Question

Background:

I whish to create a distribution that contains a few modules for personal use — it will never see the light of day (aka. CPAN). I would like to start these modules using Module::Starter to leverage easy testing, and version management.

(perl 5.12.4 on Ubuntu GNU/Linux)

Question:

How do I use Unicode filenames with Module::Starter, e.g. Local::Λ ← (Lambda)?

This doesn’t work:

$ module-starter --module Local::Λ
Invalid module name: Local::Λ at /usr/local/share/perl/5.12.4/Module/Starter/App.pm line 132.

The line that actually fails is

croak "Invalid module name: $_" unless /\A[a-z_]\w*(?:::[\w]+)*\Z/i;

in Module::Starter::Simple (line 95).

The Lambda clearly is a "word" character.

This works, by the way:

$ cat > xΛ.pm
use utf8;
package xΛ;
sub foo { print "42\n" }
1;
^D
$ perl -Mutf8 -MxΛ -e 'xΛ::foo()'
42
$ 

Things I have not yet tried:

• spending a day compiling perl 5.16 for more unicode goodness.
• hacking away at the source, and enabling anything remotely connected to unicode.
• Checking that this isn’t just a restriction on distribution names (which I am not concerned about, as CPAN will never see this).
• Checking if this could be a Getopt::Long problem.

Answer 1

The definition of ‘word character’ changes with: a) your version of perl. b) If the script in question is using your input as a binary string or a character string. Check out the /a & /u flags of recent versions of perl’s regex engine.

FYI: Because the Perl folks strive for cross-platform compatibility and there is widely varying support for Unicode in various filesystems, they’ve decided not to support Unicode Module Names at this time.

---

Resolution from Amon’s Comment:
this did the trick:

perl -Mutf8 -M5.012 -MModule::Starter::App -E'$ARGV[1]="Local::\x{039B}";Module::Starter::App->run' -- --module Local::Λ --license gpl3 --ignore git

(Using \x{...} escapes forces unicode, according to perlre).