AUTOLOAD, overload, and Roman numerals for Perl

ferreira has asked for the wisdom of the Perl Monks concerning the following question:

I just released to CPAN a distribution named Acme-Roman. (It's propagating yet, so maybe you can't reach immediately from your favourite CPAN mirror.) As an Acme module, it is a simple experience to allow using roman numerals in Perl programs in a straightforward manner.

    use Acme::Roman;

    print I + II; # III, as expected
[download]

That was done with a bit of overload, AUTOLOAD and a part which I am not satisfied with.

Some digression first. The idea came to me a long time ago from the solution to a Ruby quiz, namely

Ruby Quiz - Roman Numerals (#22)
http://rubyquiz.com/quiz22.html

(See the hightlighted solution at the Quiz Summary in the same page.) The trick in Ruby was done by defining strategic methods method_missing (which is kind of Perl AUTOLOAD) and const_missing. The Perl solution was quite natural, and I got to the point where

    use Acme::Roman;

    print I() + II(); # III
[download]

worked fine. But when I left the parentheses, it only barfed at me with

    $ perl -w -Ilib -MAcme::Roman -e 'my $x = I+II'
    Argument "II" isn't numeric in addition (+) at -e line 1.
    Argument "I" isn't numeric in addition (+) at -e line 1.
[download]

and

    $ perl -w -Ilib -MAcme::Roman -e 'print I+II'
    Name "main::I" used only once: possible typo at -e line 1.
    print() on unopened filehandle I at -e line 1.
[download]

which revealed the barewords were not resolved with the provided AUTOLOAD. I learned that if I provided an empty protoype it worked. And that's how I coded -- but that kind of sucks: I need to define the prototype of 3999 subs (I .. MMMCMXCIX) to make it work. I think that's too expensive. (Who would like to play with an Acme module which is as heavy as a useful module? ;-)

The Ruby solution is helped by the availability of const_missing method functionality. The equivalent in Perl would be to overload barewords, but overload documentation says:

Barewords are not covered by overloaded string constants.

Does anyone have any suggestion to make this lazier? Or a different way to implement that?

Comment on AUTOLOAD, overload, and Roman numerals for Perl Select or Download Code

Replies are listed 'Best First'.
Re: AUTOLOAD, overload, and Roman numerals for Perl by ikegami (Patriarch) on Jun 29, 2007 at 15:59 UTC
Barewords are autoquoted (`BAREWORD -> 'BAREWORD'`), except in two situations. 1) If a function was already declared with that name, it will be treated as a function name. 2) If a glob is expected (such as `print`'s first argument), then it will be treated as a glob (`BAREWORD -> *BAREWORD`). A bareword is only treated as a function name when a function with that name has already been declared, so `AUTOLOAD` won't work. Autoquoting of barewords can't be overriden by overload, so overload won't help. Looks to me like you're stuck using `&I`/`I()` or a source filter. Note: `use strict` will generate an error when bareword autoquoting occurs outside of a hash index.	[reply] [d/l] [select]
Re: AUTOLOAD, overload, and Roman numerals for Perl by shmem (Chancellor) on Jun 29, 2007 at 17:57 UTC
Heh. I would say that the only decent way to write an Acme::Roman module is in terms of Lingua::Romana::Perligata - it's to say: in latin... :-) --shmem _($_=" "x(1<<5)."?\n".q·/)Oo. G°\ / /\_¯/(q / ---------------------------- \__(m.====·.(_("always off the crowd"))."· ");sub _{s./.($e="'Itrs `mnsgdq Gdbj O`qkdq")=~y/"-y/#-z/;$e.e && print}	[reply]
Re: AUTOLOAD, overload, and Roman numerals for Perl by mrpeabody (Friar) on Jun 30, 2007 at 03:16 UTC
Source filter?	[reply]


more useful options
	PerlMonks