serpextract

serpextract provides easy extraction of keywords from search engine results pages (SERPs).

This module is possible in large part to the very hard work of the Piwik team. Specifically, we make extensive use of their list of search engines.

Installation

Latest release on PyPI:

$ pip install serpextract

Or the latest development version (not recommended):

$ pip install -e git://github.com/Parsely/serpextract.git#egg=serpextract

Usage

Command Line

Command-line usage, returns the engine name and keyword components separated by a comma and enclosed in quotes:

$ serpextract "http://www.google.ca/url?sa=t&rct=j&q=ars%20technica"
"Google","ars technica"

You can also print out a list of all the SearchEngineParsers currently available in your local cache via:

$ serpextract -l

Python

from serpextract import get_parser, extract, is_serp, get_all_query_params

non_serp_url = 'http://arstechnica.com/'
serp_url = ('http://www.google.ca/url?sa=t&rct=j&q=ars%20technica&source=web&cd=1&ved=0CCsQFjAA'
            '&url=http%3A%2F%2Farstechnica.com%2F&ei=pf7RUYvhO4LdyAHf9oGAAw&usg=AFQjCNHA7qjcMXh'
            'j-UX9EqSy26wZNlL9LQ&bvm=bv.48572450,d.aWc')

get_all_query_params()
# ['key', 'text', 'search_for', 'searchTerm', 'qrs', 'keyword', ...]

is_serp(serp_url)
# True
is_serp(non_serp_url)
# False

get_parser(serp_url)
# SearchEngineParser(engine_name='Google', keyword_extractor=['q'], link_macro='search?q={k}', charsets=['utf-8'])
get_parser(non_serp_url)
# None

extract(serp_url)
# ExtractResult(engine_name='Google', keyword=u'ars technica', parser=SearchEngineParser(...))
extract(non_serp_url)
# None

Naive Detection

The list of search engine parsers that Piwik and therefore serpextract uses is far from exhaustive. If you want serpextract to attempt to guess if a given referring URL is a SERP, you can specify use_naive_method=True to serpextract.is_serp or serpextract.extract. By default, the naive method is disabled.

Naive search engine detection tries to find an instance of r'\.?search\.' in the netloc of a URL. If found, serpextract will then try to find a keyword in the query portion of the URL by looking for the following params in order:

_naive_params = ('q', 'query', 'k', 'keyword', 'term',)

If one of these are found, a keyword is extracted and an ExtractResult is constructed as:

ExtractResult(domain, keyword, None)  # No parser, but engine name and keyword

# Not a recognized search engine by serpextract
serp_url = 'http://search.piccshare.com/search.php?cat=web&channel=main&hl=en&q=test'

is_serp(serp_url)
# False

extract(serp_url)
# None

is_serp(serp_url, use_naive_method=True)
# True

extract(serp_url, use_naive_method=True)
# ExtractResult(engine_name=u'piccshare', keyword=u'test', parser=None)

Custom Parsers

In the event that you have a custom search engine that you'd like to track which is not currently supported by Piwik/serpextract, you can create your own instance of serpextract.SearchEngineParser and either pass it explicitly to either serpextract.is_serp or serpextract.extract or add it to the internal list of parsers.

# Create a parser for PiccShare
from serpextract import SearchEngineParser, is_serp, extract

my_parser = SearchEngineParser(u'PiccShare',          # Engine name
                               u'q',                  # Keyword extractor
                               u'/search.php?q={k}',  # Link macro
                               u'utf-8')              # Charset
serp_url = 'http://search.piccshare.com/search.php?cat=web&channel=main&hl=en&q=test'

is_serp(serp_url)
# False

extract(serp_url)
# None

is_serp(serp_url, parser=my_parser)
# True

extract(serp_url, parser=my_parser)
# ExtractResult(engine_name=u'PiccShare', keyword=u'test', parser=SearchEngineParser(engine_name=u'PiccShare', keyword_extractor=[u'q'], link_macro=u'/search.php?q={k}', charsets=[u'utf-8']))

You can also permanently add a custom parser to the internal list of parsers that serpextract maintains so that you no longer have to explicitly pass a parser object to serpextract.is_serp or serpextract.extract.

from serpextract import SearchEngineParser, add_custom_parser, is_serp, extract

my_parser = SearchEngineParser(u'PiccShare',          # Engine name
                               u'q',                  # Keyword extractor
                               u'/search.php?q={k}',  # Link macro
                               u'utf-8')              # Charset
add_custom_parser(u'search.piccshare.com', my_parser)

serp_url = 'http://search.piccshare.com/search.php?cat=web&channel=main&hl=en&q=test'
is_serp(serp_url)
# True

extract(serp_url)
# ExtractResult(engine_name=u'PiccShare', keyword=u'test', parser=SearchEngineParser(engine_name=u'PiccShare', keyword_extractor=[u'q'], link_macro=u'/search.php?q={k}', charsets=[u'utf-8']))

Tests

There are some basic tests for popular search engines, but more are required:

$ pip install -r requirements.txt
$ nosetests

Caching

Internally, this module caches an OrderedDict representation of Piwik's list of search engines which is stored in serpextract/search_engines.pickle. This isn't intended to change that often and so this module ships with a cached version.