require_ok "open" fails

[ikegami]

require_ok "open" fails by trying to use the open operator.

This can be resolved by replacing

    # Try to determine if we've been given a module name or file.
    # Module names must be barewords, files not.
    $module = qq['$module'] unless _is_module_name($module);
 
    my $code = <<REQUIRE;
package $pack;
require $module;
1;
REQUIRE

with

    # Try to determine if we've been given a module name or file.
    # Module names must be barewords, files not.
    my $code = ( _is_module_name($module)
        ? "require '$module'"
        : "use $module ()"
    );
 
    $code = <<REQUIRE;
package $pack;
$code;
1;
REQUIRE

[exodist]

Unfortunately this change could break things. There is a lot of code on cpan that overrides require, any code testing a require override using require_ok could potentially break if use does not call the override (I am not sure it would, though putting a sub in @inc should still work).

This is another affirmation of the Test2 decision not to implement use_ok or require_ok.

I think for back-compatibility of Test::More we should simply document this limitation and move on, the fix may be more problematic than the problem being solved.

I will leave this ticket open for a while to solicit feedback from you and anyone else who wants to weigh-in on the pros and cons. If anyone has the capacity right now to look into the effects on a require-override and/or verify against a large chunk of cpan with such a change in place I would welcome that, I do not have the time at the moment. If this fix can be proven safe for back-compatibility I will consider it.

[haarg]

use will also trigger require overrides, but it would change the error messages somewhat.

A better approach would probably be to translate module names into "Module/Name.pm" form before requiring them.

[exodist]

What is the current recommended way to do that? I usually do the naive $mod =~ s{::}{/}g; $mod .= ".pm" change. But I am aware that this could be wrong on non-unix systems, and has some potential issues with unicode sequences. Is the naive approach sufficient, or is there a core module I can depend on (with sufficient version back to 5.8) that has a working implementation? Or do I need to copy a "correct" implementation and inline it?

[haarg]

$mod =~ s{::}{/}g; $mod .= ".pm" is the correct approach, including on non-unix systems. The only thing extra done by 'recommended' modules like Module::Runtime is to validate that it looks like a module name first. In the case of Test::More, that's already being done with _is_module_name.