Simplify Node Http Response Logic

Currently a http response exists in two forms:

1. context.res = bodyValue; // res is interpreted as 200 with body bodyValue
2. context.res = { body: bodyValue, headers: ..., status: ... } // res is fully specified object

Removing option 1 is something we ought to consider, as it has some negatives:

• We are forced to guess whether the returned object is a response based on existence of ‘body’ property, which in a couple cases is incorrect:
  • bodyValue has a body property (workaround is nested body inside body)
  • a response object is returned with only headers and status
• The ‘magic’ code of 1 doesn’t save lines of code
  • context.res = bodyValue; context.done(); vs context.res.send(bodyValue);
• We would decrease the matrix of things we need to test (currently, different behavior in content type handling between 1 and 2)
• Simplified documentation and examples with a single recommended way of interacting with http in node

#978 #814

[mamaso]

@stefan-akelius At present all function triggers are routed through a C# process. We use edgejs for interop between node and C#. The requests are served in C# so unfortunately we don’t have access to a node generated IncomingMessage or ServerResponse (I’d love to support those types natively as well).

[johnnliu]

figured this one out. 1.0 don’t like raw body content. can’t validate webhook that wanted text/plain response. argh.