ROADMAP.md

ROADMAP

This is a roadmap to an MVP that aims to prove the viability of this idea.

Step 1. - Opcodes & AST

• Define & Implement a minimal set of opcodes

  • All Int,Float,Bool & String operations
  • All Array, Hash and Ref operations
  • Functions & Closures
  • Skip (for now)
    • OOP
    • Glob stuff
    • I/O & Unix stuff
  • edge case behavior can be punted
    • some stuff we want to deprecate
    • other stuff may need support

• Define & Implement the ASTs needed to model these opcodes

  • we probably need to make sure the AST captures all relevant opcode data
    • so that the opcodes do not need to be the starting point
      • enabling Guacamole ASTs, etc.

Step 2. - Stabilization & Serialization

• Once we have a full set of Opcodes & AST Nodes

  • stabilize their structures
  • we need to be able to serialize/deserialize them
    • without loss of information & connections
      • this will include info about how to handle
        • breaking thing up appropriately (files, JSON nodes, etc)
        • reassembling them correctly (rebuilding connections, etc.)
    • this allows us to decouple things from perl
      • So tools can be written in other langauges
        • that operate on these structures

• Build tools to generate Opcodes from the AST

  • this is gives Guacamole, etc. a path
  • we already have the AST from the Opcodes

Step 3. -

• We now have the following:

  • A spec for the Perl AST
  • A spec for the Perl Opcodes
  • All tools built should operate on the spec level only
    • the process is reversable
      • AST => Ops
      • Ops => AST
  • All that is missing is the runtime, so ...

• Build a "reference implementation" in Perl of perl

  • this should be both
    • a AST tree walking interpreter
    • a VM that consumes opcodes
  • the key thing is to figure out
    • the stack
    • the runloop
    • stash/pad storage
  • NOTE: this doesn't have to be in Perl
    • but it is the quickest approach

Step 4. - Profit!

• ???