plotnine.stage

stage(start=None, after_stat=None, after_scale=None)

Stage allows you evaluating mapping at more than one stage

You can evaluate an expression of a variable in a dataframe, and later evaluate an expression that modifies the values mapped to the scale.

Parameters

start : str | array_like | scalar = None: Aesthetic expression using primary variables from the layer data.
after_stat : str = None: Aesthetic expression using variables calculated by the stat.
after_scale : str = None: Aesthetic expression using aesthetics of the layer.

Examples

%load_ext autoreload
%autoreload 2
%aimport plotnine

import pandas as pd
import numpy as np

from plotnine import (
    ggplot,
    aes,
    after_stat,
    stage,
    geom_bar,
    geom_text,
    geom_bin_2d,
    stat_bin_2d,
)

stage

df = pd.DataFrame({
    "var1": list("abbcccddddeeeee"),
    "cat": list("RSRSRSRRRSRSSRS")
})

(
    ggplot(df, aes("var1"))
    + geom_bar()
)

Add the corresponding count on top of each bar.

(
    ggplot(df, aes("var1"))
    + geom_bar()
    + geom_text(aes(label=after_stat("count")), stat="count")
)

Adjust the y position so that the counts do not overlap the bars.

(
    ggplot(df, aes("var1"))
    + geom_bar()
    + geom_text(
        aes(label=after_stat("count"), y=stage(after_stat="count", after_scale="y+.1")),
        stat="count",
    )
)

Note that this will work even nicely for stacked bars where adjustig the position with nudge_y=0.1 would not.

(
    ggplot(df, aes("var1", fill="cat"))
    + geom_bar()
    + geom_text(
        aes(label=after_stat("count"), y=stage(after_stat="count", after_scale="y+.1")),
        stat="count",
        position="stack",
    )
)

Create a binned 2d plot with counts

np.random.seed(123)
df = pd.DataFrame({
    "col_1": np.random.rand(1000),
    "col_2": np.random.rand(1000)
})

(
    ggplot(df, aes(x="col_1", y="col_2"))
    + geom_bin_2d(position="identity", binwidth=0.1)
)

Add counts to the bins. stat_bin_2d bins are specified using retangular minimum and maximum end-points for dimension; we use these values to compute the mid-points at which to place the counts.

First x and y aesthetics are mapped to col_1 and col_2 variables, then after the statistic consumes them and creates xmin, xmax, ymin & ymax values for each bin along with associated count. After the statistic computation the x and y aesthetics do not exist, but we create meaningful values using the minimum and maximum end-points.

Note that the binning parameters for the geom and stat combination must be the same. In this case it is the binwidth.

(
    ggplot(df, aes(x="col_1", y="col_2"))
    + geom_bin_2d(position="identity", binwidth=0.1)
    + stat_bin_2d(
        aes(
            x=stage(start="col_1", after_stat="(xmin+xmax)/2"),
            y=stage(start="col_2", after_stat="(ymin+ymax)/2"),
            label=after_stat("count"),
        ),
        binwidth=0.1,
        geom="text",
        format_string="{:.0f}",
        size=10,
    )
)

Source: stage