo).mark_area()
        .encode(
                x='X:O',
                y='V:Q',
                color='K:N')
    )
t = (alt.Chart(demotext).mark_text()
        .encode(
                x='X:O',
                y='Y:Q',
                text='K:N'
))
a+t

I think the best and easiest way it to manually construct the data for text if you have a very convoluted graph -

a = (alt.Chart(demo).mark_area()
        .encode(
                x='X:O',
                y='V:Q',
                color='K:N')
    )
t = (alt.Chart(p).mark_text()
        .encode(
                x='X:O',
                y='csum:Q',
                text='K:N'
))
a+t

Where p is -

    K   V   X   csum
0   a   8   6   18
1   b   7   7   7
2   c   8   6   4
3   d   9   3   4