WIP: New parser

[AtkinsSJ]

Current status:

• Output is overly nested and verbose
• $discard results are not discarded

Questions:

• How do we discard results? Parser.result() seems like a good place, but every combinator receives the return value. Do all of them have to handle discarding?
• Wrapping things in a.discard() is ugly, is there a nicer way?
• How to take the verbose output and make it clean? Probably a case of modifying each combinator to consider what its contents are.
• How do we then take the parsed tree and convert its contents into the objects we want? Define a function for each symbol that modifies its output?

[KernelDeimos]

The FOAM framework's combinator parser uses a concept called "actions"; an action is a function that's named after a symbol in the grammar, and it transforms the result of that symbol (the current return value of symbol(<symbol's name>)) before it gets returned.

In the face of the issues you're describing, this approach seems to solve everything. There's one little thing about it I didn't like though; these transformations coupled with the grammar, making it difficult for example to choose between outputs such as: AST, high-level AST ("abstract logic tree"), CST. If we can implement these in a way where actions are de-coupled from the grammar I think that might be really flexible.

Regarding $discard, I there are two options: Sequence, Repeat, etc all need to check for it; vs parsers always return arrays (and an abstract layer removes items with $discard)

[KernelDeimos]

Merging now so it doesn't get lost - merging into monorepo-prep-2