Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
Python wrapper for Factual's API.
Python
tree: d29e518449

Fetching latest commit…

Cannot retrieve the latest commit at this time

Failed to load latest commit information.
factual
.gitignore
LICENSE
Makefile
README.md
distribute_setup.py
setup.py

README.md

Factual Server API Wrapper

This package wraps the Factual v3 API. Its query style is SQLAlchemy-inspired and designed to make it easy to build "read" requests by chaining filter calls together.

There is also limited support for v2 API, including actions other than "read," under the "v2" subpackage.

Note that there are minor API inconsistencies between the v2 and v3 versions, both on the server side and in this wrapper. The v3 version of the wrapper is now the default.

API access

To get started, you'll need to register for Factual API OAuth credentials.

Usage

First, create a Session using your v3 API OAuth consumer key and consumer secret:

import factual
session = factual.Session(
                         consumer_key="myOAuthConsumerKey",
                         consumer_secret="myOAuthConsumerSecret"
                         )

Now, build a query using the read action on the "global" table:

query = session.read("global")

You can apply as many filters as you'd like to the query. Filters on a query are cumulative, and can be chained:

query.search("coffee")
query.filter({"locality": "Albany"}).filter({"region":"NY"})

When you're done, run() the query and retrieve the results using the records() method:

from pprint import pprint

data = query.run().records()
pprint(data)

Geographic queries

You probably want to search for places near a point. Use the within(latitude, longitude, radius_in_meters) helper method of read. within() chains and applies like any other filter, except that the last call will overwrite earlier geo filters. The underlying Factual filter API has changed between v2 and v3, but this will work for both:

query = session.read("global").within(40.7353,-73.9912,1000).search("coffee")

Pagination

Get more results using count() and page():

query.count(30).page(2) # 30 results per request, second page of results

Note that Factual's API instead uses limit and offset, so I should probably change the wrapper to match.

Categories

Factual provides categories as hierarchical strings. That is, any place marked "Food & Beverage > Bakeries" is in the "Bakeries" subcategory of "Food & Beverage."

It's possible to query for either specific subcategories or parent categories using the $bw ("begins with") filter operator. You can then search across multiple of these $bw filters by chaining them together with $or.

Because this can get pretty lengthy, the category_helpers module has a make_category_filter function. make_category_filter takes a list of category strings and combines them into a $bw/or filter. Since $bw will always include all subcategories of an supercategory listed, make_category_filter also dedupes the provided categories to generate the shortest list possible. This may mean that it will include more subcategories than you indended; but if you want to get in to $and $not $bw tangles, you're on your own.

Pass blank = True as a kwarg to make_category_filter if you also want results without categories set.

from factual import category_helpers

my_categories = [ "Food & Beverage", "Food & Beverage > Bakeries", "Shopping" ]
my_filters = category_helpers.make_category_filter(my_categories, blank=True)
# {'$or': ({'category': {'$bw': 'Food & Beverage'}}, {'category': {'$bw': 'Shopping'}}, {'category': {'$blank': True}})}

query = s.read("global").filter(my_filters)

Non-OAuth requests

It's possible to skip OAuth for the v3 API, if, for instance, there's some trouble with signing requests. Creating a Session without a consumer_secret will authenticate requests via the KEY query string parameter rather than OAuth:

non_oauth_session = factual.Session(consumer_key="myOAuthConsumerKey")

Note that Factual discourages falling back to the KEY parameter, and intends it for debugging use only, so use OAuth if possible.

Examples

from factual import *
s = Session(consumer_key="deadbeef", consumer_secret="foobar")
my_place = s.read("global").search("coffee").run().records()[0]

Building requests one piece at a time:

query = s.read("global")
query.filter({"name": "Foobar"})
if my_address != None:
    query.filter({"address": my_address})
response = query.run()
records = response.records()

Limiting categories and using the filter helper functions:

from factual.filter_helpers import *
q = s.read("global").filter(
                    or_(
                        bw_("category", "Food"), 
                        bw_("category", "Arts")
                    )
                  ).search("Foobar")
records = q.run().records()

Geo queries:

# lat, lon, radius in meters
coffee_places = s.read(places).search("coffee").within(40.7353,-73.9912,1000).run().records()

Factual v2 "Server" API

To use the v2 API, instantiate a factual.v2.Session object instead of a factual.Session object:

import factual
v2_session = factual.v2.Session(api_key="deadbeef")
...

Note that you'll need to provide a v2 API key using the api_key argument. Factual issues v2 API keys via a different process than for v3 API credentials.

In the v2 API, you can also modify a record in the Playpen:

v2_session = factual.v2.Session(api_key="deadbeef")
p = v2_session.read(USLocalPlaypen).search("coffee").count(1).run().records()[0]
p['address'] += "/Foobar"
v2_session.input(USLocalPlaypen).values(p).comment("Silly update test").run()

See also the Python documentation for session.Session and requests.Read and Factual's developer documentation.

TODO

  • Write support for v3, when available
  • Multiple search filters (search filters currently replace one another)

Changelog

  • v0.1.2 - Add async option to Request.run() to permit delaying the processing of the HTTP response. If the asynchttp module is installed, this will cause the initial call to run(async=True) to immediately return a get_response function, allowing you to defer the blocking call until the results are needed.
Something went wrong with that request. Please try again.