Hacker News Reader

Comment by wruza | Hacker News Reader

Tbh, for me it was a source of regex knowledge, in retrospect. Remove regexes from perl, what’s left?

I can name: dynamic scoping, implicit filehandles, implicit $_, strange contexts, keyword-like list functions. That’s basically it, and while it sounds fun, it doesn’t do much in a sense of code reduction.

$. $? $$ etc, well you can split lines with .split('\n'), get status as a part of result, call os.getpid().

<>, you can open() or createReadStream().

$_ is just a kids toy. “print if /…/“. Cool.

Lots of global modes under dynamic scoping is questionable. Value semantics and corresponding sigils are horrible, really. Bless is ugh. Subs are meh.

It’s all just regex. Add regex as a language construct and you get the power of perl in it. /…/.test(s), s.match(/…/), s.replace(/…/, '$&') that’s 90% of Perl. Regex saves most lines of code, everything else is just esoteric extra that doesn’t do much. Also no GC.

View on HN

$_ is not just a kid's toy.

Let's say I'm talking to a human programmer. I say, "Read in a line of text. If it ends in a newline, remove the newline." Cool.

But when I talk to a computer, I can't talk that way. It says, "If what ends in a newline?" It doesn't understand "it".

Except in Perl. In Perl, $_ is "it". It's what we're talking about if I don't specify something specific. It's where the line of text gets read into if I don't specify a destination. It's where the newline gets removed from if I don't specify a variable. It lets me ignore specifics when I don't care about those specifics. It makes Perl much closer to how we talk to each other instead of how we talk to computers.

It’s also where you (or your adhd colleague, or future self) introduce a subtle bug if one of your subs reads another stream and you forgot to local $_ in it. And you realize that you don’t always understand what “it” is either and that it is a bad idiom for formal systems. Isn’t worth it in the long run, especially in implicitness-ridden code that perl encourages. Write out your “line” and assign it back:

  if (line ~= /\n$/) {
    line = chop(line)
  }

Or use a library with helper functions akin to chomp(). This sentence-like code with $_ is a sign of wrong level of abstraction which solves small tasks with smaller components. If a language claims text processing capabilities, it should provide a function or a generator that does “read this <enc=utf-8> file and split by <what>[ omitting empty ones but keeping original line numbers in the emitted records][ also erase /\s*#.*$/-comments][ keep original lines too]” among other common tasks. All this $, $. $/ $_ could be just a couple of functions that accept a config argument with few optional callbacks for custom needs. You don’t have to create a whole language to read lines split by newline then by separator. Same for matching vars, could be just a sliding context arg to a properly designed matching api.

You may argue that other languages don’t have it either, well that’s true. Standard libraries usually suck compared to what perl does out of box.

I use Perl for one-liners all the time -- nothing can top it there.

But when the task at hand grows just slightly bigger, to the point where I know I'll need to pass filehandles to/from my own functions, my skin starts to literally itch because of how bad the filehandle situation is.

It's not plain "open F, ..." any more... Is it "*F"? Is it "\*F"? Are those the same thing? (What the hell is a "typeglob"?) Wait, can't I just use "open my $f" and everything works the way it obviously should? Well, kinda, unless you want to be able to store an existing filehandle like STDIN in there, then you have to do something different again... Do I need IO::Handle instead? (Why does that need to exist...?)

Plain perl:

  open my $fh, '<', $filename or die "Couldn't open ${filename}: $!";

Nicer:

  use File::Open qw(fopen);
  ...
  my $fh = fopen $filename;

(and File::Open is simple pure perl code so if I need to distribute the script to random machines I throw App::FatPacker at it to produce a bundled version)

"my $fh" works fine if you only want to work with files you open() yourself, but what if you want to write a function that, based on some test, returns either that or the preexisting STDOUT filehandle?

    sub foo {
      if (whatever) {
        open my $fh, ...;
        return $fh;
      } else {
        return \*STDOUT;
      }
    }

will do the job.

Thanks. I feel certain that I tried exactly this a long time ago, and it didn't work as expected, but I may be misremembering. I'll give this a shot the next time I'm in front of a keyboard -- if it works, it will be nothing short of healing.

Slight oddities converting it into a one-liner but here's a copy and paste from a shell session running in my local WSL slice:

  demeisen=; perl -e 'my $in_fh = (defined $ARGV[0] and $ARGV[0] ne "-") ? do { open my $fh, "<", $ARGV[0] or die "Failed to open $ARGV[0]: $!"; $fh } : \*STDIN; print "Line: ".<$in_fh>;'
  HELLO
  Line: HELLO
  demeisen=; perl -e 'my $in_fh = (defined $ARGV[0] and $ARGV[0] ne "-") ? do { open my $fh, "<", $ARGV[0] or die "Failed to open $ARGV[0]: $!"; $fh } : \*STDIN; print "Line: ".<$in_fh>;' -
  HELLO
  Line: HELLO
  demeisen=; echo 'HI' >tmp/hi  
  demeisen=; perl -e 'my $in_fh = (defined $ARGV[0] and $ARGV[0] ne "-") ? do { open my $fh, "<", $ARGV[0] or die "Failed to open $ARGV[0]: $!"; $fh } : \*STDIN; print "Line: ".<$in_fh>;' tmp/hi
  Line: HI
  demeisen=;

>>Remove regexes from perl, what’s left?

qw, qr, references, multiline regexes, a far advanced OO ecosystem, Data::Dumper, map, grep, pack/unpack, DBM, ``, top class unicode handling, given/when, functional programming etc etc.

In fact I've barely scratched the surface.

You must grok Perl more.

OO is the part of Perl that I grok the least. What about it makes it "far advanced" compared to, say, Java or C++ versions of OO?

https://metacpan.org/pod/Moose

Moose is definitely what I go for when I have to use OO with Perl, this more than a decade old and stable for production use cases.

Its based on Class::MOP, which is in turn based on Meta Object Protocol, the same concepts on which CLOS(Common Lisp Object System) is based on. Its always nice to have CLOS goodness in Perl. For eg- https://metacpan.org/dist/Moose/view/lib/Moose/Manual/Method... these methods like before, after, around do fix need for design patterns to a large extent.

But of course the more you explore, the more you discover the possibilities with this.

To a large extent I think Perl brings OO and functional paradigm in a far better package than Python does.

Perhaps HOP needs a new chapter for OO given how few people are aware of this.

The key idea in CLOS is multi method dispatch, which Moose does not provide.

Moose out of the box has its own set of problems, leading to other workarounds, e.g.

Moo (a similar but much less bloated solution)

MooseX::Extended - Moose with more sensible/correct defaults

https://metacpan.org/dist/MooseX-Extended/view/lib/MooseX/Ex...

OO in Perl is lower level compared to Java or C++, so e.g. instead of having a class construct, you have to simulate a class - typically by binding (blessing) a data structure to a namespace. Then any functions in the namespace become methods that can be called via the data structure (object).

In practice, this has turned out to be a mixed blessing because of how tedious it is to do this repeatedly. So over the years there have been many libraries created to make this easier, each with different features.

There's currently work underway to modernise Perl's built in OO to address these problems.

On the other hand, this makes some things easier, e.g. Design by Contract can be added to Perl just by writing a library.

Another example, adding traits to PHP required updating PHP itself, whereas in Perl there are libraries to do that.

Won’t nitpick, how much of what’s useful here is unique to perl? You could make a similar list for most languages and barely scratch the surface.

top class unicode handling

This one I remember struggling with, afair due to io vs perlio layers impedance mismatch and use utf8. Not sure about now, but perl was anything but unicode for the clueless.

given/when are deprecated and will be removed in a future release.

References are a PITA - cumbersome to use, and they make code less readable.

Perl doesn't have first class functions (you can pass or return functions via references, but that's cumbersome and less readable compared to languages with better FP support).

>>given/when are deprecated and will be removed in a future release.

given/when won't be removed, too much back wards compatibility issues.

>>References are a PITA - cumbersome to use, and they make code less readable.

Depends what you mean readable though. In python you can't tell whats a variable and whats a list, and whats a dictionary by looking. One can claim whole language is unreadable since variables are needed at every step.

>>but that's cumbersome and less readable compared to languages with better FP support

Sure let them add all the other practical goodness of Perl, then we can use them.

Consider

    % perl -v | head -2

    This is perl 5, version 40, subversion 0 (v5.40.0) built for x86_64-linux
    % 
    % perl -E 'use feature "switch"; my ($x, $y); given ($x) { $y = 1 when /^abc/ }'
    given is deprecated at -e line 1.
    when is deprecated at -e line 1.

This is also mentioned in the docs:

  Smartmatch is now seen as a failed experiment and was marked as deprecated in Perl 5.37.10. 
  This includes the when and given keywords, as well as the smartmatch operator ~~. 
  The feature will be removed entirely in the Perl 5.42.0 production release.

https://perldoc.perl.org/perldeprecation#Smartmatch

>> In python you can't tell whats a variable and whats a list, and whats a dictionary by looking.

In Perl, sigils are used to distinguish between scalars/lists/hashes. Sigils and references are not the same thing.

https://www.perl.com/article/on-sigils

https://perldoc.perl.org/perlreftut

yes, trailblazing.

yes, enables brevity.

yes, probably comprehensive offline std docs.

,...

https://perldoc.perl.org/perldiag should set off some alarms.

Reply on news.ycombinator.com