developer_guide.md

title	date	draft	weight	description
Developer guide	2021-11-18	false	30	Setting up for KCIDB development

Hacking

If you want to hack on the source code, install the package in the editable mode with the -e/--editable option, and with "dev" extra included. E.g.:

pip3 install --user --editable '.[dev]'

The latter installs kcidb executables which use the modules from the source directory, and changes to them will be reflected immediately without the need to reinstall. It also installs extra development tools, such as flake8 and pylint.

Then make sure your PATH includes the ~/.local/bin directory, e.g. with:

export PATH="$PATH":~/.local/bin

Pre-commit hooks

Git hooks are scripts that can be triggered automatically by Git when certain events occur, such as committing changes to a repository. They are typically used to enforce coding standards, perform automated tests, or perform other checks on code changes before they are committed.

There are two main components in the setup 1) .pre-commit script and 2) .github-workflow-script. The .github-workflow-script extracts the steps directly from the GitHub workflow, while a .pre-commit script will run the extracted commands.

To leverage the setup, symlink the .pre-commit script into the git pre-commit hook (.git/hooks/pre-commit). We can do that using a simple command in the terminal. Make sure you are in the root directory of the project.

ln -s ../../.pre-commit .git/hooks/pre-commit

Now, every time you run git commit, the pre-commit hook script will be executed before the commit is created. If any of the pre-commit checks fail, the commit will be aborted and you will see an error message explaining the issue.

Testing subscriptions

While developing subscription modules you will likely need to test which notifications they generate for which data. You could use the kcidb-ingest tool for that.

It takes I/O JSON data on its standard input, loads it into the database specified with the -d/--database option, and then invokes subscriptions for every modified object, outputting any generated email notifications on standard output, separated by the null character \0.

The default database is an in-memory sqlite database, which is initially empty, and is discarded after completion, but you could specify any other database, for example containing pre-loaded data to simulate updates (e.g. -d json:existing_data.json). See the output of kcidb-ingest --database-help for documentation on database specification strings.

Here's a minimal example triggering the special test subscription:

kcidb-ingest <<END
{
    "version": {"major": 4},
    "builds": [
        {
            "id": "test:1",
            "origin": "test",
            "checkout_id": "test:1",
            "valid": false
        }
    ]
}
END

The above would output a single notification email.

Guidelines

Commits

When posting PRs, please keep one logical change per commit, and do not have more than one commit per a logical change. This makes it easier to review, think about, and manipulate changes. This way if you want to revert a change, you only need to revert a single commit. Same goes for merges across branches.

For example, if you have to do a similar change across multiple files, make it a single commit. OTOH, if you're doing different changes on the same file, or even the same line, make them separate commits.

Before asking for a review, please make sure the commits are in order, and don't keep amendments made on top of the previous review as separate commits - squash them.

Please use imperative mood when writing commit messages, i.e. as if giving a command for something to change. Keep subjects short and to the point, expand in the body of the commit message. This makes it easier to read, understand, and manipulate commit history. E.g. write "Add support for querying incidents", or even just "Support querying incidents", but not "Supported querying incidents", and not "Code to support querying incidents".

Use "tags" in commit subjects to make them shorter and to indicate the scope of the change. E.g. instead of writing "Add documentation on commit requirements", write "doc: Describe commit requirements". Look at commit history of particular areas you're changing for reference on which tags to use, so they're consistent.

I/O data validation

When loading data into a database, the data should be "directly compatible" with the database's I/O schema. "Directly compatible" (awkward term, I know) with a schema means that the data adheres to a schema with the same major version number, and the same, or lower minor number. That is, it can be interpreted by the receiver without "upgrading". The kcidb.io.<SCHEMA_VERSION>.is_compatible_directly() function checks for that.

When fetching I/O data from the database, it should match the database's I/O schema exactly. It should not be blindly upgraded to the currently-supported latest schema version by the database client or the drivers, because we might want to load it up into another database, which is also using an older schema, and you cannot "downgrade" I/O data, only "upgrade".

When submitting data through the client (kcidb.Client), or via a message queue (kcidb.mq.IOPublisher, which is used internally by the client), it should be valid according to the current or a previous KCIDB schema.

However, it's up to the receiving database on the other end of that message queue to reject the data, if it's newer than the database's schema, and at the moment we have no mechanism to report that situation. The submitters must be made explicitly aware of which versions they can submit, for now.

The ORM and OO modules always deal with the latest I/O schema for simplicity.