An attempt at a JSONP endpoint for crime data
Switch branches/tags
Nothing to show

Yet another Chicago Crime API

The main purpose of building this endpoint was to power this site which needed a JSON endpoint that could handle geospatial queries, a capability which does not exist in any of the other web APIs that are available. As of right now, I’m actually hoping to get someone else to take up this slack so that I don’t actually have to maintain this anymore.

What’s inside?

The backend contains all of the crime data between January 1, 2001 and a week ago for Chicago. The stuff you get back looks kinda looks like this.

So, it basically echos the query you sent back along with some other meta and then a list of results (in this case there’s only one match).

An example of how you might put all this stuff together is over here.

How this sucker works

Right now it only responds with JSON with a rather permissive CORS header. Whether or not you actually use it in a client-side app after that is entirely up to you. There is also an endpoint at /api/print/ that can generate a PDF of your query (including overlays of beat boundaries and overlays of any arbitrary polygons you send with your query). There is also an endpoint which generates an Excel file of the results of your query at /api/report/.

The only required field is dataset_name and for the time being it should always be chicago_crimes_all.


By default, you’re limited to 2000 records. If you’d like to discuss this, I’d encourage opening an issue.

Constructing a query

If you’re familiar with Tastypie the way you interact with this API is heavily influenced by that. The basic concept is to pass in the name of the field that you’d like to query, followed either by a filter (separated from the field name by two underscores) or by the value that you’re hoping to find in the database. In practice, that looks like this:[field_name]__[filter]=[value]

So, if you wanted to find all crimes reported between May 23, 2012 and June 25, 2012 it would look like this:

If you just want to construct a query without a filter, just leave that part out. So, if you wanted to get crimes reported on May 23, 2012 and on June 25, 2012 you’d do this:

Although that would only return reports that were made exactly at midnight on those days (since the query is performed on both the date and time).

Queryable fields

I’m basically allowing queries on any fields that are present in the dataset and that make sense to perform queries on. If you’re familiar with the dataset where all this crime data originates, you should find these fields rather familiar.

year                    # a 4-digit year 
domestic                # Boolean telling whether or not the crime was in a domestic setting
case_number             # Chicago Police Department case number
id                      # Primary key carried over from Socrata 
primary_type            # Primary crime description
district                # Police district
arrest                  # Boolean telling whether or not an arrest was made
location                # GeoJSON Point of the location of the reported crime
community_area          # Chicago Community Area where the crime was reported
description             # Secondary description of the crime
beat                    # Police beat
obs_date                # Date the crime was reported
ward                    # Ward where the crime was reported
iucr                    # Illinois Uniform Crime Reporting (IUCR) Codes
location_description    # Location description
updated_on              # When the report was most recently updated
fbi_code                # FBI crime code
block                   # Street Block where the crime was reported
type                    # Index Crime type

A few things to note:

  • You can find out more about what I’m doing with the type field here. I’m basically going with how the Tribune handles that so that there is less noise in general that you get with the responses.
  • Learn more about IUCR codes and what they mean here.

Filtering queries

The filters that can be passed in along with the fields are these:

lt                  # Less than. Used for dates.
lte                 # Less than or equal to. Also for dates    
gt                  # Greater than. Yup, dates again
gte                 # Greater than or equal to. For dates.
near                # Used to find reports near a given point. See ‘Location queries’ below.
geoWithin           # Used to find reports within a given area. See ‘Location queries’ below.
in                  # Finds matches within an array of values
ne                  # Finds matches not equal to a given value
nin                 # Finds matches not within a given array of values

You can find more about the basic query operators here.

Location queries

Finally, the good stuff. Any location based query expects ‘stringified’ GeoJSON as the lookup parameter. The excellent JSON.js library has a pretty excellent tool for taking arbitrary JSON objects and turning them into strings (which is what I use). Some more modern browsers have an implementation of this already so, depending on your audience, you may be fine.

Depending on your geometry type, you’ll either use the near filter or the geoWithin filter to find what your after. If you’re looking for crime reports near a point, you’ll construct a GeoJSON Point object, stringify it and use a near filter to return what you’re after. If you’re looking for all reports within a given polygon, you’ll construct a GeoJSON Polygon, stringify it and use the geoWithin filter. Examples of what that might look like:

A GeoJSON Polygon...

    "coordinates": [
    "type": "Polygon"

...gets stringified and appended as a query parameter:


Looks a bit insane, but it works.