Sample Node subclasses and Node execution

[reelmatt]

This is my first stab at how our Node subclasses and Node execution might work. I'm still a little new to classes/inheritance in Python, so you might notice some Java-inspired patterns that could be re-written. But, I think this serves as a good foundation we can build on. This should match up with the discussions we've had, but there's definitely still some unanswered questions in this approach (custom nodes??) and room for improvement!

pyworkflow

• Added a few Node subclasses (IONode, ManipulationNode, ReadCsvNode) to provide customization to the parent Node class
• Creation/lookup of Nodes is now handled by a new node_factory.py (written as functions, not sure if a class/other implementation would be more appropriate)
  • Node basics should be fine for going to/from the session (stored in the nx.DiGraph as key: value attributes.). Now works great in conjunction with the factory as the node info can be passed directly to the factory à la node_factory(self.graph.nodes[node_id])
  • Didn't fully test pandas DataFrames, but they have to_json and read_json methods, so that should be okay.

Django views

• Modified the add_node route (to use the node_factory)
• Added an execute_node route

Testing

• Routes are still all GET-based, instead of POST. While we can skirt around the CSRF issue temporarily, I am having problems with testing via Postman/curl due to our session-based approach. When trying a route that assumes a graph exists, it complains because there is no session to speak of (is my guess at least). So I'm currently doing testing via the browser (hence the GET requests). Any session-based testing tips would be 👍
• I've attached a sample workflow.json and test.csv in vp-test-files.zip to use for testing. These should be moved to /tmp. You'll find a few comments scattered throughout in the new commits referencing file-handling. tl;dr—workflow is mostly fine, but we should take a look at files for Nodes with input/output.
• To test, you can open (/workflow/open?file=workflow.json) and then execute node (/node/{id}/execute). I added an 'Add node' test to Postman as well. (Again, hasn't succeeded in Postman, but has worked for me in the browser).

[reelmatt]

I figured out support for sessions in Postman, at least for creating the new workflow. This now allows for modification of a workflow and POSTing of Nodes. Latest commits reflect these changes and simplifies some code.

The 'retrieve nodes' route was changed to pull in information from the Node module instead of hard-coded JSON data. To enable some retrieval, I changed some information from instance -> class attributes, but not sure if this is "correct" or any public/private variable information we might need to consider.