NAME Exception::Class::TryCatch - Syntactic try/catch sugar for use with Exception::Class SYNOPSIS use Exception::Class::TryCatch; # simple usage of catch() eval { Exception::Class::Base->throw('error') }; catch my $err and warn $err->error; # catching only certain types or else rethrowing eval { Exception::Class::Base::SubClass->throw('error') }; catch( my $err, ['Exception::Class::Base', 'Other::Exception'] ) and warn $err->error; # caught() is a synonym for catch() eval { Exception::Class::Base->throw('error') }; if ( caught my $err ) { $err->isa('this') and do { handle_this($err) }; $err->isa('that') and do { handle_that($err) }; } # use "try eval" to push exceptions onto a stack to catch later try eval { Exception::Class::Base->throw('error') } and do { # cleanup that might use "try/catch" again }; catch my $err; DESCRIPTION Exception::Class::TryCatch provides syntactic sugar for use with Exception::Class using the familiar keywords "try" and "catch". Its primary objective is to allow users to avoid dealing directly with $@ by ensuring that any exceptions caught in an "eval" are captured as Exception::Class objects, whether they were thrown objects to begin with or whether the error resulted from "die". This means that users may immediately use "isa" and various Exception::Class methods to process the exception. In addition, this module provides for a method to push errors onto a hidden error stack immediately after an "eval" so that cleanup code or other error handling may also call "eval" without the original error in $@ being lost. Inspiration for this module is due in part to Dave Rolsky's article "Exception Handling in Perl With Exception::Class" in *The Perl Journal* (Rolsky 2004). The "try/catch" syntax used in this module does not use code reference prototypes the way the Error.pm module does, but simply provides some helpful functionality when used in combination with "eval". As a result, it avoids the complexity and dangers involving nested closures and memory leaks inherent in Error.pm (Perrin 2003). Rolsky (2004) notes that these memory leaks may not occur in recent versions of Perl, but the approach used in Exception::Class::TryCatch should be safe for all versions of Perl as it leaves all code execution to the "eval" in the current scope, avoiding closures altogether. USAGE "catch, caught" # zero argument form my $err = catch; # one argument forms catch my $err; my $err = catch( [ 'Exception::Type', 'Exception::Other::Type' ] ); # two argument form catch my $err, [ 'Exception::Type', 'Exception::Other::Type' ]; Returns an "Exception::Class::Base" object (or an object which is a subclass of it) if an exception has been caught by "eval" or else returns "undef" if no error exists. The exception is either popped from a hidden error stack (see "try") or, if the stack is empty, taken from the current value of $@. If the exception is not an "Exception::Class::Base" object (or subclass object), an "Exception::Class::Base" object will be created using the string contents of the exception. This means that calls to "die" will be wrapped and may be treated as exception objects. Other objects caught will be stringfied and wrapped likewise. Such wrapping will likely result in confusing stack traces and the like, so any methods other than "error" used on "Exception::Class::Base" objects caught should be used with caution. "catch" is prototyped to take up to two optional scalar arguments. The single argument form has two variations. * If the argument is a reference to an array, any exception caught that is not of the same type (or a subtype) of one of the classes listed in the array will be rethrown. * If the argument is not a reference to an array, "catch" will set the argument to the same value that is returned. This allows for the "catch my $err" idiom without parentheses. In the two-argument form, the first argument is set to the same value as is returned. The second argument must be an array reference and is handled the same as as for the single argument version with an array reference, as given above. "caught" is a synonym for "catch" for syntactic convenience. "try" try eval { #dangerous code }; catch my $err; $rv = try eval { return $scalar }; @rv = try [ eval { return @array } ]; Pushes the current error ($@) onto a hidden error stack for later use by "catch". "try" uses a prototype that expects a single scalar so that it can be used with eval without parentheses. As "eval { BLOCK }" is an argument to try, it will be evaluated just prior to "try", ensuring that "try" captures the correct error status. "try" does not itself handle any errors -- it merely records the results of "eval". "try { BLOCK }" will be interpreted as passing a hash reference and will (probably) not compile. (And if it does, it will result in very unexpected behavior.) Since "try" requires a single argument, "eval" will normally be called in scalar context. To use "eval" in list context with "try", put the call to "eval" in an anonymous array: @rv = try [ eval {return @array} ]; When "try" is called in list context, if the argument to "try" is an array reference, "try" will dereference the array and return the resulting list, In scalar context, of course, "try" passes through the scalar value returned by "eval" -- even if that is an array reference -- without modifications. When using "try" in a compound idioms like the following, be sure that the "eval" returns a true value: try eval { # code 1; } and do { # cleanup }; catch my $err; "try" must always be properly bracketed with a matching "catch" or unexpected behavior may result when "catch" pops the error off of the stack. "try" executes right after its "eval", so inconsistent usage of "try" like the following will work as expected: try eval { eval { die "inner" }; catch my $inner_err die "outer" if $inner_err; }; catch my $outer_err; # handle $outer_err; However, the following code is a problem: # BAD EXAMPLE try eval { try eval { die "inner" }; die $@ if $@; }; catch my $outer_err; # handle $outer_err; This code will appear to run correctly, but "catch" gets the exception from the inner "try", not the outer one, and there will still be an exception on the error stack which will be caught by the next "catch" in the program, causing unexpected (and likely hard to track) behavior. In short, if you use "try", you must "catch". The problem code above should be rewritten as: try eval { try eval { die "inner" }; catch my $inner_err; $inner_err->rethrow if $inner_err; }; catch my $outer_err; # handle $outer_err; REFERENCES 1. perrin. (2003), "Re: Re2: Learning how to use the Error module by example", (perlmonks.org), Available: http://www.perlmonks.org/index.pl?node_id=278900 (Accessed September 8, 2004). 2. Rolsky, D. (2004), "Exception Handling in Perl with Exception::Class", *The Perl Journal*, vol. 8, no. 7, pp. 9-13 SEE ALSO - Exception::Class - Test::Exception - Error [but see (Perrin 2003) before using] INSTALLATION To install this module, type the following: perl Build.PL ./Build ./Build test ./Build install BUGS Though this is a simple module, it may contain bugs or have unexpected behaviors. Please report bugs using the CPAN Request Tracker at http://rt.cpan.org/NoAuth/Bugs.html?Dist=Exception-Class-TryCatch AUTHOR David A Golden http://dagolden.com/ COPYRIGHT Copyright (c) 2004, 2005 by David A. Golden This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself. The full text of the license can be found in the LICENSE file included with this module.