Link Search Menu Expand Document

Epidata API Development Guide

Prerequisite: this guide assumes that you have read the frontend development guide.

This guide describes how to write and test code for the Epidata API. For preliminary steps, install docker and create a virtual network.

After reading this guide, you may want to visit the fluview_meta tutorial for an example of how to add a new endpoint to the API.


For working on the Epidata API, you’ll need the following two Delphi repositories:

You likely won’t need to modify the operations repo, so cloning directly from cmu-delphi is usually sufficient. However, since you are going to be modifying delphi-epidata sources, you’ll first need to fork the repository and then clone your personal fork. For more details, see the Delphi-specific discussion on forking and branching.

Here’s an example of how to setup your local workspace. Note that you will need to use your own GitHub username where indicated.

# collect everything in a directory called "repos"
mkdir repos && cd repos

# delphi python (sub)packages
mkdir delphi && cd delphi
git clone
git clone
cd ..

# go back up to the workspace root
cd ..

Your workspace should now look like this:

tree -L 3 .
└── repos
    └── delphi
        ├── delphi-epidata
        └── operations

build images

We now need images for the Epidata API web server and the epidata database. These are both based on core Delphi images which are defined in the operations repo which you cloned above. The base images are built first, followed by the derived epidata-specific images.

From the root of your workspace, all of the images can be built as follows:

docker build -t delphi_web \
  -f repos/delphi/operations/dev/docker/web/Dockerfile .

docker build -t delphi_web_epidata \
  -f repos/delphi/delphi-epidata/dev/docker/web/epidata/Dockerfile .

docker build -t delphi_database \
  -f repos/delphi/operations/dev/docker/database/Dockerfile .

docker build -t delphi_database_epidata \
  -f repos/delphi/delphi-epidata/dev/docker/database/epidata/Dockerfile .


At this point you’re ready to bring the stack online. To do that, just start containers for the epidata-specific web and database images. As an aside, the output from these commands (especially the web server) can be very helpful for debugging. For example:

# launch the database
docker run --rm -p 13306:3306 \
  --network delphi-net --name delphi_database_epidata \

# launch the web server
docker run --rm -p 10080:80 \
  --network delphi-net --name delphi_web_epidata \

unit tests

Once the server containers are running, you can run unit tests.

First, build the delphi_python image per the backend development guide. Your test sources will live in, and be executed from within, this image.

Then run the test container:

  docker run --rm --network delphi-net delphi_python \
  python3 -m undefx.py3tester.py3tester \

The final line of output should be similar to the following:

  All 48 tests passed! 68% (490/711) coverage.

manual tests

You can test your changes manually by:

  1. inserting test data into the relevant table(s)
  2. querying the API using your client of choice (curl is handy for sanity checks)

Here’s a full example based on the fluview endpoint:

  1. Populate the database (particularly the fluview table) with some fake data. For example:
  echo 'insert into fluview values \
    (0, "2020-04-07", 202021, 202020, "nat", 1, 2, 3, 4, 3.14159, 1.41421, \
    10, 11, 12, 13, 14, 15)' | \
  mysql --user=user --password=pass \
    --port 13306 --host epidata

Note that the host and port given above are “external” values, which are locally visible. You’ll need the mysql client installed locally to run the above command.

In case you don’t have the mysql client installed on your machine, and don’t want to install it, you can simply use the binary that’s bundled with the mariadb docker image, which you should already have from building the delphi_database image. In that case, use the “internal” values, which are visible to containers on the same virtual network. For example:

  echo 'insert into fluview values \
    (0, "2020-04-07", 202021, 202020, "nat", 1, 2, 3, 4, 3.14159, 1.41421, \
    10, 11, 12, 13, 14, 15)' | \
  docker run --rm -i --network delphi-net mariadb \
  mysql --user=user --password=pass \
    --port 3306 --host delphi_database_epidata epidata

Note that for these inserts, absence of command-line output is a sign of success. On the other hand, output after the insertion likely indicates failure (like, for example, attempting to insert a duplicate unique key).

  1. Query the API directly:
  curl -s \
    'http://localhost:10080/epidata/api.php?source=fluview&epiweeks=202020&regions=nat' | \
  python3 -m json.tool

The pipe to python’s built-in JSON formatter is optional, but handy. You should expect to see the following response from the server:

      "result": 1,
      "epidata": [
              "release_date": "2020-04-07",
              "region": "nat",
              "issue": 202021,
              "epiweek": 202020,
              "lag": 1,
              "num_ili": 2,
              "num_patients": 3,
              "num_providers": 4,
              "num_age_0": 10,
              "num_age_1": 11,
              "num_age_2": 12,
              "num_age_3": 13,
              "num_age_4": 14,
              "num_age_5": 15,
              "wili": 3.14159,
              "ili": 1.41421
      "message": "success"

Alternatively, you could query the API using one of the available client libraries. However, this would require you to modify the base URL within the client’s code, and there is some additional amount of boilerplate involved in calling the client and displaying the result. For these reasons, client libraries are better candidates for automated integration tests (and unit tests, in the case of the python client) than one-off manual tests.

integration tests

Writing an integration test is outside of the scope of this document. However, a number of existing integration tests exist and can be used as a good starting point for additional tests. For example, see the tests for the fluview API endpoint and the covidcast ingestion pipeline.

To run the existing tests and any new tests that you write, you must follow the backend development guide within the same workspace, so that the delphi_python image is created with any changes you have made (e.g. adding new integration tests). That image will contain the test driver and the source code of your integration tests. Then, run the tests inside a container based on that image. Note that the container of tests will need to be attached to the virtual network delphi-net in order to see and communicate with the web and database servers.

More concretely, you can run Epidata API integration tests like this:

  1. Build server images as described in the building section above.

  2. Launch the server containers as described in the test section above.

  3. Build the delphi_python image per the backend development guide. Your test sources will live in, and be executed from within, this image.

  4. Run integration tests in a container based on the delphi_python image:

  docker run --rm --network delphi-net delphi_python \
  python3 -m undefx.py3tester.py3tester --color \

You should see output similar to the following (edited for brevity):

  test_privacy_filtering (repos.delphi.delphi-epidata.integrations.test_covid_survey_hrr_daily.CovidSurveyHrrDailyTests
  Don't return rows with too small of a denominator. ... ok
  test_round_trip (repos.delphi.delphi-epidata.integrations.test_covid_survey_hrr_daily.CovidSurveyHrrDailyTests)
  Make a simple round-trip with some sample data. ... ok

  test_round_trip (repos.delphi.delphi-epidata.integrations.test_fluview.FluviewTests)
  Make a simple round-trip with some sample data. ... ok

  ✔ All 3 tests passed! [coverage unavailable]
  1. Bring down the servers, for example with the docker stop command.