Creating Marimekko chart

[Sunamur]

I am trying to create a Marimekko chart - Altair's grammar seems capable enough to generate them.

That is my best attempt yet:

data = {
    'year':[2017,2017,2017,2017,2018,2018,2018,2018,2019,2019,2019,2019],
    'yearly_weight':[340,340,340,340,170,170,170,170,520,520,520,520],
    'category':['food','medicine', 'rent','other','food','medicine', 'rent','other','food','medicine', 'rent','other',],
    'category_weight':[12,15,14,13,16,22,35,15,17,17,10,23]
}

df = pd.DataFrame(data)
alt.Chart(df).mark_bar().encode(
    x='year:O',
    y=alt.Y('category_weight:Q',stack="normalize" ),
    color='category:N',
    size='yearly_weight:Q'
).properties(width=100,height=100)

The problem seems to be that X ticks may only be spaced evenly, which is not desirable when creating Marimekko.

Is there any way to fix this?

[jakevdp]

Hmm... I don't think this is well supported by the grammar, but you can do something like this by transforming the data. Not perfect, but a rough idea:

import altair as alt
import pandas as pd

df = pd.DataFrame({
    'year': [2017, 2018, 2019],
    'year_weight': [340, 170, 520],
    'food': [12, 16, 17],
    'medicine': [15, 22, 17],
    'rent': [14, 35, 10],
    'other': [13, 15, 23]
})

alt.Chart(df).transform_stack(
    stack='year_weight',
    as_=['x1', 'x2'],
    offset='normalize',
    groupby=[],
).transform_fold(
    ['food', 'medicine', 'rent', 'other'],
    as_=['category', 'category_weight'],
).transform_stack(
    stack='category_weight',
    groupby=['year'],
    offset='normalize',
    as_=['y1', 'y2'],
).mark_rect().encode(
    x='x1:Q', x2='x2:Q',
    y='y1:Q', y2='y2:Q',
    color='category:N',
    tooltip=['year:N', 'category:N']
)

[Sunamur]

Thanks for reply, @jakevdp !
I would remove first normalize (so that the X axis is more informative) and add borders to divide the blocks.

import altair as alt
import pandas as pd



df = pd.DataFrame({
    'year': [2017, 2018, 2019],
    'year_weight': [340, 170, 520],
    'food': [12, 16, 17],
    'medicine': [15, 22, 17],
    'rent': [14, 35, 10],
    'other': [13, 15, 23]
})

alt.Chart(df).transform_stack(
    stack='year_weight',
    as_=['x1', 'x2'],
    groupby=[],
).transform_fold(
    ['food', 'medicine', 'rent', 'other'],
    as_=['category', 'category_weight'],
).transform_stack(
    stack='category_weight',
    groupby=['year'],
    offset='normalize',
    as_=['y1', 'y2'],
).mark_rect(strokeWidth=0.3).encode(
    x=alt.X('x1:Q', title='Yearly expenses'), 
    x2='x2:Q',
    y=alt.Y('y1:Q',title='Expense item share'), 
    y2='y2:Q',
    fill='category:N',
    tooltip=['year:N', 'category:N'],
    stroke=alt.value('black'),
).configure_axis( grid=False )

Biggest challenge now would be adding nominal X labels (year in our case) without tooltips or legend.

Big fan of Altair, almost completely switched to it for doing work stuff!