Feed Seeker

It slant rhymes with "heat seeker"

A library for finding atom, rss, rdf, and xml feeds from web pages. Produced at the mediacloud project. An incremental improvement over feedfinder2, which was itself based on feedfinder, written by Mark Pilgrim, and maintained by Aaron Swartz until his untimely death.

Quickstart

By default, the library uses requests to grab html and inspect it and find the most likely feed url:

from feed_seeker import find_feed_url

>>> find_feed_url('https://github.com/ColCarroll/feed_seeker')
'https://github.com/ColCarroll/feed_seeker/commits/master.atom'

To do a more thorough search, use generate_feed_urls, which returns more likely candidates first.

from feed_seeker import generate_feed_urls

>>> for url in generate_feed_urls('https://xkcd.com'):
...     print(url)
...
https://xkcd.com/atom.xml
https://xkcd.com/rss.xml

For the most thorough search, add a spider argument to do depth-first spidering of urls on the same hostname. Note the below call takes nearly four minutes, compared to 0.5 seconds for find_feed_url.

>>> for url in generate_feed_urls('https://github.com/ColCarroll/feed_seeker', spider=1):
...     print(url)
...
https://github.com/ColCarroll/feed_seeker/commits/master.atom
https://github.com/ColCarroll/feed_seeker/commits/a8f7b86eac2cedd9209ac5d2ddcceb293d2404c9.atom
https://github.com/ColCarroll/feed_seeker/commits/3b5245b46a10fb3647a1f08b8e584b471683fbbd.atom
https://github.com/ColCarroll/feed_seeker/commits/659311b8853c4c4a67e3b4bc67a78461d825a064.atom
https://github.com/ColCarroll/feed_seeker/commits/3e93490cb91f7652325c2fe41ef29a5be4558d6a.atom
https://github.com/index.atom
https://github.com/articles.atom
https://github.com/dfm/feedfinder2/commits/master.atom
https://github.com/ColCarroll.atom
https://github.com/blog.atom
https://github.com/blog/all.atom
https://github.com/blog/broadcasts.atom

Installation

The library is not yet available on PyPI, so installation is via github only for now:

pip install git+https://github.com/ColCarroll/feed_seeker

Differences with feedfinder2

The biggest difference is that all functions are implemented as generators, and are evaluated lazily. Candidate feed links are actually accessed and inspected to determine whether or not they are a feed, which can be quite time consuming. We expose a function to find the most likely feed link, and another to lazily generate links in rough order from most prominent to least.

There are also a few more heuristics based on our experience at mediacloud.