In Perl 6, you can ask the sequence operator to build a desired sequence for you. It can be arithmetic or geometric progression. All you need is to show the beginning of the sequence to Perl, for example:

.say for 3, 5 ... 11;

This prints numbers 3, 5, 7, 9, and 11. Or:

.say for 2, 4, 8 ... 64;

This code prints powers of 2 from 2 to 64: 2, 4, 8, 16, 32, and 64.

I am going to try understanding how that works in Rakudo. First of all, look into the src/core/operators.pm file, which keeps a lot of different operators, including a few versions of the ... operator. The one we need looks really simple:

multi sub infix:<...>(\a, Mu \b) {
    Seq.new(SEQUENCE(a, b).iterator)
}

Now, the main work is done inside the SEQUENCE sub. Before we dive there, it is important to understand what its arguments a and b receive.

In the case of, say, 3, 5 ... 11, the first argument is a list 3, 5, and the second argument is a single value 11.

These values land in the parameters of the routine:

sub SEQUENCE(\left, Mu \right, :$exclude_end) {
    . . .
}

What happens next is not that easy to grasp. Here is a screenshot of the complete function:

It contains about 350 lines of code and includes a couple of functions. Nevertheless, let’s try.

What you see first, is creating iterators for both left and right operands:

my \righti := (nqp::iscont(right) ?? right !! [right]).iterator;

my \lefti := left.iterator;

Then, the code loops over the left operand and builds an array @tail out of its data:

while !((my \value := lefti.pull-one) =:= IterationEnd) {
    $looped = True;
    if nqp::istype(value,Code) { $code = value; last }
    if $end_code_arity != 0 {
        @end_tail.push(value);
        if +@end_tail >= $end_code_arity {
            @end_tail.shift xx (@end_tail.elems - $end_code_arity)
                unless $end_code_arity ~~ -Inf;

            if $endpoint(|@end_tail) {
                $stop = 1;
                @tail.push(value) unless $exclude_end;
                last;
            }
        }
    }
    elsif value ~~ $endpoint {
        $stop = 1;
        @tail.push(value) unless $exclude_end;
        last;
    }
    @tail.push(value);
}

I leave you reading and understand this piece of code as an exercise, but for the given example, the @tail array will just contain two values: 3 and 5.

> .say for 3,5...11;
multi sub infix:<...>(\a, Mu \b)
List    # nqp::say(a.^name);
~~3     # nqp::say('~~' ~ value);
~~5     # nqp::say('~~' ~ value);
elems=2 # nqp::say('elems='~@tail.elems);
0=3     # nqp::say('0='~@tail[0]);
1=5     # nqp::say('1='~@tail[1]);

This output shows some debug data print outs that I added to the source code to see how it works. The green comments show the corresponding print instructions.

That’s it for today. See you tomorrow with more stuff from the sequence operator. Tomorrow, we have to understand how the list 3, 5 tells Perl 6 to generate increasing values with step 1.

Today, we will see how Perl 6 helps to keep our programs better.

Redeclaration of a variable

Examine the following program:

my $x = 1;
my $x = 2;
say $x;

You can immediately see that this program is not entirely correct. Either we meant to assign a new value to $x or to create a new variable with a different name. In either case, compiler has no idea and complains:

$ perl6 redecl.pl 
Potential difficulties:
    Redeclaration of symbol '$x'
    at /Users/ash/redecl.pl:2
    ------> my $x⏏ = 2;
2

You see a runtime warning, while the program does not stop. Let us find out where it happens in the source code.

When you declare a variable, the grammar matches the corresponding text and calls the variable_declarator action method. It is quite compact but nevertheless I will not quote it completely.

class Perl6::Actions is HLL::Actions does STDActions {
    . . .

    method variable_declarator($/) {
        . . .
    }

    . . .
}

By the way, you can see here how Perl 6 treats a variable name:

 my $past := $<variable>.ast;
 my $sigil := $<variable><sigil>;
 my $twigil := $<variable><twigil>;
 my $desigilname := ~$<variable><desigilname>;
 my $name := $sigil ~ $twigil ~ $desigilname;

The name of a variable is a concatenation of a sigil, a twigil and an identifier (which is called desigiled name in the code).

Then, if we’ve got a proper variable name, check it against an existing lexpad:

if $<variable><desigilname> {
    my $lex := $*W.cur_lexpad();
    if $lex.symbol($name) {
        $/.typed_worry('X::Redeclaration', symbol => $name);
    }

If the name is known, generate a warning. If everything is fine, create a variable declaration:

make declare_variable($/, $past, ~$sigil, ~$twigil, $desigilname,
                      $<trait>, $<semilist>, :@post);

Redeclaration of a routine

Now, let us try to re-create a subroutine:

sub f() {}
sub f() {}

This may only be OK if the subs are declared as multi-subs. With the given code, the program will not even compile:

===SORRY!=== Error while compiling /Users/ash/redecl.pl
Redeclaration of routine 'f' (did you mean to declare a multi-sub?)
at /Users/ash/redecl.pl:6
------> sub f() {}⏏<EOL>

This time, it happens in a much more complicated method, routine_def:

method routine_def($/) {
     . . .

     my $predeclared := $outer.symbol($name);
     if $predeclared {
         my $Routine := $*W.find_symbol(['Routine'], :setting-only);
         unless nqp::istype( $predeclared<value>, $Routine)
                && nqp::getattr_i($predeclared<value>, $Routine, '$!yada') {
              $*W.throw($/, ['X', 'Redeclaration'],
                        symbol => ~$<deflongname>.ast,
                        what => 'routine',
              );
         }
     }

The exception

The code of the exception is rather simple. Here it is:

my class X::Redeclaration does X::Comp {
    has $.symbol;
    has $.postfix = '';
    has $.what = 'symbol';
    method message() {
        "Redeclaration of $.what '$.symbol'"
        ~ (" $.postfix" if $.postfix)
        ~ (" (did you mean to declare a multi-sub?)" if $.what eq 'routine');
    }
}

As you see, depending on the value of $.what, it prints either a short message or adds a suggestion to use the multi keyword.

Welcome back! As you might notice, there was a small gap in the daily post flow.

Before we are back to the Rakudo internals, a couple of words about some changes here. First of all, every post is now marked with either 🦋 or 🔬 (or with indistinguishable rectangles □ if your browser cannot display an emoji :-). These characters mean two categories of posts here: a butterfly stands for Perl 6 syntax, while a microscope is for Perl 6 internals. In the first category, only user-level aspects or Perl 6 are discussed. In the second, we dig into the source codes of Rakudo. All the past post are updated accordingly.

The second change is that I will occasionally post more articles in the Perl 6 syntax category because I found out that non-Russian speakers often like my Russian blog posts. Those posts are mostly short texts explaining interesting features of Perl 6, such as the =~= operator or promises.

* * *

OK, now we have to talk about the EVAL routine. It is defined in the src/core/ForeignCode.pm file as a multi-function. Let us see at their signatures:

proto sub EVAL($code is copy where Blob|Cool, 
               Str() :$lang = 'perl6',
               PseudoStash :$context, *%n)

multi sub EVAL($code, 
               Str :$lang where { ($lang // '') eq 'Perl5' },
               PseudoStash :$context)

Notice that one of the function is a proto, while another is the only multi-candidate. Unlike many other cases that you can see in the sources of Rakudo, this proto routine contains code. Refer to one of the recent blog posts to see how it works.

We start with an example from the first part of the article.

EVAL('say 123');

Here, the passed value is Str, and it is caught by the proto sub, as its first argument can be Cool.

The sub creates a compiler for the given language (which is Perl 6 by default).

my $compiler := nqp::getcomp($lang);

The next step in the sub is to make a string out of $code. In this first example, this task is trivial.

$code = nqp::istype($code,Blob) ?? $code.decode(
    $compiler.cli-options<encoding> // 'utf8'
) !! $code.Str;

Finally, the string is compiled:

my $compiled := $compiler.compile:
    $code,
    :outer_ctx($eval_ctx),
    :global(GLOBAL),
    :mast_frames(mast_frames),
    |(:optimize($_) with nqp::getcomp('perl6').cli-options<optimize>),
    |(%(:grammar($LANG<MAIN>), :actions($LANG<MAIN-actions>)) if $LANG);

After compilation, you get an object of the ForeignCode type. This class is a child class of Callable, so the object can be called and returned (actually, it’s not quite clear how it happens):

$compiled();

Now you can understand that single quotes in the second example with curly braces still create an executable code:

EVAL('say {456}');

Here, the whole string is compiled as it was a Perl 6 code, and the code block there is a code block, which Perl should execute, and thus say gets a code block, so it calls its gist method to prepare the output:

> {456}.gist
-> ;; $_? is raw { #`(Block|140388575216888) ... }