Re^3: Module Announcement: Perl-Critic-1.01

[Rhandom]

You are correct. I can't argue that that is what happens when you return undef.

Now - the larger question is, who is still designing Perl module interfaces that return non-scalar results.

I can't remember the last time I designed a function to return a list of values - much less variably return a list or a single scalar depending upon context. To me it is a code smell to not return a single scalar value (the single value could be arrayref or hashref - but it is still a single value). I know that may not sound Perlish - but as a user of the modules I produce I prefer not have to guess what context I need to call items in. It also saves memory to return arrayrefs or hashrefs.

Yes I still use map and grep and caller and times and stat, but they are some of the last things I use that return lists. Perl 6 will fix some of these issues.

[Jenda]

So I guess you'd rather do

my $pointless_reference = $obj->getPosition();
my ($x, $y) = @$pointless_reference[0,1];

or

my ($x, $y) = @{$obj->getPosition()};

instead of

my ($x, $y) = $obj->getPosition();

right? Restricting myself to returning just one value just because most languages do not allow anything more seems silly to me. Especially compared to such dirty tricks as updating the values of some variables passed by reference or something.

To me bending backwards to always return a single scalar is a code smell. A code smell sugesting that the author writes C or some other language, but definitely not Perl. Even though the code is full of sigils and there are regexps scattered around.

[Rhandom]

Absolutely!

I think that the following is perfectly fine.


my ($x, $y) = @{ $obj->getPosition }

Depending upon the interface I think that it may be better to return a point reference or object rather than the xy coordinates themselves.


my $point = $obj->getPosition;
do_something_with_point($point);
printf "X is %.2f\n", $point->x;
# or maybe $point->[0] - possibly bad
# or maybe $point->{x} - probably bad
# but really $point->x is the best


It is funny that you should use a geometry example. Many of the C or C++ libraries that I have seen that deal with graphics primitives prefer to pass around a point struct rather than the individual x or y.

Just because it is possible to make interchanged data as terse as possible doesn't mean it is the right thing to do.

[Jenda]

C or C++ cannot do anything better anyway. I think we can't convince each other so let's just hope we do not meet in one company and on the same project since this is one of the seemingly unimportant style issues that drive people crazy ;-)