++ed by:

8 PAUSE users

Marc Chantreux
and 1 contributors

Write your own generators

Have you ever miss the shell pipe in Perl? Such a clever operator: it streams data from programs to programs on demand, which means that nothing will compute more than expected by the whole stream, any part the pipe can stop the whole stream.

    seq 1000

will compute 1000 lines

    seq 1000 | sed 5q

will compute only 5 lines as sed ends its job there. We somehow miss it in perl. Sure, we have grep and map but they are only acting in complete arrays.

Perlude comes with all most common filters, i doubt you'll missing one. If you do so: please feedback and i'll probably add it! So your job is about writing generators (or using those written in Perlude::Stuff and Perlude::Sh). If you have to write a generic one, please contribute.

To write them, you have to understand the Perlude conventions. When you're using an iterator, say range 1,4, you have to think about the whole list of potential results. Those can be written as

    ( 1, 2, 3, 4     )
    ( 1, 2, 3, 4,    )
    ( 1, 2, 3, 4, () )

so () is used as a list terminator. your iterator must return one scalar by call and last it work sending a terminator. As exemple:

    sub read_file {
        open my $fh, shift; 
        sub { <$fh> // () }

now you can write

    now {say}
        filter {/foo/}
        take 5,
        read_file "test.txt"

which is equivalent to

    sed 5q test.txt | grep foo 

Not only it's easy to read and write, its behaviour is also the best expected:

  • it reads one record, use it and forget it before reading the next record. This is a memory friendly behavior

  • whoever in the pipe can decide to stop it. For example: it's fine for grep to release 3 records only.

Writing unix filters is really easy. Also note that filters/generators compositions rules are simple and powerfull

    G | F => G 
    F | F => F

If you wrote shell, powershell, perl6 or any other functionnal langage, you probably miss it coming back to perl5.

Basically, on demand lists are just iterators. Perlude is just a pleasant way to deal with them stealing keywords from haskell Perlude.


As example: What are the 5 first naturals containing a 3?

A perl implementation would be:

    ( my $_=0, my $count=0
    ; $count <= 5
    ; $_++ )
    { if (/3/) { $count++; say } }

Hard to read ... and worth: nothing is reusable at all

The shell counterpart would be

    nat () { while {true} { print $[i++] } }
    nat | grep 3 | head -n5

There are things to understand about the shell elegance:

  • there is no need of a counter variable, neither a for loop: head is the single command which handles it for you.

  • the implementation of nat is bare simple: you just focus on your nat problem, you don't care how many elements the filter could need.

  • you added nat to your toolkit, it's much more pain to resuse it in perl ... before Perlude

also, it's easy to create a new function 'top5' by passsing a an argument to head (looks like a partial application):

    top5 () { head -n5 }
    contains3 () { grep 3 }
    nat | contains3 | top5

No perl builtin provide this power.

I can haz nat in perl ?

Perlude is a set of functions that takes closures as arguments, and returns others

nat is the basic closure exemple:

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

a reusable way to write it would be:

    sub nat_from {
        my $x = shift;
        sub { $x++ }

    sub nat { nat_from 0 }

now you can use Perlude keywords on this functions

    sub evens_in { filter { not( $_ % 2 ) } shift }
    sub top5 { take 5, shift }

Other dynamic langages stuff

    Ruby       : Rubylude was written by Nono after RMLL'11 https://github.com/nono/Rubylude
    Javascript :
        http://weepy.github.com/kaffeine/ was quoted it the python pipe intro but i guess it's useless
        as http://jashkenas.github.com/coffee-script/ is javascript made right.
    Python     : https://github.com/JulienPalard/Pipe with an introduction here: http://dev-tricks.net/pipe-infix-syntax-for-python