An Introduction to Pretenders
The pretenders project creates flexible fakes for external network services. These external services can be faked by setting pre-canned answers, defining expectations, or just querying them on received data after [test] execution.
They are loosely designed along the pattern of Record/Replay/Verify.
The server side of Pretenders is written in Python3:
- Because I firmly believe all new projects should use Python3, unless there are very compelling reasons against it.
- Because since these will be run as standalone servers, compatibility is not such an issue
In the cases we implement a client, we will be making this runnable in Python 3.x and Python 2.6/2.7.
The initial service to be mocked by pretenders is HTTP. Future servers we are considering to support include SMTP and AMQP. These they represent the vast majority of the services that the software we write interacts with.
Some example use cases include:
- Mocking external HTTP-based APIs at the wire level in module integration tests
- Facilitating the manual testing and debugging of front-end Ajax code where the back-end has not been developed yet
- Writing fully automated tests for any system that sends email, and where we want to easily verify outgoing mails
Pretenders will provide a unified RESTful API to verify the interactions of our code with external services, regardless of the service protocol.
pretenders.http - HTTP server
Typical usage is to mock RESTful/SOAP APIs of external services. This will normally require pre-programming the service with responses, and enquiring the service later about received requests. This can be done with specialised assertions / matchers, or by setting expectations and verifying fulfilment of such expectations.
The pre-programming step may be done by using specific proprietary HTTP headers, or by using an alternative HTTP port to the one used for mocking.
Implementation is based on the bottle web microframework.
One of our goals will be that the wire protocol is simple enough that you do not need any specialised client. That said, we will be providing client libraries (at least one in Python) to simplify interacting with the server, and to provide a comfortable API to use in tests.