REST-ful microservice framework for scientific computing

rest, server, scientific, computing, microservice
pip install remisc==0.0.3



Another framework!? What's the point? Aren't there a thousand-and-one web frameworks out there already? Don't some of them already support REST-oriented operations? (Spoiler: Sort of...) Who cares about scientific computing anyways?

Here's the problem... Who needs a full-up MVC framework if all you're doing is passing system model data back and forth, or hosting a simulation service in a cloud environment? You don't. You need something model-centric, which doesn't assume clients are using a web browser. You need something that's going to provide the supporting infrastructure, like WSGI mapping and automated operation & model documentation, with the least possible overhead of work and processing power. Focus on the unique part of your work--defining operations and the models they use--and let remisc take care of everything else.


A new microservice is defined by subclassing remisc.service.Service. Methods of this class can be decorated by @remisc.service.isop. These operations take two inputs, the urlparse object of the request and a dictionary of any arguments parsed from the URL's query string segment. They should return a textual response. A request will be mapped to an operation method by matching the method name to the top-level directory in the request URL (i.e., if a service is hosted on, the request to will look for the method test of the host Service-derived class.

The base Service class includes an app() method that defines a WSGI-compliant application interface used to parse request arguments, map them to the appropriate method, and wrap the response generation process to return content.


Frequently, users must provide--or expect to receive--data formatted in accordance with specific models. Such data exchange formats can be defined by any class co-located with a Service-derived class, so long as they are decorated by @remisc.dxm.isdxm.


Several default operations are built into the base Service class:


Any request without a top-level directory name will be routed to the _root method. As with any other method, this operation can be overridden by the user.


This built-in operation returns an empty response. This is one way to hide error handling--if you're really into that sort of thing.


This is one of the most powerful behaviors of the remisc framework. The _help operation returns documentation of all operations and data exchange models for the current Service-derived object. Operations are determined by evaluating all instance methods for the @isdxm decorator. Data exchange models are determined by evaluating all classes co-located in the same module as the Service-derived class for the @isdxm decorator. The response maps a JSON object to the docstrings for each operation and data exchange model, so users can determine what operations the service provides and how it may utilize them.


By default, the remisc package includs a server module that utilizes the core module wsgiref's simple_server model to host a Service. All Service-derived classes include a WSGI-compliant app() method invoked by the server host process. For production runs, it is strongly advised to invoke that methods from a production-level WSGI server instead.

Getting Started

In Action

When the remisc.server.main method is invoked directly, the reference WSGI server in Python's core module wsgiref is used to host an instance of a given remisc.service.Service class. (If no class is provided, the base class is used instead.) This can be done procedurally:

>>> from remisc import server
>>> server.main()
Serving "remisc.service.Service" with wsgiref.simple_server @

If you use your web browser to load, you will see the default operation implemented by the remisc.service.Service._root() method ("Here is the base."). You will also see, in the Python environment where you launched the server, how that request was mapped to an operation by the WSGI application interface implemented by

"" => "_root" - - [{date} {time}] "GET / HTTP/1.1" 200 {response time}
"" => "_null" - - [{date} {time}] "GET /favicon.ico HTTP/1.1" 200 {response time}

Note that most browsers implicitly request favicon.ico with each page request; this is one useful application of the _null method, which is hard-coded to reply to such requests with an empty response.

You can view the base responses directly by browsing to the following URLs while your test server is still running. Once you have seen the responses, press CTRL-C in the Python environment to stop the server.

Your Own Service

Subclassing remisc.service.Service will let you define your own operations and override those already implemented. Make sure you decorate each method that implements an operation with @remisc.service.isop. Such methods should take two arguments: a urlparse object capturing the original request, and an args dictionary constructed from the URL query string. The operation should return the textual content of the response.

For example, let's implement a simple service Joker that implements a joke operation:

>>> from remisc import service
>>> class Joker(service.Service):
>>>     @service.isop
>>>     def joke(self, urlobj, args):
>>>         return 'Why did the spam cross the road?\n\nTo evade the dead parrot!'

We can host this service by passing the class to the remisc.server.main function as the Svc parameter:

>>> server.main(Svc=Joker)

Now, try browsing to For a bonus, note that your new operation has automatically been added to the response at!