Feature Request: Support list inputs for plotly express x, y arguments

[samjett247]

First off, thanks for the awesome plotly.express package, its fulfilling a great need for interactive plotting capabilities.

One problem I've encountered frequently is that I have variables I want to plot in separate columns in my dataframe, and I have to make a subplot to plot the distributions or correlations of all of these variables. For these purposes, it would be very helpful if px.scatter, px.box , and the other px methods could accept the lists of column names as x or y variables. Note that this is different from faceting by column or row, because for the facet: (a) the yaxes and/or xaxes are typically shared, and (b) the "column" or row to be faceted on is a single column. Would love to hear the communities thoughts about this idea. Example code below.

import plotly.express as px
from plotly.subplots import make_subplots
import pandas as pd

temp = [
    {"Clinic": "A", "Subject": "Bill", "Height(cm)": 182, "Weight(kg)": 101, "BloodPressureHigh": 128, "BloodPressureLow": 90},
    {"Clinic": "A", "Subject": "Susie", "Height(cm)": 142, "Weight(kg)": 67, "BloodPressureHigh": 120, "BloodPressureLow": 70},
    {"Clinic": "B", "Subject": "John", "Height(cm)": 202, "Weight(kg)": 89, "BloodPressureHigh": 118, "BloodPressureLow": 85},
    {"Clinic": "B", "Subject": "Stacy", "Height(cm)": 156, "Weight(kg)": 78, "BloodPressureHigh": 114, "BloodPressureLow": 76},
    {"Clinic": "B", "Subject": "Lisa", "Height(cm)": 164, "Weight(kg)": 59, "BloodPressureHigh": 112, "BloodPressureLow": 74} 
]
df = pd.DataFrame(temp)
# This is the call I want to be able to make
# px.box(data_frame=df, x='Clinic', y=['Height(cm)', 'Weight(kg)', "BloodPressureHigh", "BloodPressureLow"]) 

# Instead, this is what I have to do
fig = make_subplots(rows=1, cols=4)
for j,y in enumerate(['Height(cm)', 'Weight(kg)', "BloodPressureHigh", "BloodPressureLow"]):
    localfig = px.box(data_frame=df, x='Clinic', y=y) 
    # Or the below if you want to plot scatter
    # localfig = px.scatter(data_frame=df, x='Height(cm)', y=y)
    trace1 = localfig['data'][0]
    fig.add_trace(trace1, row=1, col=j+1)
fig.show()

[samjett247]

Thanks @nicolaskruchten for the helpful example - just what I needed!

[nicolaskruchten]

We've actually gone ahead an implemented this feature after all! https://medium.com/plotly/beyond-tidy-plotly-express-now-accepts-wide-form-and-mixed-form-data-bdc3e054f891

[eisenlohr]

Support for wide format is very nice to have. However, what I am struggling with is the issue that a px.line plot connects traces resulting from different columns as demonstrated in this MWE:

df = pd.DataFrame(data={'who':['a','a','b','b'],
                        'x':[0,1,0,1],
                        'score':[1.0,2,3,4],
                        'miss':[3.2,2.5,1.3,1.5],
                        })
px.line(df,x='x',y=['miss','score'],color='who',width=300,height=300).show()

My hope/expectation was that above code results in the union of below two separate plots, i.e. without connecting the data across different columns

px.line(df,x='x',y=['miss'],color='who',width=300,height=300).show()
px.line(df,x='x',y=['score'],color='who',width=300,height=300).show()

[nicolaskruchten]

You can resolve this by adding line_group="variable", to get the lines to be split by variable as well as by color.

df = pd.DataFrame(data={'who':['a','a','b','b'],
                        'x':[0,1,0,1],
                        'score':[1.0,2,3,4],
                        'miss':[3.2,2.5,1.3,1.5],
                        })
px.line(df,x='x',y=['miss','score'], color='who', line_group="variable", width=300,height=300).show()

[nicolaskruchten]

The default value of color is "variable" so you get this effect automatically, but if you remap color to something else and you want to maintain the line-grouping, you'll need to make this explicit.

[nicolaskruchten]

Perhaps this should be the default :)

[nicolaskruchten]

@eisenlohr as of 4.8.2, line_group="variable" by default for px.line :)