Re: How to iterate through a PDL piddle?

[zentara]

You need a dog, not puppies. :-) Does this work?

#!/usr/bin/perl
use strict;
use warnings;
use PDL::LiteF;
#dog dosn't work on huge values

my $u = pdl [1,2,3,4];
my $dim = 4;
print join "\t", $u->dog;
print "\n";

[arrac]

Thanks @zentara. It works! I can't believe that it works though. There should be a more obvious way like  $u->list which gives an array reference. Or a custom iterator on a piddle which can replace  foreach.

[etj]

If you feel the urge to use an iterator in PDL, quite often that is a sign you could instead use "array programming" type constructs (see https://en.wikipedia.org/wiki/Array_programming), which lets all the processing be done inside the fast C loops, rather than going backwards and forwards between Perl and PDL.

[dkogan]

Errr..... there IS a 'list' function $u->list acts the same as $u->dog except list breaks the piddle up into individual elements while dog breaks it up into N-1-dimensional slices. In this particular case, where $u is one-dimensional, $u->list is identical to $u->dog

[etj]

• dog returns a list of PDL slices, as Dima says, so a 2,3,4 ndarray dog-ed would return 4 ndarrays each dimensioned 2,3 (and as they're "slices", updates to them would be reflected in the original ndarray)
• list returns the ndarray flattened (so the whole thing in a single flat list) as Perl scalars, so a 2,3,4 would be a list of 24 Perl scalars
• unpdl (as of 2.006 from 2013, as provided by the mighty Joel Berger), which returns the ndarray as an array-ref with the same (possibly nested) structure as the input ndarray.