Noflare

Noflare is a a drop-in replacement for selected Cloudflare Workers features that makes it easier to test Cloudflare Worker applications.

Noflare does not use a local wrangler/miniflare session and does not depend on the Cloudflare runtime.

Motivation

The library was created to make Cloudflare Workflows easier to test. In particular:

• Idempotency
• Pause/Resume of workflows (not implemented in wrangler/miniflare)

Workflows

Cloudflare Workflows embody critical business logic that is important to test.

It's particularly important to test that all the steps are idempotent.

It's also important to test that workflows can run concurrently.

Workflows typically communicate with external services, and mocking those is non-trivial when the tests run in one process and the workflow in another.

We wanted a way to test Workflows where the tests and the workflow run in the same process. This makes it easier to stub external services and verify how the workflow interacts with them.

Another reason we wanted to run them in the same process is that we want a sub-second feedback cycle.

Noflare achieves this by providing a mechanism to inject stub adapters for external services.

Just write your workflows as normal with the following changes:

1. Replace the import of cloudflare:workers with @oselvar/noflare
2. Replace the use of this.env with this.adapters - see
3. Use createCloudflareWorkflow to convert your Noflare Workflow to a Cloudflare Workflow

See CalculateCubeWorkflow.ts for an example.

Try it out interactively

# Terminal 1
npm start

# Terminal 2
curl http://localhost:9875?value=3

## Replace id with the response from the first request
curl http://localhost:9875?instanceId=id

Now, let's test it:

npx vitest

Tips

Noflare might highlight idempotency bugs in your workflows. Here are some tips to help make them idempotent:

• Avoid nested if statements and loops around workflow steps.
• Visualize your workflows with @oselvar/c4. (Coming soon).
• If you use D1 (Sqlite), UNIQUE constraints might help you avoid inserting duplicate data.