Recollections from the Raku Core Summit

The Raku Programming Language Collect, Conserve and Remaster Project

Recollections from the Raku Core Summit

Originally published on 18 June 2023 by Jonathan Worthington.

The first Raku Core Summit, a gathering of folks who work on “core” Raku things, was held on the first weekend of June, and I was one of those invited to attend. It’s certainly the case that I’ve been a lot less active in Raku things over the last 18 months, and I hesitated for a moment over whether to go. However, even if I’m not so involved day to day in Raku things at the moment, I’m still keen to see the language and its ecosystem move forward, and – having implemented no small amount of the compiler and runtime since getting involved in 2007 – I figured I’d find something useful to do there!

The area I was especially keen to help with is RakuAST, something I started, and that I’m glad I managed to bring far enough that others could see the potential and were excited enough to pick it up and run with it.

One tricky aspect of implementing Raku is the whole notion of BEGIN time (of course, this is also one of the things that makes Raku powerful and thus is widely used). In short, BEGIN time is about running code during the compile time, and in Raku there’s no separate meta-language; anything you can do at runtime, you can (in principle) do at compile time too. The problem at hand was what to do about references from code running at compile time to lexically scoped symbols in the surrounding scope. Of note, that lexical scope is still being compiled, so doesn’t really exist yet so far as the runtime is concerned. The current compiler deals with this by building up an entire flattened table of everything that is visible, and installing it as a fake outer scope while running the BEGIN-time code. This is rather costly, and the hope in RakuAST was to avoid this kind of approach in general.

A better solution seemed to be at hand by spotting such references during compilation, resolving them, and fixating them – that is, they get compiled as if they were lookups into a constant table. (This copies the suggested approach for quasiquoted code that references symbols in the lexical scope of where the quasiquoted code appears.) This seemed promising, but there’s a problem:

`my $x = BEGIN %*ENV<DEBUG> ?? -> $x { note "Got $x"; foo($x) } !! -> $x { foo($x) };`

It’s fine to post-declare subs, and so there’s no value to fixate. Thankfully, the generalized dispatch mechanism can ride to the rescue; we can:

When compiling Raku code, timing is everything. I knew this and tried to account for it in the RakuAST design from the start, but a couple of things in particular turned out a bit awkward.

I got a decent way into this restructuring work during the core summit, and hope to find time soon to get it a bit further along (I’ve been a mix of busy, tired, and had an eye infection to boot since getting back from the summit, so thus far there’s not been time for it).

I also took part in various other discussions and helped with some other things; those that are probably most worth mentioning are:

Thanks goes to Liz for organizing the summit, to Wendy for keeping everyone so well fed and watered, to the rest of attendees for many interesting discussions over the three days, to TPRF and Rootprompt for sponsoring the event, and to Edument for supporting my attendance.