Our MVC_controller.py supported the following actions:
Actions
-------
c - create an object to add to the collection
r - retrieve an object from the collection and display it
u - update an object in the collection
d - delete an object from the collection
l - list all the objects in the collection
q - exit
We want to support the same functionality using HTTP, with a browser as the client, and our same shelve as the back-end datastore.
Now how exactly can we map these CRUD actions onto the HTTP methods? There are two issues we have to resolve.
Remember that the form of an HTTP request is METHOD URL
VERSION so we need to identify the URLs for each method, i.e. we
have to say GET from somewhere,
POST to somewhere etc.
Some of these are easier to figure out than others:
GET and DELETE will obviously be GET uid to retrieve the
html representation of object uid, and DELETE uid to delete
the object with UID uid.
UPDATE is not so clear. Clearly we will want to issue UPDATE
uid to update the object with UID uid, but we will have to send it
the information to use in creating the updated object which presumes an
earlier interaction. Remember that in our console-based controller we
first got the modified values by interacting with the user and
then called the update method in model.py. In an HTTP system the
initial interaction will be achieved by presenting the user with an HTML
form to fill in and then submitting this form to trigger the UPDATE.
This highlights the second issue: Some of those actions invoked an interaction between the user and the controller, but HTTP is stateless, that is each request is handled alone without knowledge of previous requests. (Now lots of approaches to web apps try to get around this by managing sessions, but that adds complexity, fragility and makes many desirable performance features of the web unavailable, e.g. caching. )
So we will need to call one form to fill in the modified values and then PUT that form to trigger the actual update.
Similarly using POST to create a new object will involve two steps, first requesting a form to use to specify the characteristics of the new object and second a POST request to actually trigger its creation. Now where do we post to to create a new object? To a special object creation URL.
Finally there is listall to consider. We will do that by GETting the server root.