Apache Zeppelin comes with basic charts out of the box and even offers the ability to call into the underlying NVD3 library. Sometimes, though, you just want some Plot.ly goodness. While this awesome visualization library can be imported directly within Jupyter, having it work on Zeppelin needs a bit more of work. This repository is one way of making that happen.
It does two things:
- The Zeppelin web app sources
plotly-latest.min.jsfrom the CDN. - It overrides Plot.ly's offline module with the code contained here such that calls to it will render via Zeppelin's display system instead of Jupyter's.
These steps apply to the node that runs Zeppelin.
-
Install the following Python libraries:
-
Clone this repository in a commonly accessible path.
-
Update
index.htmlin zeppelin-web such that it sourcesplotly-latest.min.jsduring startup. The cleaner way would be to modify Zeppelin's code and then compile. But I assume you have already have Zeppelin set up and you just want to get Plot.ly running on it as fast as possible. In this case, we just extractindex.htmlfromzeppelin-web-*.war, insert the script tag for plotly.js, and then addindex.htmlback into the same WAR file. -
Find the
zeppelin-web-*.warfile inside the root folder of your Zeppelin install (i.e.zeppelin-web-0.6.0-incubating-SNAPSHOT.war). Back it up somewhere so you can restore the original copy if you need to. -
Extract
index.htmljar xvf zeppelin-web-0.6.0-incubating-SNAPSHOT.war index.html -
Edit
index.htmlto insertplotly-latest.min.jsscript tag just before the<body>tag.<link rel="stylesheet" ng-href="assets/styles/looknfeel/{{looknfeel}}.css"> <link rel="stylesheet" href="assets/styles/fontImport.css"> <script src="https://cdn.plot.ly/plotly-latest.min.js"></script> <body ng-class="{'bodyAsIframe': asIframe}"> -
Insert this modified
index.htmlback into the WAR file.jar uvf zeppelin-web-0.6.0-incubating-SNAPSHOT.war index.html
Your modifications to the setup are now done and you can just start or restart your Zeppelin instance. In case it fails to start, you probably mistyped something when editing index.html. You can always restore the original WAR file you backed up if all else fails.
-
At the top of a Zeppelin notebook, copy and execute the following code. Replace with the actual path where you cloned the repository.
%pyspark # Load this module from a non-standard path (non-installed package) import sys sys.path.insert(0, "PATH_TO_YOUR_REPO/zeppelin-plotly") import offline # Override Plotly's module loading to use our patch instead. sys.modules["plotly"].offline = offline sys.modules["plotly.offline"] = offline # Import cufflinks (Pandas, etc) like usual. import cufflinks as cf # We need to use Cufflinks in offline mode. cf.go_offline() -
You can then use Plotly and Cufflinks (offline mode) as usual in the suceeding notebook paragraphs.
%pyspark import pandas as pd import numpy as np df = pd.DataFrame(np.random.randn(1000, 4), columns=['a', 'b', 'c', 'd']) df.scatter_matrix()
Check out the showcase notebook included inside the sample folder if you want to see the different visualizations available on Plot.ly. Just import the json file like how you would any other Zeppelin notebookand then point the sys.path.insert bit to the actual location of your cloned repo.
- Maybe contribute this back to the Plot.ly and/or Zeppelin projects so that integration becomes even easier.
- Write tests