General terminology and practices¶
These are common conventions we keep on building as the project gains new contributors and grows in complexity.
As we gather more concrete examples, this page will have one section for each category to demonstrate a before and after.
|We use Numpy convention with markdown to help generate more readable API references. For public APIs, we always include at least one Example to ease everyone's experience when using an IDE.
|We use black and Ruff to enforce beyond good practices PEP8. We use type annotations and enforce static type checking at CI (mypy).
|Core utilities always accept
service as a constructor parameter, can work in isolation, and are also available in other languages implementation.
|Utilities are not as strict as core and focus on community needs: development productivity, industry leading practices, etc. Both core and general utilities follow our Tenets.
|Specific exceptions live within utilities themselves and use
Error suffix e.g.
|We follow conventional commits. We do not enforce conventional commits on contributors to lower the entry bar. Instead, we enforce a conventional PR title so our label automation and changelog are generated correctly.
|API reference docs are generated from docstrings which should have Examples section to allow developers to have what they need within their own IDE. Documentation website covers the wider usage, tips, and strive to be concise.
|We treat it like a product. We sub-divide content aimed at getting started (80% of customers) vs advanced usage (20%). We also ensure customers know how to unit test their code when using our features.
We group tests in different categories
|When to write
|Verify the smallest possible unit works.
|Networking access is prohibited. Prefer Functional tests given our complexity.
|Lightning fast (nsec to ms)
|Guarantee functionality works as expected. It's a subset of integration test covering multiple units.
|No external dependency. Prefer Fake implementations (in-memory) over Mocks and Stubs.
|Fast (ms to few seconds at worst)
|Gain confidence that a Lambda function with our code operates as expected.
|It simulates how customers configure, deploy, and run their Lambda function - Event Source configuration, IAM permissions, etc.
|Ensure critical operations won't increase latency and costs to customers.
|CI arbitrary hardware can make it flaky. We'll resume writing perf test after we revamp our functional tests with internal utilities.
|Fast to moderate (a few seconds to a few minutes)