First: Wikipedia, tell us what “Boilerplate Text” is. Everybody got it? Good.

My Perl programs, the ones I write now, and, unless it breaks a thing, the stuff that I pull up and modify, all start with this.

#!/usr/bin/env perl
use strict ;
use warnings ;
use utf8 ;
use feature qw{ postderef say signatures state } ;
no warnings qw{ experimental::postderef experimental::signatures } ;

The hashbang is a thing I pulled from a Python program, and what you get from it is, instead of hardcoding system Perl (/usr/bin/perl), it pulls whichever Perl is the first in your path. This could be system perl, or one you compiled yourself, our your group’s official version.

use strict and use warnings are two of the keys to Modern Perl. What are they? Well, Perl has, by default, a wildly dynamic nature, where this is a valid first use of the variable $foo:

$foo = 1 ;

Let us throw context here. Yes, $foo is a common metasyntactic variable — “I need a variable and I don’t care what I call it; I’ll just choose foo — but let’s pretend this relates to your travel plans, saying if a meal should be provided, and should have been $food. You have created a new variable by misspelling an old variable, and it is valid. If you use strict, you get this error:

Global symbol "$foo" requires explicit package name (did you forget to declare "my $foo"?) at ./test.pl line 10.
Execution of ./test.pl aborted due to compilation errors.

warnings go a little deeper. Instead of operator overloading, Perl overloads the variables. Unlike, for example, Javascript, where two strings are concatenated with +, Perl uses that just for mathematical addition, but will find a numerically allowable take on a string and use that, so "23" + 1 will be 24, and "a" + 1 is NaN thus 0 plus one, so one. But “a” in math is probably not what you want, so Perl warns you:

Argument "a" isn't numeric in addition (+) at ./test.pl line 10.

use utf8 means that the internal variable representation mentioned in warnings has strings as UTF-8 instead of ASCII. There is much more to utf8 than I ever use, but I always put it in my programs. 🤷

use feature turns on a number of features. You can use them, but without those features, your Perl will still work for the ancient program your work relies on. The ones I use:

  • say: Time was, if you wanted a newline, you would have to print "\n", and if your string was not to be interpreted, print 'like this', you would have to have it be an interpreted, double-quote string print "like this\n" so Perl could handle the newline, or do print 'like this' . "\n", and isn’t appending newlines just always ugly? I still have reason to print, but 99% of my output is done with say.
  • state: I use this rarely, but it is occasionally useful. Say we wanted an iterator function, where each time it is called, it returns the next digit. This — sub iterator { my $x = 1 ; return $x++ } — would not work, but would instead just return 2. You could use lexical scoping — { my $x = 1 ; sub iterator { return $x++ } } — but that’s double the curly quotes. Using state — sub iterator { state $x = 1 ; return $x++ } — and $x will retain value between calls.
  • postderef: This is a thing I like and use, but I really get where people don’t. Between YAML, JSON and DBI, I get and use a lot of references, and I’m much more likely to create a hashref than a hash, or an arrayref for an array. If I want to use $hashref, it used to be keys %{$hashref}, and it could get worse if it’s really $data->{hashref} you want the keys for. I find keys $data->{->%* and $arrayref->@* very readable, and that keeps it in a way that $json->encode($arrayref) doesn’t complain about.
  • signatures: This is mostly another readability thing. Take this function: sub squared { my $number = shift @_ ; return $number ** 2 } (Example code doesn’t have to be a smart thing to write, it just has to make a point.) Most languages would put $number in the function definition, and that’s what signatures gives us: sub squared ( $number )){ return $number ** 2 }

Both signatures and postderef are new, and as such, they currently carry warnings about their use, although I cannot coerce the postderef warning right now.

The signatures feature is experimental at ./test.pl line 15.

A little more on signatures (I have to justify putting them in the title). It does a little more than just keeping you from having to shift out of @_. I don’t like long argument lists, because then you start getting problems with forgetting order — “Is it mysub( $foo, $bar, $blee, $quuz) or mysub( $foo,$bar, $quuz, $blee)?” — which is another argument for using hashrefs and calling $ref->{foo} inside, but if you have a small number (three or less), you can have a manageable set of signatures.

sub mysub ( $foo, $bar=1 , $blee='armada' , $quuz='fmep') { ... }

This auto-sets lexical scope, so that the $foo inside mysub() is not the same as the $foo outside. Also, it presets initial variables, so $bar is 1 for mysub('foo') but 21 for mysub('foo',21). It also ensures that all arguments are there: if I try mysub(), I get:

Too few arguments for subroutine 'main::mysub' at ./test.pl line 34.

You could do a lot of that without signatures…

sub mysub {
    my ( $foo, $bar, $blee, $quuz ) = @_;
    croak 'too few arguments' unless defined $foo;
    $bar  = defined $bar  ? $bar  : 1;
    $blee = defined $blee ? $blee : 'armada';
    $quuz = defined $quuz ? $quuz : 'fmep';
}

But don’t you find that verbose and ugly? I’d rather get as quickly as I can to the details that matter.

Note: @bokutin asked about //=. I admit that this is an operator I hardly use and never remember. The above non-signature sub would be better written as:

sub mysub {
    my ( $foo, $bar, $blee, $quuz ) = @_;
    croak 'too few arguments' unless defined $foo;
    $bar  //= 1;
    $blee //= 'armada';
    $quuz //= 'fmep';
}

I agree, that’s better. I still like signatures.

That is my standard Perl boilerplate. I have a number of go-to modules, which I’ll get into in another blog post, but this is a code block I use for everything. If you have suggestions for different cool features to start using, like lexical_subs, I’d be glad to know how you use them.

If you have any questions or comments, I would be glad to hear it. Ask me on Twitter or make an issue on my blog repo.