Setup Selenium headless (without display) on Ubuntu Server 16.04

Data management is the first skill of a data scientist, in my opinion. The management task generally includes import data from one form into another one (generally a database). But sometimes, we need to collect the data ourselves, for example from Internet.

The data on internet generally stored as HTML, XML, JSON, etc. Web pages can be classified into two categories: static and dynamic pages. There are a lot of softwares/libraries/packages for extracting data from static web pages, however for dynamic pages, there is only several tools (Selenium/PhantomJS) available. Among them, Selenium is the most well knowned.

I have been using the package XML, Rcurl of R to extract data from static web pages for years, I tried several times to extract information from dynamic web pages, but found no lucky. Now I feel more and more interesting data stored in dynamic web pages, so I gave myself a push to handle it. There is a R package called RSelenium, which can driver a web browser and extract data from dynamic web pages. Unfortunately, I can’t make it work on my Linux server. (It works on my Windows PC).

I spend a lot time to setup the Selenium on my Ubuntu 16.04, but never get succeeded until I realized that I have no display device (screen). Selenium use web browser to render web page content, and web browser need a display device to show the content. Then, what we do? Can we use selenium with display? Sure, use a virtual display device.

Step 1: install selenium

sudo apt-get install python-pip
sudo pip install selenium

Step 2: install geckodriver

tar -zxvf geckodriver-v0.11.1-linux64.tar.gz
sudo mv geckodriver /usr/bin/

Above command download and unzip deckodriver. The third line command move the excutable deckodriver to a directory where it located in computer’s environment list. So, instead move it /usr/bin/, you can add its path to system variable PATH: export PAHT=$PATH:directory where geckodriver located.

Step 3: install Xvfb & PyVirtualDisplay

Xvfb is a software that simulates a display doing everything in memory and not showing any screen output, while PyVirtualDisplay is a Python wrapper for Xvfb. It allows you to easily work with a virtual display in Pythonref. Use the following command to install them.

sudo apt-get install xvfb
sudo pip install pyvirtualdisplay

OK, now open python in the terminal, and run the following script to test wheter it work.

from pyvirtualdisplay import Display
from selenium import webdriver

display = Display(visible=0, size=(1920, 1080)).start()
browser = webdriver.Firefox()
print browser.title.encode('utf8', 'replace')

Another way: PhantomJS

Above step let Firefox run without display through Xvfb, so now the firefox is actually a headless web browser. But there is another web browser is designed as headless: (PhantomJS)[]. If you are familiar with PhantomJS, you can instead it, instead of running commands in above step.

tar xjf phantomjs-2.1.1-linux-x86_64.tar.bz2
export PATH=$PATH:<DIRECT>/phantomjs-2.1.1-linux-x86_64/bin

Run the following script to tes:

from selenium import webdriver
browser = webdriver.PhantomJS()
print browser.title.encode('utf8', 'replace')

Written on October 28, 2016