How can I set multiple conditions for colours of a bar chart? Specifically, I want the color scheme for values < 0 to be "reds", values > 0 to "lightgray" and the "Gold" name to be "goldenrod" color. I am using Altair 5.5.

Here is what I have tried:

np.random.seed(1)
test = pd.DataFrame(
    {
        'Name': ['Cat_A', 'Cat_B', 'Cat_C', 'Cat_D', 'Gold'],
        'Value': np.random.randint(-10,10,5),
    },
).sort_values(by='Value', ascending=False).reset_index(drop=True).reset_index()
test.rename(columns={'index': 'Rank'}, inplace=True)

color = (
    alt.when(alt.datum.Value < 0)           # check if Value < 0
    .then(alt.Color("Rank:O").scale(scheme='reds').legend(None)) # then 'reds'
    #.when(alt.datum.Name == 'Gold')        # check if Name is 'Gold'
    #.then(alt.value('goldenrod'))          # If true, use 'goldenrod'
    .otherwise(alt.value('lightgray'))      # Default color if neither is true
)
alt.Chart(test).mark_bar(tooltip=True).encode(
    x=alt.X('Value:Q').title('Change in Score'),
    y=alt.Y('Name:N', sort=test['Name'].to_list()).title(None),
    color=color
)

I can only get two conditions to work, and not the "goldenrod" color (which I have commented out above in order to get this):

How can I set multiple conditions for colours of a bar chart? Specifically, I want the color scheme for values < 0 to be "reds", values > 0 to "lightgray" and the "Gold" name to be "goldenrod" color. I am using Altair 5.5.

Here is what I have tried:

np.random.seed(1)
test = pd.DataFrame(
    {
        'Name': ['Cat_A', 'Cat_B', 'Cat_C', 'Cat_D', 'Gold'],
        'Value': np.random.randint(-10,10,5),
    },
).sort_values(by='Value', ascending=False).reset_index(drop=True).reset_index()
test.rename(columns={'index': 'Rank'}, inplace=True)

color = (
    alt.when(alt.datum.Value < 0)           # check if Value < 0
    .then(alt.Color("Rank:O").scale(scheme='reds').legend(None)) # then 'reds'
    #.when(alt.datum.Name == 'Gold')        # check if Name is 'Gold'
    #.then(alt.value('goldenrod'))          # If true, use 'goldenrod'
    .otherwise(alt.value('lightgray'))      # Default color if neither is true
)
alt.Chart(test).mark_bar(tooltip=True).encode(
    x=alt.X('Value:Q').title('Change in Score'),
    y=alt.Y('Name:N', sort=test['Name'].to_list()).title(None),
    color=color
)

I can only get two conditions to work, and not the "goldenrod" color (which I have commented out above in order to get this):

• altair

Share Improve this question Follow asked Feb 21 at 4:44 footfalconfootfalcon 64177 silver badges1818 bronze badges

Add a comment  | 

1 Answer 1

Sorted by: Reset to default 0

I believe the basic usage of chaining when().then() multiple times is that it implies a priority: i.e. the FIRST when() that is true will take effect.

In the example shown here, "Gold" has a value < 0, and therefore (were it not commented) it would still trigger the first when() and be colored with the Reds scale, and this plot would look the same as it does now.

On the other hand, if name=="Gold" was the first when().then(), I think you would get the result you want.

-- UPDATE --

I'm just coming back to Altair ver 5+, and it seems the new when().then() syntax does not like mixing in alt.Color() encodings, and returns an error for me "TypeError: Chained conditions cannot be mixed with field conditions."

So instead, here's an old school hacky way of doing it. Still works, I tested it this time:

condition1 = alt.condition(alt.datum.Name == 'Gold', alt.value('goldenrod'), alt.Color("Rank:O").scale(scheme='reds').legend(None))
condition2 = alt.condition(alt.datum.Value >= 0, alt.value('lightgray'), alt.Color("Rank:O").scale(scheme='reds').legend(None))
combined_condition = condition1.copy()
combined_condition['condition'] = [condition1['condition'], condition2['condition']]


alt.Chart(test).mark_bar(tooltip=True).encode(
    x=alt.X('Value:Q').title('Change in Score'),
    y=alt.Y('Name:N', sort=test['Name'].to_list()).title(None),
    color=combined_condition
)

Is cobbling together two alt.condition()s elegant? Nope, but I hope this helps. It might break with small variations because I don't know if this hack is intended, but it works for this example.