[Python+Altair (Vega-lite)]: Streamgraph from excel-like Pandas' Dataframe

Question

I have a Pandas' Dataframe I would like to visualize using Altair, but it looks like the library syntax is yet above my skills:

                 Time     WL1       WL2      WL3      WL4
0 2017-04-05 09:15:00  103448.0  100776.0  71581.1  81118.2
1 2017-04-05 09:30:00  114730.0  113206.0  68600.1  78322.2
2 2017-04-05 09:45:00  117252.0  114189.0  68208.5  77261.0
3 2017-04-05 10:00:00  115569.0  111300.0  67454.7  75347.4
4 2017-04-05 10:15:00  117504.0  114956.0  68709.9  76345.1  
5 2017-04-05 10:30:00  116123.0  112637.0  66210.9  74544.4
6 2017-04-05 10:45:00  111050.0  106342.0  61699.3  70165.2
7 2017-04-05 11:00:00  111221.0  108409.0  66124.6  74842.1
8 2017-04-05 11:15:00  111604.0  106853.0  66950.5  75782.9
9 2017-04-05 11:30:00  113026.0  108071.0  68222.2  77848.0

I cannot find the syntax to create a Streamgraph like this, taken from Altair's tutorial (06-AreaCharts):

I can visualize the single lines but it is not the best representation for my data:

LayeredChart(data,
    layers=[
        Chart().mark_line().encode(
            x='Time:T',
            y='WL1:Q',
        ),

        Chart().mark_line().encode(
            x='Time:T',
            y='WL2:Q',
        ),
        Chart().mark_line().encode(
            x='Time:T',
            y='WL3:Q',
        ),

        Chart().mark_line().encode(
            x='Time:T',
            y='WL4:Q',
        ),
    ]
)

Is there a way to create a Streamgraph without going through a LayeredChart, directly from the multiple-column dataframe?

score 2 · Accepted Answer · answered Jun 01 '17 at 22:28

2

This page contains the code for the stream graph.

By the way, you may want to compare the code with the vega-lite json spec to allow you to better understand the Altair syntax.

answered Jun 01 '17 at 22:28

g3o2

271
3
7

yoonghm · Answer 2 · 2019-05-13T13:21:42.437

It may be very late but I hope this post is useful for others who want to use altair with data from local or remote csv, json files.

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

time = np.array([
    '2017-04-05 09:15:00',
    '2017-04-05 09:30:00',
    '2017-04-05 09:45:00',
    '2017-04-05 10:00:00',
    '2017-04-05 10:15:00',
    '2017-04-05 10:30:00',
    '2017-04-05 10:45:00',
    '2017-04-05 11:00:00',
    '2017-04-05 11:15:00',
    '2017-04-05 11:30:00',
])

wl1 = np.array([
    103448.0,
    114730.0,
    117252.0,
    115569.0,
    117504.0,
    116123.0,
    111050.0,
    111221.0,
    111604.0,
    113026.0,
])

wl2 = np.array([
    100776.0,
    113206.0,
    114189.0,
    111300.0,
    114956.0,
    112637.0,
    106342.0,
    108409.0,
    106853.0,
    108071.0,
])

wl3 = np.array([
    71581.1,
    68600.1,
    68208.5,
    67454.7,
    68709.9,
    66210.9,
    61699.3,
    66124.6,
    66950.5,
    68222.2,
])

wl4 = np.array([
    81118.2,
    78322.2,
    77261.0,
    75347.4,
    76345.1,
    74544.4,
    70165.2,
    74842.1,
    75782.9,
    77848.0,
])

df = pd.DataFrame({'time': time, "wl1": wl1, "wl2": wl2, "wl3": wl3, "wl4": wl4})
df = df.melt('time') # Unpivot df with 'time' as identifier

# Save the data into csv and json format and upload them to a public website
lcsv  = './timeseries.csv'
ljson = './timeseries.json'
rcsv  = 'https://raw.githubusercontent.com/yoonghm/data/master/timeseries.csv'
rjson = 'https://raw.githubusercontent.com/yoonghm/data/master/timeseries.json'
df.to_csv(lcsv, index=False)           # Remove index
df.to_json(ljson, orient='records')    # In records orietation

def plot_line(data, source=str(data)):
    return (
        alt.Chart(data, title='Source: ' + source).mark_line(point=True).encode(
            x=alt.X('time:T',  scale=alt.Scale(zero=False)),
            y=alt.Y('value:Q', scale=alt.Scale(zero=False)),
            color='variable:N',
            tooltip=['time:T', 'value:Q']
        ).properties(
            width=300, height=400
        )
    )

a = plot_line(df, source='DataFrame')
b = plot_line(lcsv, lcsv)
c = plot_line(ljson, ljson)
d = plot_line(rcsv, rcsv)
e = plot_line(rjson, rjson)

a & b & c & d & e

After creating your vega-lite plot, you can make use of vega-lite to handle streaming data as shown in the this post and that post.

[Python+Altair (Vega-lite)]: Streamgraph from excel-like Pandas' Dataframe

2 Answers2