CouchDB + Spatial = Awesome

Overview

• What is CouchDB?
• Why CouchDB relevant to GIS.
• How to use CouchDB (and GeoCouch).

What is CouchDB?

A NoSQL document store

• Data is stored as JSON fragments.
• Data can be collated and queried using views.
• I like to think of it as a distributed file system that actually works.
• And, it's also a webserver.

Where Does CouchDB fit?

• I think the question is more, where doesn't it?

• Suitable for:

  • Small test applications
  • Websites
  • Distributed Data Stores
  • Use it as a Distributed File System that works.

Getting Started with CouchDB

Not Installing CouchDB

• By far the easiest way to get started with CouchDB, is to not install it.
• One of the best ways to do this is to use IrisCouch: http://www.iriscouch.com/.
• If you do need to install it, then I'd recommend looking at http://refuge.io - the guys there are doing some great work making the install easier.

Interacting with CouchDB

• All interaction with CouchDB is done through a sensible, RESTful API.
• There are also many libraries for working with the API.
• My preference is for those which provide an interface as close to the actual HTTP API as possible.
• And, there is a pretty friendly web admin utility called Futon.

Simple GET

curl http://sidelab.iriscouch.com/seattle_neighbourhood/c0d5867266787565c510f5a7e20000f2

{
 "lat/lng": "47.612237, -122.290868",
 "_id": "c0d5867266787565c510f5a7e20000f2",
 "_rev": "3-3b75f0b5202e28c14de19addcc2b8d3d",
 "Website": "http://www.seattleschools.org/schools/madrona/",
 "Address": "1121 33rd Ave",
 "Latitude": "47.612237",
 "City Feature": "Alternative Schools",
 "Longitude": "-122.290868",
 "Common Name": "Madrona K-8"
}

NOTE: While this data has lat/lng data it is not in an ideal form, for GeoCouch to use.

Simple PUT

New data can be added through use of the HTTP PUT method:

curl -X PUT http://sidelab.iriscouch.com/test/test -d '{"type":"Feature", "color":"orange" ,"geometry":{"type":"Point","coordinates":[11.395,48.949444]}}'

{
    "ok":true,
    "id":"test",
    "rev":"1-2eeb1e5eee6c8e7507b671aa7d5b0654"
}

So much more

• I consider CouchDB to be quite a deep product, that is easy to get started with but a while to master.

• We will touch on other aspects of CouchDB as relevant to this talk.

CouchDB and Location Based Services

Why?

From my personal observations, location based data (with particular reference to Point of Interest data) generally contains:

• Some core common attributes (title / name / etc)
• Some Geospatial information (be that points, lines, polygons, etc)
• And then a whole pile of custom attributes.

Effectively storing these custom attributes in an RDBMS, is generally painful and ugly.

CouchDB offers an alternative

And you've got options!

Option 1: Hybrid Stack

• Involves separating* data into relevant stores:
  • Geospatial data should be extracted and placed in a traditional GIS.
  • Extended attribute information should remain in Couch only.

Hybrid Stack Pros & Cons

PROS

• Access to more complicated geospatial operations through familiar tool.

CONS

• Have to ensure mechanisms are in place to keep data in sync.

Options 2: Pure Couch

• All data remains in CouchDB (as it does in Option 1)
• GeoSpatial data is indexed within Couch, courtesy of GeoCouch.

Pure Couch Pros & Cons

PROS

• Highly self contained, with options of deployment to mobile

CONS

• Limited Geospatial operations available
• Geospatial operation speed
• Data size - see Todd Jackson's post.

Importing Data

Many Options

• Scripting and interaction with the HTTP API.

  Check out Todd Jackson's post at:

  http://boomphisto.blogspot.com.au/2011/10/serving-up-geojson-while-having-rest-on.html

• Tools such as shp2geocouch (ruby): https://github.com/maxogden/shp2geocouch (Not sure if this is still activately maintained).

• GDAL >= 1.9.0 support for CouchDB

GDAL Example

Given you lot are spatial types, I'm going to recommend using GDAL (>= 1.9.0), and in particular the ogr2ogr tool.

ogr2ogr -f couchdb "couchdb:http://yourname.iriscouch.com/parking" sourcedata.shp -nln boundaries -overwrite

Credentials can be passed in using the COUCHDB_USERPWD environment variable:

COUCHDB_USERPWD=admin:yourpass ogr2ogr -f couchdb "couchdb:http://yourname.iriscouch.com/parking" sourcedata.shp -nln boundaries -overwrite

More info: http://www.gdal.org/ogr/drv_couchdb.html

NOTE: Use the HTTP API for Updates!

• Performing full imports using the either the shp2geocouch or ogr2ogr tools would not be encouraged in a production situation, as it will impact CouchDB replication and the _changes feed.

GeoCouch

What is GeoCouch

• GeoCouch (see https://github.com/couchbase/geocouch) is a set of spatial extensions for CouchDB.

• Given CouchDB's capabilities as a webserver this opens up opportunities for much lighter weight, but still very capable geospatial web apps.

• Checkout a great talk from Volker Mische, GeoCouch's creator here: http://www.couchbase.com/slides/geocouch/CCSF

Using GeoCouch

• The simplest way to get started with GeoCouch is through using a free couch instance at http://iriscouch.com/.

• Uploading design docs into CouchDB is the hardest part.

  This generally true of CouchDB, but CouchApps can help

Using GeoCouch - Writing a CouchApp

• Some clever individuals came up with the concept of a CouchApp (see http://couchapp.org/).

• Writing a CouchApp means you are deploying both your application and data into couch, and serving it using the internal webserver.

CouchApp Uploaders

There are a number of tool available to help with the process of deploying a CouchApp into a CouchDB instance (including IrisCouch, CloudAnt, etc):

• CouchApp (Python): https://github.com/couchapp/couchapp
• CouchApp (Node.js): https://github.com/mikeal/node.couchapp.js
• Kanso (Node.js): http://kan.so/

Demo

see: https://github.com/DamonOehlman/demo-kanso-geocouch

Summary

• You should investigate CouchDB. Sooner rather than later.
• While not designed as a GIS platform, it does have a place I believe.
• GeoCouch has a way to go yet, but it's worth keeping an eye on.

Questions?