Splash and Jupyter

Splash provides a custom Jupyter (previously known as IPython) kernel for Lua. Together with Jupyter notebook frontend it forms an interactive web-based development environment for Splash Scripts with syntax highlighting, smart code completion, context-aware help, inline images support and a real live WebKit browser window with Web Inspector enabled, controllable from a notebook.


To install Splash-Jupyter using Docker, run:

$ docker pull scrapinghub/splash-jupyter

Then start the container:

$ docker run -p 8888:8888 -it scrapinghub/splash-jupyter


Without -it flags you won’t be able to stop the container using Ctrl-C.

This command should print something like this:

Copy/paste this URL into your browser when you connect for the first time,
to login with a token:

To view Jupyter, open the suggested location in a browser. It should display an usual Jupyter Notebook overview page.


In older Docker setups (e.g. with boot2docker on OS X) you may have to replace ‘localhost’ with the IP address Docker is available on, e.g. a result of boot2docker ip in case of boot2docker or docker-machine ip <your machine> in case of docker-machine.

Click “New” button and choose “Splash” in the drop-down list - Splash Notebook should open.

Splash Notebook looks like an IPython notebook or other Jupyter-based notebooks; it allows to run and develop Splash Lua scripts interactively. For example, try entering splash:go("you-favorite-website") in a cell, execute it, then enter splash:png() in the next cell and run it as well - you should get a screenshot of the website displayed inline.


By default, notebooks are stored in a Docker container; they are destroyed when you restart an image. To persist notebooks you can mount a local folder to /notebooks. For example, let’s use current folder to store the notebooks:

$ docker run -v `/bin/pwd`/notebooks:/notebooks -p 8888:8888 -it splash-jupyter

Live Webkit window

To view Live Webkit window with web inspector when Splash-Jupyter is executed from Docker, you will need to pass additional docker parameters to share the host system’s X server with the docker container, and use the --disable-xvfb command line flag:

$ docker run -e DISPLAY=unix$DISPLAY \
             -v /tmp/.X11-unix:/tmp/.X11-unix \
             -v $XAUTHORITY:$XAUTHORITY \
             -e XAUTHORITY=$XAUTHORITY \
             -p 8888:8888 \
             -it scrapinghub/splash-jupyter --disable-xvfb


The command above is tested on Linux.

From Notebook to HTTP API

After you finished developing the script using Splash Notebook, you may want to convert it to a form suitable for submitting to Splash HTTP API (see execute and run).

To do that, copy-paste (or download using “File -> Download as -> .lua”) all relevant code. For run endpoint add return statement to return the final result:

-- Script code goes here,
-- including all helper functions.
return {...}  -- return the result

For execute add return statement and put the code inside function main(splash):

function main(splash)
    -- Script code goes here,
    -- including all helper functions.
    return {...}  -- return the result

To make the script more generic you can use splash.args instead of hardcoded constants (e.g. for page urls). Also, consider submitting several requests with different arguments instead of running a loop in a script if you need to visit and process several pages - it is an easy way to parallelize the work.

There are some gotchas:

  1. When you run a notebook cell and then run another notebook cell there is a delay between runs; the effect is similar to inserting splash:wait calls at the beginning of each cell.
  2. Regardless of sandbox settings, scripts in Jupyter notebook are not sandboxed. Usually it is not a problem, but some functions may be unavailable in HTTP API if sandbox is enabled.