Improve comments in core.js and util/codegen.js

[mikolalysenko]

That code is not adequately documented and currently very confusing. Sorry.

[vorg]

I would love to see that happening. At the moment i can't say I understand how regl works at I think that's crucial for long term commitment and bigger involvement in the library. Eg. I would love to help with https://github.com/regl-project/regl/issues/312 but have no idea where to start. Commenting on each line of code doesn't make sense but a general architecture and the data flow would help a lot.

[crobi]

• What would help me understand regl most is a short paragraph summarizing how the regl core works, especially the code generation part.
• What was most unclear to me from the readme is the statement "regl uses dynamic code generation and partial evaluation to remove almost all overhead". What overhead does this remove? Does it reduce the number of draw calls (merge VBOs, use instanced rendering)? Does it reduce the number of WebGL state changes (after statically proving that some state doesn't change between regl commands)? Does it just avoid repeatedly analyzing the properties of the argument of the regl() function? To me, this partial evaluation sounds like to most interesting contribution of this project, adding some details about that in the readme would be nice.
• I tried to understand the regl core by debugging the examples, but this turned out to be very difficult:
  • Examples at https://regl-project.github.io/regl/www/gallery.html use inline minified code that can't be unminified by Chrome or Firefox, so you can't set useful breakpoints.
  • Examples at http://regl.party/examples run in an iframe that loads its code from a blob, breakpoints don't seem to work here at all.
  • Debugging generated code is difficult in general.

[vorg]

Does it just avoid repeatedly analyzing the properties of the argument of the regl() function?

That's my current understanding and unfortunately my current big project require more and more of reduce the number of WebGL state changes, which seems to be on the user side? cc: @mikolalysenko