6macros: discussing the finer points of Perl 6 macros, Qtrees, and how to stay sane | irclog: irclog.perlgeek.de/6macros/today
Set by moderator on 26 March 2015.
01:48 ilbot3 joined
moderator 6macros: discussing the finer points of Perl 6 macros, Qtrees, and how to stay sane | irclog: irclog.perlgeek.de/6macros/today
08:14 Ven joined 10:44 Ven joined 12:35 Ven joined 15:16 Ven joined
vendethiel ideone.com/3J0H2y 19:54
that's an example of CL-style macros, masak :-)
I mean, late-binding :P
masak looks 19:57
vendethiel I'm also thinking about macros for a language 19:58
I wanted the language to be simple and moduler, as in: 19:59
cat foo | parse | emit-bytecode | vm
cat foo | parse | desugar* | emit-bytecode | vm
but -- how do I add macros to that? I need the vm to run code, and I need the ast2bc to compile macros...
masak good question. 20:06
"Lisp in Small Pieces" has this notion about "internal macros" vs "external macros".
sounds like you're touching upon that notion.
vendethiel mmh. 20:07
perl6 just had to borrow the notion of "BEGIN3
BEGIN* for macros
masak see #perl6 20:10
vendethiel mmh 20:11
you've still solved the "how" problem
now need to get all the ramifications correctly
masak aye 20:14
vendethiel I'm a bit at loss here.
masak my example ought to have worked and been a faithful translation of yours. 20:15
vendethiel no 20:16
there's no quasi in my code
I'm not *generating* that push. it's run as macro time
masak if it's run at macro time, how come at runtime the list is empty at the first print? 20:17
vendethiel as I said -- welcome to common lisp macros ;-)
masak there's something I'm missing here
vendethiel and "late binding"
it's run as it's parsed
masak I mean, it seems *better*, but I don't see how it could possibly work
vendethiel well
it can't have reader macros
if it didn't work like that
masak I must say it works much more like how people expect Perl 6 macros to work. 20:18
I'd very much like to see a consistent model based on this :)
vendethiel yeaaah
masak problem is, Perl 6 provides two "phases" inside a macro 20:19
vendethiel another BIG pain point for the "BIG PLANS" i have for perl6 is the other part of that
masak the compile-time phase, outside the quasi
and the runtime phase, inside the quasi
vendethiel it's the "constant are BEGIN time"
that makes me incredibly sad :(
masak huh
vendethiel I can't do any compile-time metaprogramming
because
role A[::T] { constant NEXT = X[T]; } will fail
because "T" is Mu at the time the constant is parsed 20:20
masak bring it up with TimToady
vendethiel well, it can't work differently, can it?
constants have to be BEGIN-time
but I want them role-parameterization-time
masak so maybe we need a `let` or similar for runtime constants.
vendethiel but I don't want quite "runtime" constants :P
masak *nod* 20:21
vendethiel i still want something that runs at compile-time
masak you know what you could do?
vendethiel no?
paste that in #perl6?
:P
masak you could declare it as a second parameter in the role, with a default
then it would evaluate at the right time
maybe not ideal
but it works today
vendethiel care to give me an example?
masak role A[::T, $NEXT = X[T]] {} 20:22
vendethiel aaah.
masak :)
vendethiel that's actually an interesting idea.
not quite "pretty", tho, but...
masak it works.
vendethiel
.oO( I'll just write a macro to do that )
 20:24
masak :P 20:25
yep, that should probably be possible once I'm done... 20:26
vendethiel so, there's something in progress? :-)
did you get the CL macro system, or do you want me to add a bit on that?
masak "it's lazy and magical, and ponies" 20:27
is what I got
vendethiel haha
masak sorry for being a bit brusque. still doing $work stuff :) 20:28
vendethiel (defmacro mac (arg) (print arg) `(print arg))
then
(mac 5)
it'll first evaluate the macro
printing arg first
then substitute the AST
then evaluate that
masak so, in some sense, the `(mac 5)` remains in the code until runtime? 20:29
vendethiel yeah
there's only runtime
there's nothing else
just like I said before
masak *nod*
vendethiel if everything wasn't runtime, lexer macros wouldn't be possible 20:30
masak I... see.
well, Perl 6 *does* have two phases.
they can even be separated to different physical machines, etc.
vendethiel ;-)
masak and macros most certainly run in the first phase. that won't change.
vendethiel definitely :)
masak do you think that, given the two-phase thing, the type of macros you describe would still be possible? 20:32
vendethiel the type of macros? 20:33
with the print before and after the macro use?
no
lexer macros? probably :P
only because we define our lexer modifications declaratively
whereas cl runs code
masak needs to learn more about lexer macros, then 20:34
vendethiel changing the reader as-you-go
masak would you say Perl 6 has a reader? 20:37
vendethiel I'd say every language has a "reader", in the form of a lexer
in c, I can "augment" the reader by declaring types
masak of all the languages I know, Perl 6 has a lexer the least... :)
vendethiel in c++, I can "change" the reader by declaring templates
s/change/augment/ # copy pasta fail :-) 20:38
masak I see your point, though
vendethiel and in perl6, I just have more control
(though I have even more in lisp..obviously :P)
masak obviously 20:39
vendethiel: did you see irclog.perlgeek.de/6macros/2015-03-29#i_10359341 ? 20:46
vendethiel definitely didn't
vendethiel didn't backlog properly
I'm not sure you can make your AST types so good 20:48
that it's impossible to build an invalid ast :-)
masak I want to allow building an invalid AST 20:50
that's why the hardening mechanism is necessary
and some things that are disallowed in parsing probably should be allowed in hardening 20:51
vendethiel referring-using-strings needs to be banned, but it was never a question in my eyes
"symbols" were always gonna be necessary
masak oh, ok
good to know
thanks for that bit of feedback alone
that basically answers it for me
vendethiel I agree that I don't see a solution with strings :) it can't be made sane! 20:52
elixir, which has strings and "symbols", uses another "data type"
masak everybody else seems to try to do it with strings, in my experience
vendethiel because no other solution around?
masak sure, it's a local minimum, if that's what you mean 20:55
it's just not I particularly strive for in Perl 6
we can do better
vendethiel I agree,I agree:) 20:58
also 21:00
strings aren't really viable with our compile-time lexical variables errors :) 21:01
masak right. 21:06
22:19 vendethiel joined

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!