++ed by:
SHARYANTO PERLANCAR TOBYINK MARCC MISHIN

8 PAUSE users

Marc Chantreux
CPAN version Endorse eiro on Coderwall

SYNOPSIS

If you're used to a unix shell, Windows Powershell or any langage comming with the notion of streams, perl could be frustrating as functions like map and grep only works with arrays.

The goodness of it is that | is an on demand operator that can easily compose actions on potentially very large amount of data in a very memory and you can control the amount of consummed data in a friendly way.

Perlude gives a better | to Perl: as it works on scalars which can be both strings (like unix shell), numbers or references (like powershell).

In Perlude::Tutorial i show examples

The big difference is there is no | operator, so the generator is used as function parameter instead of lhs of the pipe (still, the ease of composition remains). So the perlude notation of

    seq 1000 | sed 5q

is

    take 5, range 1, 1000

this code returns a new iterator you want to consume, maybe to fold it in a array, maybe to act on each lastly generated element with the keyword now (as "now, compute things you learnt to compute").

    my @five = fold take 5, range 1, 1000;
    map {say} take 5, range 1, 1000;

a classical, memory agressive, Perl code would be

    map {say} (1..1000)[0..4]

Note that

    map {say} (1..4)[0..1000]

is an error when

    now {say} take 1000, range 1,4

Perlude stole some keywords from the Haskell Prelude (mainly) to make iterators easy to combine and consume.

    map say, grep /foo/, <STDIN>;

Perlude provides "streamed" counterpart where a stream is a set (whole or partial) of results an iterator can return.

    now {say} filter {/foo/} lines \*STDIN;

Now we'll define the concepts under Perlude. the functions provided are in the next section.

an iterator

is a function reference that can produce a list of at least one element at each calls. an exhausted iterator returns an empty list.

Counter is a basic example of iterator

    my $counter = sub {
        state $x = 0;
        $x++
    };

If you use Perl before 5.10, you can write

    my $counter = do {
        my $x = 0;
        sub {$x++}
    };

(see "Persistent variables with closures") in the perldoc perlsub.

an iteration

one call of an iterator

    print $counter->();

a stream

the list of all elements an iterator can produce (it may be infinite).

the five first elements of the stream of $counter (if it wasn't previously used) is

    my @top5 = map $counter->(), 1..5;

the perlude counterpart is

    my @top5 = fold take 5, $counter;

a generator

is a function that retuns an iterator.

    sub counter ($) {
        my $x = $_[0];
        # iterator starts here
        sub { $x++ }
    }

    my $iterator = counter 1;
    print $iterator->();
    

a filter

is a function that take an iterator as argument and returns an iterator, applying a behavior to the elements of the stream.

such behavior can be removing or adding elements of the stream, exhaust it or applying a function in the elements of it.

some filters are Perlude counterparts of the perl map and grep, other can control the way the stream is consumed (think of them as unix shell filters).

a consumer

filters are about combining things nothing is computed as long as you don't use the stream. consumers actually starts to stream (iterate on) them (think python3 list() or the perl6 &eager).

to sumarize

A stream is a list finished by an empty list (which makes sense if you come from a functional language).

        (2,4,6,8,10,())

A an iterator is a function that can return the elements of an iterator one by one. A generator is a function that retuns the iterator

        sub from_to { # the generator
                my ( $from, $to ) = @_;
                sub { # the iterator
                        return () if $from > $to;
                        my $r = $from;
                        $from+=2;
                        return $r
                }
        }

note that perlude authors are used to implicit notations so we're used to write more like

        sub {
                return if $from > $to;
                (my $r, $from) = ( $from, $from + 2 );
                $r;
        }

(see the code of the &lines generator)

Examples

find the first five zsh users

    my @top5 =
        fold
        take 5,
        filter {/zsh$/}
        lines "/etc/passwd";

A math example: every elements of fibo below 1000 (1 element a time in memory)

    use Perlude;
    use strict;
    use warnings;

    sub fibo {
        my @seed = @_;
        sub {
            push @seed, $seed[0] + $seed[1];
            shift @seed
        }
    }

    now {say} takeWhile { $_ < 1000 } fibo 1,1;

Used to shell? the Perlude version of

    yes "happy birthday" | sed 5q

is

    sub yes ($msg) { sub { $msg } }
    now {say} take 5, yes "happy birthday" 

A sysop example: throw your shellscripts away

    use Perlude;
    use strictures;
    use 5.10.0;

    # iterator on a glob matches stolen from Perlude::Sh module
    sub ls {
        my $glob = glob shift;
        my $match;
        sub {
            return $match while $match = <$glob>;
            ();
        }
    }

    # show every txt files in /tmp
    now {say} ls "/tmp/*txt

    # remove empty files from tmp

    now { unlink if -f && ! -s } ls "/tmp/*"

    # something more reusable/readable ?

    sub is_empty_file { -f && ! -s }
    sub empty_files_of { filter {is_empty_file} shift }
    sub rm { now {unlink} shift }

    rm empty_files_of ls "/tmp/*./txt";

Function composition

When relevant, i used the Haskell Prelude documentation descriptions and examples. for example, the take documentation comes from http://hackage.haskell.org/packages/archive/base/latest/doc/html/Prelude.html#v:take.

Functions

generators

range $begin, [ $end, [ $step ] ]

A range of numbers from $begin to $end (infinity if $end isn't set) $step by $step.

    range 5     # from 5 to infinity
    range 5,9   # 5, 6, 7, 8, 9
    range 5,9,2 # 5, 7, 9

cycle @set

infinitly loop on a set of values

    cycle 1,4,7

    # 1,4,7,1,4,7,1,4,7,1,4,7,1,4,7,...

records $ref

given any kind of ref that implements the "<>" iterator, returns a Perlude compliant iterator.

    now {print if /data/} records do {
        open my $fh,"foo";
        $fh;
    };

as_open

just easier (yet safer?) to use wrapper on the sub described in perldoc -f open (also "open" in perlfunc).

the goal is to have an wrapper on open does a coercion (just return @_ if nothing to do). so

  • don't carre about prototype (so you can call it with an array, not only a list)

  • return a FILEHANDLE instead of having a side effect on the first variable

  • just return a FILEHANDLE passed as argument (so it's a coercion from @_ to an open handler).

        open FILEHANDLE
        open EXPR
        open MODE,EXPR
        open MODE,EXPR,LIST
        open MODE,EXPR,REF

lines @openargs

if $openargs[0] is a string, &open @openargs (nothing done there if it's already a file handler).

return an iterator that chomp the records of the open file.

so

    now {say} lines "/etc/passwd" 

can be written like

    now {say} apply { chomp; $_ } do {
        open my $fh, "/etc/passwd";
        sub {
            return unless defined my $line = <$fh>;
            chomp $line;
            $line;
        }
    }

filters

filters are composition functions that take a stream and returns a modified stream.

filter $xs

the Perlude counterpart of grep.

    sub odds () { filter { $_ % 2 } shift }

apply

the Perlude counterpart of map.

    sub double { apply {$_*2} shift }

take $n, $xs

take $n, applied to a list $xs, returns the prefix of $xs of length $n, or $xs itself if $n > length $xs:

    sub top10 { take 10, shift }

    take 5, range 1, 10
    # 1, 2, 3, 4, 5, ()

    take 5, range 1, 3
    # 1, 2, 3, ()

takeWhile $predicate, $xs

takeWhile, applied to a predicate $p and a list $xs, returns the longest prefix (possibly empty) of $xs of elements that satisfy $p

    takeWhile { 10 > ($_*2) } range 1,5
    # 1, 2, 3, 4

drop $n, $xs

drop $n $xs returns the suffix of $xs after the first $n elements, or () if $n > length $xs:

    drop 3, range 1,5
    # 4 , 5

    drop 3, range 1,2
    # ()

dropWhile $predicate, $xs

dropWhile $predicate, $xs returns the suffix remaining after dropWhile $predicate, $xs

     dropWhile { $_ < 3 } unfold [1,2,3,4,5,1,2,3] # [3,4,5,1,2,3]
     dropWhile { $_ < 9 } unfold [1,2,3]           # []
     dropWhile { $_ < 0 } unfold [1,2,3]           # [1,2,3]

misc

unfold $array

unfold returns an iterator on the $array ref so that every Perlude goodies can be applied. there is no side effect on the referenced array.

    my @lower = fold takeWhile {/data/} unfold $abstract

see also fold

pairs $hash

returns an iterator on the pairs of $hash stored in a 2 items array ref.

    now {
        my ( $k, $v ) = @$_;
        say "$k : $v";
    } pairs {qw< a A b B >};

aims to be equivalent to

    my $hash = {qw< a A b B >};
    while ( my ( $k, $v ) = each %$hash ) {
        say "$k : $v";
    }

except that:

  • pairs can use an anonymous hash

  • can be used in streams

  • i hate the while syntax

consumers

now {actions} $xs

read the $xs stream and execute the {actions} block with the returned element as $_ until the $xs stream exhausts. it also returns the last transformed element so that it can be used to foldl.

(compare it to perl6 "eager" or haskell foldl)

fold $xs

returns the array of all the elements computed by $xs

    say join ',',      take 5, sub { state $x=-2; $x+=2 } # CODE(0x180bad8)
    say join ',', fold take 5, sub { state $x=-2; $x+=2 } # 0,2,4,6,8

see also unfold

nth $xs

returns the nth element of a stream

    say fold nth 5, sub { state $x=1; $x++ }
    # 5

chunksOf

non destructive splice alike (maybe best named as "traverse"? haskell name?). you can traverse an array by a group of copies of elements

    say "@$_" for fold chunksOf 3, ['a'..'f'];
    # a b c
    # d e f

Composers

concat @streams

concat takes a list of streams and returns them as a unique one:

    concat map { unfold [split //] } split /\s*/;

streams every chars of the words of the text

concatC $stream_of_streams

takes a stream of streams $stream_of_streams and expose them as a single one. A stream of streams is a steam that returns streams.

    concatC { take 3, range $_ } lines $fh

take 3 elements from the range started by the values of $fh, so if $fh contains (5,10), the stream is (5,6,7,10,11,12)

concatM $apply, $stream

applying $apply on each iterations of $stream must return a new stream. concatM expose them as a single stream.

    # ls is a generator for a glob

    sub cat { concatM {lines} ls shift }
    cat "/tmp/*.conf"

Perlude companions

some modules comes with generators so they are perfect Perlude companions (send me an exemple if yours does too).

Path::Iterator::Rule

    use aliased qw(Path::Iterator::Rule find);

    now {print}
        take 3,
        find->new
        -> file
        -> size('>1k')
        -> and( sub { -r } )
        -> iter(qw( /tmp ));

you can use filter instead of and:

    now {print}
        take 3,
        filter {-r}
        find->new
        -> file
        -> size('>1k')
        -> iter(qw( /tmp )); 

Path::Tiny

    use Path::Tiny;

    now {print} take 3, path("/etc")->iterator;  

    now {print}
        take 3,
        apply {chomp;$_}
        records path("/etc/passwd")->openr_utf8( {qw( locked 1 )});

curry

a very friendly way to write iterators. i rewrote the exemple from the TAP::Parser doc:

    use TAP::Parser;
     
    my $parser = TAP::Parser->new( { tap => $output } );
     
    while ( my $result = $parser->next ) {
        print $result->as_string;
    }

with Perlude

    now {print $_->as_string."\n"} do {
        my $parser =
            TAP::Parser
            -> new( { tap => path("/tmp/x")->slurp });
        sub { $parser->next // () }
    }

with Perlude and curry

    now {defined and print $_->as_string."\n"}
        TAP::Parser
        -> new( { tap => path("/tmp/x")->slurp })
        -> curry::next;

TODO / CONTRIBUTONS

feedbacks and contributions are very welcome

    http://github.com/eiro/p5-perlude
  • Improve general quality: doc, have a look on http://cpants.cpanauthors.org/dist/perlude, https://metacpan.org/pod/Devel::Cover.

  • Explore test suite to know what isn't well tested. find bugs :)

        * see range implementation # what if step 0 ? 
        * pairs must support streams and array
        * provide an alternative to takeWhile to return the combo breaker
        * explore AST manipulations for futher optimizations 
  • deprecate open_file and lines (or/and find a companion) as it is out of the scope of Perlude and open_file seems scary (anything to avoid the open prototype?).

  • reboot Perl::builtins

    remove the hardcoded f namespace and use use aliased instead.

  • ask for BooK and Dolmen if they mind to remove Perlude::Lazy as noone seems to use it anymore.

  • Perlude::XS anyone ?

  • Something to revert the callback mechanism: how to provide a generic syntax to use Anyevent driven streams or "callback to closures" (for example: Net::LDAP callback to treat entries onfly)

  • provide streamers for common sources CSV, LDAP, DBI (see p5-csv-stream)

KNOWN BUGS

not anymore, if you find one, please email bug-Perlude [at] rt.cpan.org.

AUTHORS

  • Philippe Bruhat (BooK)

  • Marc Chantreux (eiro)

  • Olivier Mengué (dolmen)

CONTRIBUTORS

Burak Gürsoy (cpanization)

ACKNOWLEDGMENTS

  • Thanks to Nicolas Pouillard and Valentin (#haskell-fr), i leanrt a lot about streams, lazyness, lists and so on. Lazyness.pm was my first attempt.

  • The name "Perlude" is an idea from Germain Maurice, the amazing sysop of http://linkfluence.com back to early 2010.

  • Former versions of Perlude used undef as stream terminator. After my talk at the French Perl Workshop 2011, dolmen suggested to use () as stream terminator, which makes sense not only because undef is a value but also because () is the perfect semantic to end a stream. So Book, Dolmen and myself rewrote the entire module from scratch in the hall of the hotel with a bottle of chartreuse and Cognominal.

    We also tried some experiments about real lazyness, memoization and so on. it becomes clear now that this is hell to implement correctly: use perl6 instead :)

    I was drunk and and mispelled Perlude as "Perl dude" so Cognominal collected some quotes of "The Big Lebowski" and we called ourselves "the Perl Dudes". This is way my best remember of peer programming and one of the best moment i shared with my friends mongueurs.