Geocoder tester

Run search queries against a geocoder that supports geocodejson spec.

Intalling

• create a python >= 3.4 virtualenv environment
• git clone https://github.com/geocoders/geocoder-tester && cd geocoder-tester
• pip install -r requirements.txt

Running

Simply:

py.test

For a global help, type:

py.test -h

Tests are split by geographical area, so you can run only a subset of all the tests, for example because your local database only contains a small area, or because you want to focus on some data.

Available subsets: germany, france, iledefrance, italy.

If you want to run only a subset of the tests, run for example

py.test -m iledefrance

What if I want to have details about the failures?

py.test --tb short

How can I stop at first failing test?

py.test -x

Can I change the api URL I'm testing against?

py.test --api-url http://photon.komoot.de/api/

If you want to test not only against another photon instance, but against a nominatim or pelias service, supply the optional --api-type parameter and specify either photon, nominatim or pelias.

py.test --api-url https://nominatim.openstreetmap.org/ --api-type nominatim

Note that support for pelias is still rudimentary.

Can I limit the number of tests to be run (even if my filter select thousands of tests) ?

py.test --max-run 100

Or, to limit the number of failures:

py.test --maxfail 10

Can I have a geojson to compare failures ?

py.test --geojson

Sometimes, the data used by the search engine is not perfect, so to run the checks without comparing diacritics and such:

py.test --loose-compare

You can compare two runs (for example to compare two branches). First, save the report from the first run:

py.test --save-report path/to/report.log

Then compare when running a new version

py.test --compare-report path/to/report.log

Note: in compare mode, only new failures will appear as "FAILED" and their traceback will be rendered; already known failures will appear as "xfail" and in yellow instead of red. If you want those known to fail tests not to be run at all (thus you'll don't know how many of them now pass), you can use the --skip-xfail command line argument.

Adding search cases

We support python, CSV and YAML format.

Before creating a new file, check that there isn't a one that can host the test you want to add.

How do I name my file? Just make it start with test_, and chose the right extension according to the format you want to use: .py, .csv or .yml.

Where do I save my file? Chose the right geographical area, and if you create a new area remember to create all levels, like france/iledefrance/paris.

Remember to check the tests already done to get inspiration.

You generally want to use YAML format if you are managing tests by hand in your text editor, CSV if you are generating test cases from a script, and python test cases if you need more control.

Python

They are normal python tests. Just note that you have two utils in base.py: search and assert_search that can do a lot for you.

CSV

One column is mandatory: query, where you store the query you make. Then you can add as many expected_xxx columns you want, according to what you want to test. For example, to test the name in the result, you will store the expected value in the column expected_name; for an osm_id it will be expected_osm_id, and so on. Note on expected_coordinate format: it should be of the form lat,lon,tolerated deviation in meters, e.g. 51.0,10.3,700.

Optional columns:

• limit: decide how many results you want to look at for finding your result (defaul: 1)
• lat, lon: if you want to add a center for the search
• comment: if you want to take control of the ouput of the test in the command line
• lang: language
• skip: add a skip message if you want a test to be always skipped (feature not supported yet for example)

YAML

The spec name is the query, then one key is mandatory: expected, which then has the subkeys you want to test against (name, housenumber…). Optional keys: limit, lang, lat and lon, skip. You can add categories to your test by using the key mark (which expects a list), that you can then run with -m yourmarker.

License

Geocoder-tester is available under a MIT license. See LICENSE.txt for more information.

Some of the test cases under geocoder_tester/world/ are derived from data with other licenses:

• OpenFlights airport data available under ODbL
• Wikidata available under CC0

Please refer to the license files in the appropriate subdirectories. When no separate license file is present, the tests are considered to be in the public domain. You may use them without any restrictions.