Lweb

HWweb

It is 2021. Use python3! The default `python` command on vyuka server is python 2.7. Some of the packages do not work with python2. If you type `python3` you will get python3.

Sometimes you may be interested in processing data which is available in the form of a website consisting of multiple webpages (for example an e-shop with one page per item or a discussion forum with pages of individual users and individual discussion topics).

In this lecture, we will extract information from such a website using Python and existing Python libraries. We will store the results in an SQLite database. These results will be analyzed further in the following lectures.

Scraping webpages

In Python, the simplest tool for downloading webpages is requests package:

import requests
r = requests.get("http://en.wikipedia.org")
print(r.text[:10])

Parsing webpages

When you download one page from a website, it is in HTML format and you need to extract useful information from it. We will use beautifulsoup4 library for parsing HTML.

Parsing a webpage:

import requests
from bs4 import BeautifulSoup

text = requests.get("http://en.wikipedia.org").text

parsed = BeautifulSoup(text, 'html.parser')

Parsed variable contains HTML tree. Since our homework is mostly cared about user comments, we need to find appropriate element with user comments in the HTML tree. Here we show several examples of finding elements of interest and extracting data from them.

To select all nodes with tag <a> you might use:

>>> links = parsed.select('a')
>>> links[1]
<a class="mw-jump-link" href="#mw-head">Jump to navigation</a>

Getting inner text from the element is done by:

>>> links[1].string
'Jump to navigation'

To select and <li> element and traversing children of it, you can use following code. Also note difference between string and text (text contains text from all descendants):

>>> li = parsed.select('li')
>>> li[10]
<li><a href="/wiki/Fearless_(Taylor_Swift_album)" title="Fearless (Taylor Swift album)"><i>Fearless</i> (Taylor Swift album)</a></li>
>>> for ch in li[10].children:
...   print(ch.name)
... 
a
>>> ch.string
>>> ch.text
'Fearless (Taylor Swift album)'

Here are examples of more complicated selection of HTML elements.

parsed.select('li i')  # tag inside a tag (not direct descendant), returns inner tag
parsed.select('li > i')  # tag inside a tag (direct descendant), returns inner tag
parsed.select('li > i')[0].parent  # gets the parent tag
parsed.select('.interlanguage-link-target') # select anything with class = "..." atribute
parsed.select('a.interlanguage-link-target') # select <a> tag with class = "..." atribute
parsed.select('li .interlanguage-link-target') # select <li> tag followed by anything with class = "..." atribute
parsed.select('#top') # select anything with id="top"

• In your code, we recommend following the examples at the beginning of the documentation and the example of CSS selectors. Also you can check out general syntax of CSS selectors.
• Information you need to extract is located within the structure of the HTML document
• To find out, how is the document structured, use Inspect element feature in Chrome or Firefox (right click on the text of interest within the website). For example this text on the course webpage is located within LI element, which is within UL element, which is in 4 nested DIV elements, one BODY element and one HTML element. Some of these elements also have a class (starting with a dot) or an ID (starting with #).

• Based on this information, create a CSS selector and then extract relevant text from selected nodes.

Parsing dates

To parse dates (written as a text), you have two options:

• datetime.strptime

>>> import datetime
>>> datetime_str = '9.10.18 13:55:26'
>>> datetime.datetime.strptime(datetime_str, '%d.%m.%y %H:%M:%S')
datetime.datetime(2018, 10, 9, 13, 55, 26)

• dateutil package. Beware, that default setting prefer "month.day.year" format. This can be fixed with "dayfirst" flag.

>>> import dateutil.parser
>>> dateutil.parser.parse('2012-01-02 15:20:30')
datetime.datetime(2012, 1, 2, 15, 20, 30)
>>> dateutil.parser.parse('02.01.2012 15:20:30')
datetime.datetime(2012, 2, 1, 15, 20, 30)
>>> dateutil.parser.parse('02.01.2012 15:20:30', dayfirst=True)
datetime.datetime(2012, 1, 2, 15, 20, 30)

Other useful tips

• Don't forget to commit changes to your SQLite3 database (call db.commit()).
• SQL command CREATE TABLE IF NOT EXISTS can be useful at the start of your script.
• Use screen command for long running scripts. Screen allows to you keep script running, even if you close the ssh connection (or it drops due to bad wifi).
• To get out if screen use Ctrl+A and then press D.
• To get back in write screen -r. (Or screen -r -d if screen is attached somewhere else in the background).
• All packages are installed on our server. If you use your own laptop, you need to install them using pip (preferably in an virtualenv).