mattmakesmaps

Configure PostGIS with Travis CI

Thanks to a great presentation by foundatron on the continuous integration framework, Travis CI, I've finally been motivated to push for higher test coverage in my python projects.

Some of these projects require connectivity to a PostGIS database for testing, and with Travis CI that process is extremely straightforward. This post includes notes on configuring a Travis CI environment with PostgreSQL and PostGIS, then loading that empty database with data.

PostgreSQL/Postgis Versions #

NOTE: These versions are current as of January 2014.

Configuration #

NOTE: For background reading on Travis CI, check out the Getting Started and Build Configuration docs.

As stated in the database docs, a Travis CI environment comes with PostgreSQL installed and ready-to-go. This also includes supporting utilities such as: psql,pg_dump, and pg_restore. Since we're using PostgreSQL 9.1+ and PostGIS 2.0+, PostGIS can be installed using the CREATE EXTENSION syntax.

The .travis.yml configuration file should now be modified to execute psql calls in the before_script block to:

  1. Create the database
  2. Create the PostGIS/PostGIS Topology Extensions
  3. Populate the database with test data. (See Below for Instructions)

A reference example .travis.yml looks like this:

language: python
python:
  - "2.7"

# command to install dependencies
install:
  - "pip install -r requirements.txt --use-mirrors"

# database creation
before_script:
  - psql -c 'create database travis_postgis;' -U postgres
  - psql -c 'CREATE EXTENSION postgis;' -U postgres -d travis_postgis
  - psql -c 'CREATE EXTENSION postgis_topology;' -U postgres -d travis_postgis
  - psql -f test_data.sql -U postgres -d travis_postgis

# command to run tests
script: nosetests --verbose

Creating Test Data #

The pg_dump utility can be used to export the contents of a PostGIS database into a sql file, suitable for loading into a newly created database by the Travis worker. Key to this is to have testing data that you want to export inside of a separate schema (e.g. not in public). With pg_dump we can specify an individual schema of data tables to export, otherwise we'd be stuck with exporting potentially incompatible PostGIS function signitures in addition to our data. Boundless has a good article on backup strategies for PostGIS.

From our source PostGIS database, export tables from our schema of interest (test_data) into test_data.sql

$ pg_dump --schema=test_data travis_postgis > test_data.sql

As seen in the section above, the resulting sql file, test_data.sql is then executed as part of the Travis before_script block, populating the testing database with geometries.

NOTE: I've used the same name travis_postgis for both my local database as well as the testing database used by Travis CI. This isn't required.

Wrap Up #

Thanks to the combination of PostgreSQL's easy to use CREATE EXTENSION syntax along with Travis CI's flexible build configurations, setting up a custom PostGIS database is fairly simple.

An example repository with a .travis.yml file and an example unit test can be found here: https://github.com/mattmakesmaps/travis-postgis-test

This repository's Travis CI build log can be found here: https://travis-ci.org/mattmakesmaps/travis-postgis-test