Exporting Vis From Widget

Note

You can follow along this tutorial in a Jupyter notebook. [Github] [Binder]

In this tutorial, we look at the Happy Planet Index dataset, which contains metrics related to well-being for 140 countries around the world. We demonstrate how you can select visualizations of interest and export them for further analysis.

import pandas as pd
import lux
df = pd.read_csv("lux/data/hpi.csv")
df.default_display = "lux" # Set Lux as default display

Note that for the convienience of this tutorial, we have set Lux as the default display so we don’t have to Toggle from the Pandas table display everytime we print the dataframe.

Exporting one or more visualizations from recommendations

In Lux, you can click on visualizations of interest and export them into a separate widget for further processing.

df
1) scroll through Correlation, then 2) click on any 3 visualization (let's say 2nd, 5th and something towards the end), then 3) click on the export button and make sure the blue message box show up
bookmarked_charts = df.exported
bookmarked_charts
add screenshot of exported VisList (include the Out[] __repr__ string) in screenshot

From the dataframe recommendations, the visualization showing the relationship between GDPPerCapita and Footprint is very interesting. In particular, there is an outlier with extremely high ecological footprint as well as high GDP per capita. So we click on this visualization and click on the export button.

df
1) scroll and find the vis for GDPPerCapita and Footprint 2) select and export this vis
vis = df.exported[0]
vis
add screenshot of exported vis

Setting Vis as the Updated Intent

Often, we might be interested in other visualizations that is related to a visualization of interest and want to learn more. With the exported Vis, we can update the intent associated with dataframe to be based on the selected Vis to get more recommendations related to this visualization.

df.intent = vis
df
add screenshot

Accessing Widget State

We can access the set of recommendations generated for the dataframes via the properties recommendation.

df.recommendation
add screenshot

The resulting output is a dictionary, keyed by the name of the recommendation category.

df.recommendation["Enhance"]
add screenshot

You can also access the vis represented by the current intent via the property current_vis.

df.current_vis
add screenshot

Exporting Visualizations as Code

Let’s revist our earlier recommendations by clearing the specified intent.

df.clear_intent()
df
1) click on `Occurrence` tab, then 2) hover around the SubRegion v.s. Number of Records chart

Looking at the Occurrence tab, we are interested in the bar chart distribution of country SubRegion.

vis = df.recommendation["Occurrence"][0]
vis
add screenshot

To allow further edits of visualizations, visualizations can be exported to code in Altair or as Vega-Lite specification.

print (vis.to_Altair())
add screenshot

This can be copy-and-pasted back into a new notebook cell for further editing.

import altair as alt
visData = pd.DataFrame({'SubRegion': {0: 'Americas', 1: 'Asia Pacific', 2: 'Europe', 3: 'Middle East and North Africa', 4: 'Post-communist', 5: 'Sub Saharan Africa'}, 'Record': {0: 25, 1: 21, 2: 20, 3: 14, 4: 26, 5: 34}})

chart = alt.Chart(visData).mark_bar().encode(
    y = alt.Y('SubRegion', type= 'nominal', axis=alt.Axis(labelOverlap=True), sort ='-x'),
    x = alt.X('Record', type= 'quantitative', title='Count of Record'),
)
chart = chart.configure_mark(tooltip=alt.TooltipContent('encoding')) # Setting tooltip as non-null
chart = chart.configure_title(fontWeight=500,fontSize=13,font='Helvetica Neue')
chart = chart.configure_axis(titleFontWeight=500,titleFontSize=11,titleFont='Helvetica Neue',
            labelFontWeight=400,labelFontSize=8,labelFont='Helvetica Neue',labelColor='#505050')
chart = chart.configure_legend(titleFontWeight=500,titleFontSize=10,titleFont='Helvetica Neue',
            labelFontWeight=400,labelFontSize=8,labelFont='Helvetica Neue')
chart = chart.properties(width=160,height=150)
chart
add screenshot

You can also export this as Vega-Lite specification and vis/edit the specification on Vega Editor.

print (vis.to_VegaLite())
add screenshot of what this looks like in Vega Editor

Let’s say now we are interested in the scatter plot of the HPIRank and HappyPlanetIndex.

vis = df.recommendation["Correlation"][0]

Since the dataset used to create the scatterplot is large, Lux infers the variable name used locally for the data, and uses that as the data in the printed code block.

print (vis.to_Altair())
screenshot of code with df

If we wanted to include the actual data in the returned codeblock, we would use to_Altair(standalone=True)

print (vis.to_Altair(standalone=True))
screenshot of code with embedded data