Re^2: using a variable to specify which perl module to use (bareword)

require "Text::CSV" is much different than require Text::CSV. For your code to work, you'd need to s-::-/-g on $module for the require (but not for the import).

You can also use a lexical. Note that my declarations happen when the statement is compiled but my initializations happen when the statement is run. So:

use strict;

my $module;
BEGIN {
    $module= ... ? 'Text::CSV' : 'Text::CSV_XS';
    my $file= $module;
    $file =~ s-::-/-g;
    require "$file.pm";
    $module->import();
}

my $csv = $module->new();
[download]

Using eval "use $module" avoids this complication. Which to prefer is mostly a matter of taste (unless there is "risk" of $module ending up containing something strange).

- tye

Comment on Re^2: using a variable to specify which perl module to use (bareword) Select or Download Code

Replies are listed 'Best First'.
Re: Re^2: using a variable to specify which perl module to use (bareword) by William G. Davis (Friar) on Jul 15, 2003 at 18:45 UTC
Actually, no, it's not a matter of taste. It's best to avoid eval STRING whenever possible since it forces dynamic runtime recompilation, and is thus rather slow. An eval BLOCK is always preferable. If you can represent the package names as literals (barewords) within your program, use this: `if (something()) { eval { require Text::CSV; Text::CSV->import; } } else { eval { require Text::CSV_XS; Text::CSV_XS->import; } }` [download] The eval BLOCK syntax as well allows you to catch any fatal errors, storing them in $@: `die "Couldn't load module: $@" if ($@);` [download]	[reply] [d/l] [select]
Re^4: using a variable to specify which perl module to use (bareword) by tye (Sage) on Jul 15, 2003 at 19:14 UTC
It's best to avoid eval STRING whenever possible since it forces runtime recompilation, and is thus rather slow. Avoiding string eval is generally good advice. Always avoiding it just for speed reasons is premature micro optimization. If you were doing this in a loop and profiling showed that it was actually causing you a noticeable slow-down, then that would be a good reason to refactor the code to avoid it. My taste leads me to prefer much simpler code that runs a few milliseconds slower nearly every time. You left out the code to set what class to use when creating the object(s). If you add that in, then you are having to say each package name 3 times. I'd rather use the code I provided above. (: Thanks for mentioning the fatal error trapping. In this situation, eval's trapping of errors is a disadvantage. I should have mentioned that the extra step would be required if the "simpler" `eval "use $module"` were used. I'd just `die "$@\n" if $@;` so that the error message is as expected. Also note that you need to put the die code inside the BEGIN block in your code above. - tye	[reply] [d/l] [select]
Re: Re^2: using a variable to specify which perl module to use (bareword) by broquaint (Abbot) on Jul 15, 2003 at 21:51 UTC
require "Text::CSV" is much different than require Text::CSV Ah yes, of course. I've written code to do the appropriate munging of a bareword many times, don't know why I forgot it this time. Anyhow my code should really look like this `BEGIN { (my $file = ($module = somecond() ? "Text::CSV" : "Text::CSV_XS")".pm") =~ s{::}(/)g; require $file; $module->import; }` [download] HTH `_________ broquaint`	[reply] [d/l]