How to extract Wikipedia infoboxes and wikitables using Pandas

wikipedia-infoboxes-in-pandas.ipynb

wikipedia-infoboxes-in-pandas.py

# -*- coding: utf-8 -*-
# <nbformat>3.0</nbformat>

# <codecell>

from pandas.io.html import read_html
page = 'https://en.wikipedia.org/wiki/University_of_California,_Berkeley'
infoboxes = read_html(page, index_col=0, infer_types=False, attrs={"class":"infobox"})
wikitables = read_html(page, index_col=0, infer_types=False, attrs={"class":"wikitable"})

print "Extracted {num} infoboxes".format(num=len(infoboxes))
print "Extracted {num} wikitables".format(num=len(wikitables))

# <codecell>

infoboxes[0]

# <codecell>

infoboxes[1]

# <codecell>

wikitables[0]

# <codecell>

wikitables[1]

# <markdowncell>

# The `infer_types=False` argument is needed to turn off automatic type inference for Pandas <0.14, otherwise if date-like text appears in the table the data type will automatically be inferred as a date for the whole column, not just the particular entry, resulting in a table full of `NaT`s for non-date entries. In version >=0.14 the `infer_types` argument will be removed so it will no longer cause this kind of problem.

# <codecell>

malformed = read_html(page, index_col=0, attrs={"class":"infobox"})
malformed[0]

# <markdowncell>

# The `index_col=0` argument will turn the zeroth-column into a set of labels for the table rows which is what we want for infoboxes and some (but not all) wikitables.
# 
# Using `index_col` means that we can then refer directly to the the entries by their label, e.g.:

# <codecell>

infoboxes[0].xs(u'Motto').values[0]

# <markdowncell>

# Leaving out the argument the labels will instead be a numeric index and the zeroth-column will be part of the data.

# <codecell>

no_lefthand_labels = read_html(page, infer_types=False, attrs={"class":"infobox"})
no_lefthand_labels[0]