Get more context for task execution

When declaring a task, you can request for more context to be passed to the task function:

@app.task(..., pass_context=True)
def mytask(context: procrastinate.JobContext, ...):

This serves multiple purposes. The first one is for introspection, logs, etc. The JobContext object contains all sort of useful information about the worker that executes the job.

The other useful feature is that you can pass arbitrary context elements using App.run_worker (or App.run_worker_async) and its additional_context argument. In this case the context the task function receives will have an additional_context attribute corresponding to the elements that were passed:

def mytask(context: procrastinate.JobContext):
    http_session = context.additional_context["http_session"]
    return await http_session.get("")

async with AsyncSession() as http_session:
    await app.run_worker_async(additional_context={"http_session": http_session})

It may not be a good practice to use this additional_context object to share data from tasks to tasks. In order to keep the least surprising behavior, Procrastinate will try to keep modifications of this dictionary in one task from being visible by other tasks: tasks receive a shallow copy of this dict instead of the dict itself.

That being said, the values kept in this dict are not processed by Procrastinate. Any task mutating a value inside this dict will impact what all the concurrent and following tasks will read.

Note that if you start a worker, providing it an additional_context dict, and then modify the dict, the dict the tasks will receive will also be a shallow copy of the dict at the time the worker started running.