This repository contains code and experiments relating to parameterised configuration and in-browser customisation of Opendesk products.

You can view the published demos at https://opendesk-on-demand.firebaseapp.com/

Clone the repo, cd into the folder, run e.g.: python -m SimpleHTTPServer and open your browser at localhost:8000/demo.html. Play with the controls to see the model adapt. When ready, save and check the console output.

Opendesk products are described using the winnow data format. We use winnow:

1. to describe the "world of possibilities" of a product family and its potential configuration options
2. to describe the specific choices and individual product configurations that customers and makers choose and manufacture (as per this specification pipeline)

In-browser customisation must therefore:

• present configuration options that are based on, or at least compatible with, a product's winnow options
• generate choice documents in the winnow format

The way we're proposing to do this is as follows:

1. ensure that any customisable parameters are added to / present in the product options
2. publish dynamic product configurations (see example) that:
   • have a 3D model exported as a .obj file that:
     • groups objects by layers #
     • can be loaded in browser by three.js
   • have a config.json file (see example) that:
     • exposes a subset of the product options as parameters
     • specifies the parameter aware transformation functions to apply to object geometry values by layers (i.e.: objects on different layers will have different transformation functions applied)
3. write a cross-language compiler that:
   • in Python:
     • parses the .obj and config.json#layers data
     • transforms it into a denormalised abstract-syntax-tree where each object node carries a copy of the transformations that should be applied to it (derived from the layers that the object is on)
   • in Javascript:
     • generates new .obj code by walking the AST and calculating actual object geometry values for a given configuration by applying the chosen parameter values to the transformation functions carried by any dynamic nodes
     • render the revised .obj string using THREE.OBJLoader.parse
4. control this using a controller UI built from the config.json#parameters

N.b.: in future, we can also consider writing the transformation functions in Python and transpiling to Javascript from Python, in order to support an an alternative 'no-webgl' Python server-side code-generation step that uses Blender to render images.