Implementation goals for Perlcon 2019

[masak]

(It's possible some of this issue should be organized as a project, but I haven't learned enough about GitHub projects yet.)

tl;dr

I plan to go to YAPC::Europe^WPerlcon in Riga, and I plan to speak there. I would personally be very surprised if it wasn't 007 and macros I ended up talking about.

So, working backwards from that talk, there are some goals I would like for 007 to have met at that point.

(I'm not so good with deadlines, or rather, they have proven to be mostly wishful thinking. See #335. No plan of attack survives its first meeting with the enemy — and by "enemy", I guess I mean "my general lack of momentum" — but even so, it's better to have a plan and to make goals than not to.)

Hit-the-piñata 🎉 goals

• Separate the logic from the Qtypes (#492)
• Refurbished object system
  • Classes (#32)
  • Enums (#265)
• New runtime (replacing the old one)
  • Compilation visualizer (really good during a talk — in fact, optimized for the talk)
  • Various optimizations
• Examples cheating (which would also allow us to close #194)
• Rudimentary regexes (#174 and #348)

Shoot-for-the-stars :star: goals

• Hacker News completeness
• Massive cheating
• Contextual macros
• Modules
  • Module installer
  • All of it by not cheating
• Complete regex system
  • Grammars
  • A new parser
    • Better error messages
  • is parsed
  • Undoing all the cheating (examples + massive) and replacing it with the real deal
• SEND + MORE = MONEY example implemented as separate languages with transformations between them
• No loose ends/"this is unimplemented" left in the documentation
• More ambitious phrasing in README.
• v1.0
• Language renamed Alma

More details, please

I expect to come back here and write things out more concretely, including how things depend on each other, and what we expect to be challenging. But the above is a decent start. (Update: Finally wrote one, see comment below.)

[masak]

I don't know exactly why, but somehow the metaphor of striking a piñata while blindfolded feels like a very apt one for the 007 project.

[masak]

Talk accepted! Yay!

[vendethiel]

I should definitely consider going then :)

[masak]

@vendethiel I will be sorely disappointed unless you at least consider it 😉

[masak]

Closing this one, as PerlCon has now come and gone.

(I still hope to find time to expand on the roadmap for those goals, because it's still a good set of goals even after the conf.)

[masak]

A more detailed roadmap

My current focus on Bel leaves me in practice with zero tuits to advance Alma. But I don't consider the Alma project dead, or finished, or stable; it's more like it reached quite far in its Phase I, where Alma was an experimental interpreter. With the improved perspective I now have from taking a few steps away from the project, I see quite clearly that it's necessary for Alma to enter a Phase II, with a couple of big milestones to drive it:

• New runtime/Compile to bytecode. I'm thinking a small, well-defined subset of Alma can be implemented in LLVM. As a first step, maybe not even the Qtypes need to be exposed in userspace.
• Implement the ambitious parser. As pointed out in #293 and #485, the really interesting uses of macros depend on parser technology which, to be blunt, doesn't seem to exist yet. (I have some personal notes where it simply says "parser handover; parser relinquishing". This is crucial both for DSLs and for syntax extension, and it will probably figure heavily in the unquote story as well.) At this point I don't care if the parser is bootstrapped in Alma itself; that can come later. The important thing is a clear proof-of-concept that this kind of parser works.
• Object system/MOP. This one isn't and never was a central part of Alma's mission, but it still feels like an important thing to get into place.
• Reflected Qtypes, quasis, macros. Should let us close #335, which is mostly blocking on #30, which is completely blocking on the ambitious parser.
  • Edit: Having the assert macro from #417 as a driving use case here is probably a good idea.
• Userland regex/grammar story, which of course should mesh well with is parsed, and the Alma grammar.
• A strong slangs/langs story.
• Modules.
• Control flow primitives/low-level code à la #145.
  • Edit: The conjectural part of the assert macro from #417 might interact with and drive this bit. This one is not at all clear, though, as it's a macro that acts "outside" of its macro argument, on upstream local data dependencies.

Again, the question is where I'll be able to come by one or more clones of myself with the tuits to do the above, but... there it is: an updated roadmap for Alma.