Notes for Contributors

This page gives general information about developer workflows valid for the entire project. For more specific information about developing for the client or server part of the project, refer to:

Git Branches

  • master: This is the stable branch. Release tags are always on this branch and the latest release is always the head of this branch.

  • develop: This is the main development branch and can be ahead of the master branch. One should never directly merge and push to develop but perform a pull request on GitHub. The PR can only be merged if approved by another developer and all automatic tests pass.

  • wip_xx: Feature branches for feature number xx start with wip_xx (one can add a short name to the branch name). They always branch off develop, and as explained above, are merged to develop via GitHub pull requests.

  • release/vx.y.z: These are release branches (for version vx.y.z). They always branch off from develop. Once the release process is complete (see below), the release branch is merged to both master and develop via GitHub pull requests.

Linting and Other Checks

To ensure code quality and consistency, we perform several checks using various tools. Below is a list of the checks that should be performed:

  • Code Formatting: Use black to automatically format the code. In lomas/server/lomas_server and lomas/client/lomas_client:

    black .

  • Code Style and Static Analysis: Use flake8 to verify formatting and perform static code analysis. In lomas/server/lomas_server and lomas/client/lomas_client:

 flake8 .

  • Static Type Checking: Use mypy for static type checking. Note that both the server and the client have their own mypi.ini files to ignore specific warnings. In lomas/server and lomas/client:

 mypy .

  • Additional Static Analysis: Use pylint for further static analysis. Note that both the server and the client have their own .pylintrc files to ignore specific warnings. In lomas/server/lomas_server and lomas/client/lomas_client:

 pylint .

To streamline the process, you can use the run_linter.sh script in ``lomas`. The first time you run this script, use the following command to install dependencies:

chmod +x run_linter.sh
./run_linter.sh  --install-deps

For subsequent runs, simply execute:

./run_linter.sh

There should be no error or warning, otherwise the linting github action will fail. All configurations are in lomas/server/pyproject.toml and lomas/client/pyproject.toml.

As detailed below, we rely on GitHub workflows to automatically run these checks on pull requests, ensuring consistency and quality across all contributions.

GitHub Workflows

This project uses a number of GitHub workflows to automate various CI/CD tasks. These task can also be manually run in a local environment during development. Please refer to the workflow files in .github/workflows/ for further details. The table below gives an overview of which workflows are triggered by what events.

Workflow / Trigger

PR to develop

PR to master

Push to develop

Push to release/**

Push to master

GitHub release

Tests and Linters

Yes

Yes

No

No

No

No

Docker build and push

No

No

Yes (tag = git sha)

No

Yes (tag = git sha)

Yes (tags = latest and semver (x.y.z))

Client library push

No

No

No

No

No

Yes (must manually adjust version)

Helm charts push

No

No

No

Yes (must manually adjust version)

No

No

Documentation push

No

No

Yes (for latest)

No

No

Yes (for stable, must manually add version)

Security with CodeQL*

Yes

Yes

No

No

No

No

Of these workflows, three of them need manual intervention to adjust the version number:

  • Client library push: The ‘version’ and the ‘install_requires’ must be set in client/setup.py (‘install_requires’ should match the list of library in requirements.txt).

  • Helm chart push: The chart version (version) and app version (AppVersion) of the server and the client must be updated in server/deploy/helm/charts/lomas_server/Chart.ymland client/deploy/helm/charts/lomas_client/Chart.yaml.

  • Documentation push: If a new version is released, it must be added to the docs/versions.yaml file. For more details on the generation of the documentation, please refer to docs and the docs/build_docs.py script.

*The Security with CodeQL workflow is also triggered every Monday at 9am.

Release Workflow

The following actions must take place in this order when preparing a new release:

  1. Create a release/vx.y.z branch from develop.

  2. Fix remaining issues.

  3. Adjust versions for the client, core and server libraries (in the different setup.py), the helm charts, as well as for the documentation.

  4. Create a GitHub PR from this branch to develop AND master (make sure you are up to date with develop by rebasing on it)

  5. Once merged, manually create a release on GitHub with the tag vx.y.z.

The workflows listed in the previous section will take care of building and publishing the different items (docker images, pip packages, etc.).

Note: Helm charts are updated when there is a push on the release/vx.y.z branch. If you have a specific deployment that rely on the Chart, you can test it before finishing the release. Then, do not forget to update the chart and app versions of your specific deployment.

Adding a DP Library

It is possible to add DP libraries quite seamlessly. Let’s say the new library is named ‘NewLibrary’ Steps: 0. Add the necessary requirements in lomas/lomas_server/requirements.txt and lomas/lomas_client/requirements.txt

  1. Add the library the the DPLibraries StrEnum class in lomas/lomas_core/constants.py (DPLibraries.NEW_LIBRARY = "new_library") and add the NewLibraryQuerier option in the querier_factory (in lomas/lomas_server/dp_queries/dp_libraries/factory.py).

  2. Create a file for your querier in the folder lomas/lomas_server/dp_queries/dp_libraries/new_library.py. Inside, create a class NewLibraryQuerier that inherits from DPQuerier (lomas/lomas_server/dp_queries/dp_querier.py), your class must contain a cost method that return the cost of a query and a query method that return a result of a DP query.

  3. Add the three associated API endpoints .

  • a. Add the endpoint handlers in lomas/lomas_server/routes/routes_dp.py: /new_library_query (for queries on the real dataset), /dummy_new_library_query (for queries on the dummy dataset) and /estimate_new_library_cost (for estimating the privacy budget cost of a query).

  • b. The endpoints should have predefined pydantic BaselModel types. Aadd BaseModel classes of expected input NewLibraryModel, DummyNewLibraryModel, NewLibraryCostModel in lomas/lomas_server/utils/query_models.py and add the request case in the function model_input_to_lib().

  • c. The endpoints should have predefined default values example_new_library, example_dummy_new_library in lomas/lomas_server/utils/query_examples.py.

  1. Add tests in lomas/lomas_server/tests/test_new_library.py to test all functionnalities and options of the new library.

  2. Add the associated method in lomas-client library in lomas/client/lomas_client/client.py. In this case there should be new_library_query for queries on the private and on the dummy datasets and estimate_new_library_cost to estimate the cost of a query.

  3. Add a notebook Demo_Client_Notebook_NewLibrary.ipynb in lomas/client/notebook/ to give example of the use of the library.

External Loggers

Some libraries have ‘custom object’ parameters which are not readily serialisable. In those cases, a logger library can be made to serialise the object in the client (before sending them to the server via FastAPI) and then deserialise them in their DPQuerier class in the server.

Some examples are avalaible here:

  • opendp_logger for opendp pipelines: https://github.com/opendp/opendp-logger

  • diffprivlib_logger for diffprivlib pipelines: https://github.com/dscc-admin-ch/diffprivlib-logger

  • smartnoise_synth_logger for smartnoise_synth table transformer constraints: https://github.com/dscc-admin-ch/smartnoise-synth-logger

Do not forget to add these libraries in the requirements.txt files.

Adding a Data Connector (for private dataset in various databases)

Here is the explanation of how to add a new data connector named NewDataConnector for the example.

  1. Add the new dataset store to the NewDataConnector StrEnum class in lomas/lomas_server/constants.py.

  2. Add the NewDataConnector option in the data_connector_factory function (in lomas/lomas_server/data_connector/factory.py).

  3. Create a file for your dataset store in the folder lomas/lomas_server/data_connector/new_data_connector.py. Inside, create a class NewDataConnector that inherits from DataConnector (lomas/lomas_server/data_connector/data_connector.py), your class must contain a get_pandas_df method that return a dataframe of the dataset.

  4. Add tests in lomas/lomas_server/tests/ to test all functionnalities of the new data connector.