skids

diff --git a/src/vm/moar/ops/container.c b/src/vm/moar/ops/container.c
index aaa7e63..b925ad6 100644
--- a/src/vm/moar/ops/container.c
+++ b/src/vm/moar/ops/container.c
@@ -85,13 +85,15 @@ static void rakudo_scalar_store(MVMThreadContext *tc, MVMObject *cont, MVMObject
     if (!obj) {
         MVM_exception_throw_adhoc(tc, "Cannot assign a null value to a Perl 6 scalar");
     }
-    else if (STABLE(obj)->WHAT == get_nil()) {
-        obj = rcd->the_default;

skids

diff --git a/src/core/Signature.pm b/src/core/Signature.pm
index 129b4e9..0cbe262 100644
--- a/src/core/Signature.pm
+++ b/src/core/Signature.pm
@@ -138,16 +138,26 @@ my class Signature { # declared in BOOTSTRAP

 multi sub infix:<eqv>(Signature $a, Signature $b) { $a.perl eq $b.perl }
 Perl6::Metamodel::Configuration.set_multi_sig_comparator(
-    -> \a, \b { my $sa = a.signature.^find_private_method('gistperl')(
-                     a.signature, True, :where(-> $ { Nil }));

skids

$ time perl6 -e 'my %sw := SetHash.new; my $size = 5000000; for (0..^$size).pick xx 200000 -> $s is copy, $e is copy { ($s,$e) = $e, $s if $e < $s; %sw{+$s} +^= 1; %sw{+$e + 1} +^= 1 if $e < $size - 1; }; my $accum; my $o = 0; for %sw.keys.sort.rotor(2 => -1) -> ($s, $e) { $o +^= 1; $accum += $e - $s if $o; LAST { $accum += $size - $e unless $o }}; $accum.say'
2495015

real    1m25.943s
user    1m25.832s
sys     0m0.076s

# Just need two orders of magnitude improvement :-)

# Algorithm note: decompose each run of switch throws into throwing all the switches right (inclusive) of the leftmost switch,

skids

use nqp;
use QAST:from<NQP>;
sub EXPORT(|) {
    role Control::Never {
        rule statement_control:sym<never> {
            <sym><.kok> <xblock>
        }
    }
    role Control::Never::Actions {
        method statement_control:sym<never>(|c) {

skids

So this is why I get lost in NQP/MoarVM.

Objective: figure out how to create a non-resumable subclass of Exception

Step 1: Look at Exception, notice "method resumable" checks $!ex at key 'resume' and "method resume" just calls nqp::resume
Step 2: Well, I could just override those, but what if the nqp op is called directly somewhere.  Better investigate
Step 3: OK $!ex is made by calling "nqp::newexception". rgrep nqp tree for it
step 4: OK that's mapped to MoarVM "newexeception" op.  Resist the urge to say "this goes to VM level too?!?"
        since exceptions are pretty low level VM things. and I knew that would happen in this case.

skids

There's a similar problem when using inline doc style: declarative POD cannot be easily nested inside a section, if that is what you want.  For example if you are following manpage conventions, you'd want to start a DESCRIPTION block, then drop out of POD and put declaratives around your module, then end the DESCRIPTION block (explicitly, or implicitly by starting, say, a SEE ALSO block.

There's actually a way to get part of the way (=alias directives) but it is NYI.   Having the ability to =alias before a declarative pod directive and populate the alias with the rendered pod would be nice.   e.g.

=alias DOCSFORFOO
#| does foo things
sub foo {
}

=begin DECRIPTION

skids

So, after saying yesterday on IRC that a bottom-up design shoud come first for packish things
I somehow came over with a case of hypocrisy and thought a bit about top-downish stuff.

Suppose we had a role-like MOP construct (bikeshedding welcome on names):

packing foo {
    template bar {
        # pack template bottom level constructs
        # templates can nest, and can come from mixing other packings

skids

Definitely not a complete list of modules with while/loop in last statement of sub or whatnot.
Many are likely harmless but might bear correcting just for sake of sane return value.
Mostly I just did directories of authors whos names I remembered seeing a lot.

awwaiid/p6-lrep
azawawi/perl6-electron
colomon/mandelbrot
colomon/io-prompter
colomon/Test-Junkie
hoelzro/p6-priorityqueue

skids

    regex ℙ_production {
        ( <ℙ_new-lexical-item-name> ) <ℙ_lexical_separator>
        '::=' <ℙ_lexical_separator>
        # This next bit should be producing a prestructured lol-ish thing.
        # Two things NYI|buggy prevent that:
        #    1) @<alternatives>=(...)*
        #    2) /(...) :my $f = $0; ($f)/
        # So though it may appear the below several lines would be simpler
        # put as [ $<a>=a | $<b>=b | $<c>=c ]* % <sep>
        # they are kept in a future-looking form.

skids

A consideration of ‹ and › as quasi injection delimiters, or part thereof.

The compose-key (and thus texas mappings) are .< and .>

Advantage #0: easy to type
Disadvantage #1: texas not really symmetrical
Advantage #2: .> is not syntax for anything, I think (?)
Disadvantage #3: .< is syntax for $_<
Disadvantage #4: visual confusion with < and >