Re^4: Abusing Map

[AnomalousMonk]

Never do in a loop, what can be done outside of it.

I agree, and that's why I like, in particular, the for-loop and, less so, map (and let's just pass over the grep) versions that iterate over ranges of 1 .. $#array or 0 .. $#array-1 and so will never enter the loop. (Again, I haven't actually tested all the preceding code in this thread, only stared at it, but it looks ok.)

Wouldn't your version fail for an array with one element?

That depends on the meaning of "fail."

c:\@Work\Perl\monks>perl -wMstrict -MData::Dump -le
"my @ra = (1);
 my $i = $#ra;
 ;;
 my @rb;
 $rb[ $i ] = $ra[ $i ] + $ra[ --$i ] while $i > 0;
 dd \@rb
"
[]

But maybe an operation that iterates pairwise over an array with less than a pair of elements should throw an exception or at least issue a warning. Who can say? Only the Great Specificator above.

[BrowserUk]

I agree, and that's why I like, in particular, the for-loop and, less so, map (and let's just pass over the grep) versions that iterate over ranges of 1 .. $#array or 0 .. $#array-1 and so will never enter the loop

The problem with the map/grep and postfix for is that they generate a list as big as the array.

Small lists: meh! But for big lists -- where big frequently is in the 100s of millions on my machine and billions on client/AWS X1 instances -- those lists cost big time.

Postfix while and until avoid that whilst still giving clean, concise code. Combined with each, its good for big hashes also.

[AnomalousMonk]

... for big lists ...

I agree about the time/space advantages, sometimes life-saving advantages, of in-place mutation for big lists.

... postfix for ... generate[s] a list as big as the array.

I thought a for-loop, postfix or otherwise, was guaranteed to have a space complexity (if that's the correct term) of zero | 1 (per ikegami here) — if it's written right! That, e.g., either
    for (@humongous_array) { $_ = $_ + $_ }
or
    $_ = $_ + $_ for @humongous_array;
are guaranteed to be just fine (if @humongous_array could fit in memory in the first place), whereas
    @humongous_array = map $_ + $_, @humongous_array;
can kill you in terms of both time and space. In fact, isn't it the case that beginning with Perl version 5.mumble, map called in void context produces no intermediate list/array, so even
    map $_ = $_ + $_, @humongous_array;
should be perfectly safe (although for aesthetic reasons, I'm not a fan of this particular usage)?

And let's just not talk about Fight Club | grep anymore.

[BrowserUk]

The problem comes when you need to build a list of indexes as with the OPs code:

sub checkMem{ `tasklist /nh /fi "pid eq $_[0]"`; };;

print checkMem( $$ );;
 perl.exe                      3040 Console                    1      
+9,544 K

$#a = 1e6;;
print checkMem( $$ );;
 perl.exe                      3040 Console                    1     1
+7,428 K

map{ $_ } 0 .. $#a;;
print checkMem( $$ );;
 perl.exe                      1224 Console                    1     7
+3,272 K           ### 56MB to build the list of 1e6 indices ###

[ikegami]

I thought a for-loop, postfix or otherwise, was guaranteed to have a space complexity (it that's the correct term) of zero

Of 1 (constant).