How can I make one chart "span" the height / width of multiple charts in concatenated plots?

Question

In the following example, I'd like to have exp fill the entire width of the plot (so, as wide as sin and cos combined). By default, exp fills only one column.

How can I do this, ideally using some "auto" setting (without explicitly setting a width= to a number)?

import altair as alt
import numpy as np
import pandas as pd

x = np.linspace(0, 20, 100)
source = pd.DataFrame({
  'x': x,
  'sin': np.sin(x),
  'cos': np.cos(x),
  'exp': np.exp(x),
})

base = alt.Chart(source).mark_line().encode(x='x')
sin = base.encode(y='sin')
cos = base.encode(y='cos')
exp = base.encode(y='exp')
(sin | cos) & exp

Answer 1

A similar solution for width:

w = 200
spacing = 64

base = alt.Chart(source).mark_line().encode(x='x')
sin = base.encode(y='sin').properties(width=w, height=100)
cos = base.encode(y='cos').properties(width=w, height=100)
exp = base.encode(y='exp').properties(width=2*w+spacing, height=100)

row = alt.hconcat(sin, cos).properties(
    bounds='flush',
    spacing=spacing
)
                  
alt.vconcat(row, exp).configure_axisY(labelLimit=spacing)

Not aligned, but maybe visually nice enough with less code:

w = 200

base = alt.Chart(source).mark_line().encode(x='x')
sin = base.encode(y='sin').properties(width=w, height=100)
cos = base.encode(y='cos').properties(width=w, height=100)
exp = base.encode(y='exp').properties(width=2*w, height=100)

row = alt.hconcat(sin, cos)
alt.vconcat(row, exp).properties(center=True)