-- Perl 5.8.7 documentation --
Switch

NAME

Switch - A switch statement for Perl

VERSION

This document describes version 2.10 of Switch, released Dec 29, 2003.

SYNOPSIS

	use Switch;
	switch ($val) {
		case 1		{ print "number 1" }
		case "a"	{ print "string a" }
		case [1..10,42]	{ print "number in list" }
		case (@array)	{ print "number in list" }
		case /\w+/	{ print "pattern" }
		case qr/\w+/	{ print "pattern" }
		case (%hash)	{ print "entry in hash" }
		case (\%hash)	{ print "entry in hash" }
		case (\&sub)	{ print "arg to subroutine" }
		else		{ print "previous case not true" }
	}

BACKGROUND

[Skip ahead to "DESCRIPTION" if you don't care about the whys and wherefores of this control structure]

In seeking to devise a "Swiss Army" case mechanism suitable for Perl, it is useful to generalize this notion of distributed conditional testing as far as possible. Specifically, the concept of "matching" between the switch value and the various case values need not be restricted to numeric (or string or referential) equality, as it is in other languages. Indeed, as Table 1 illustrates, Perl offers at least eighteen different ways in which two values could generate a match.

	Table 1: Matching a switch value ($s) with a case value ($c)
        Switch  Case    Type of Match Implied   Matching Code
        Value   Value   
        ======  =====   =====================   =============
        number  same    numeric or referential  match if $s == $c;
        or ref          equality
	object  method	result of method call   match if $s->$c();
	ref     name 				match if defined $s->$c();
		or ref
        other   other   string equality         match if $s eq $c;
        non-ref non-ref
        scalar  scalar
        string  regexp  pattern match           match if $s =~ /$c/;
        array   scalar  array entry existence   match if 0<=$c && $c<@$s;
        ref             array entry definition  match if defined $s->[$c];
                        array entry truth       match if $s->[$c];
        array   array   array intersection      match if intersects(@$s, @$c);
        ref     ref     (apply this table to
                         all pairs of elements
                         $s->[$i] and
                         $c->[$j])
        array   regexp  array grep              match if grep /$c/, @$s;
        ref     
        hash    scalar  hash entry existence    match if exists $s->{$c};
        ref             hash entry definition   match if defined $s->{$c};
                        hash entry truth        match if $s->{$c};
        hash    regexp  hash grep               match if grep /$c/, keys %$s;
        ref     
        sub     scalar  return value defn       match if defined $s->($c);
        ref             return value truth      match if $s->($c);
        sub     array   return value defn       match if defined $s->(@$c);
        ref     ref     return value truth      match if $s->(@$c);

In reality, Table 1 covers 31 alternatives, because only the equality and intersection tests are commutative; in all other cases, the roles of the $s and $c variables could be reversed to produce a different test. For example, instead of testing a single hash for the existence of a series of keys (match if exists $s->{$c} ), one could test for the existence of a single key in a series of hashes (match if exists $c->{$s} ).

As perltodo observes, a Perl case mechanism must support all these "ways to do it".

DESCRIPTION

The Switch.pm module implements a generalized case mechanism that covers the numerous possible combinations of switch and case values described above.

The module augments the standard Perl syntax with two new control statements: switch and case . The switch statement takes a single scalar argument of any type, specified in parentheses. switch stores this value as the current switch value in a (localized) control variable. The value is followed by a block which may contain one or more Perl statements (including the case statement described below). The block is unconditionally executed once the switch value has been cached.

A case statement takes a single scalar argument (in mandatory parentheses if it's a variable; otherwise the parens are optional) and selects the appropriate type of matching between that argument and the current switch value. The type of matching used is determined by the respective types of the switch value and the case argument, as specified in Table 1. If the match is successful, the mandatory block associated with the case statement is executed.

In most other respects, the case statement is semantically identical to an if statement. For example, it can be followed by an else clause, and can be used as a postfix statement qualifier.

However, when a case block has been executed control is automatically transferred to the statement after the immediately enclosing switch block, rather than to the next statement within the block. In other words, the success of any case statement prevents other cases in the same scope from executing. But see "Allowing fall-through" below.

Together these two new statements provide a fully generalized case mechanism:

        use Switch;
        # AND LATER...
        %special = ( woohoo => 1,  d'oh => 1 );
        while (<>) {
            switch ($_) {
                case (%special) { print "homer\n"; }      # if $special{$_}
                case /a-z/i     { print "alpha\n"; }      # if $_ =~ /a-z/i
                case [1..9]     { print "small num\n"; }  # if $_ in [1..9]
                case { $_[0] >= 10 } {                    # if $_ >= 10
                    my $age = <>;
                    switch (sub{ $_[0] < $age } ) {
                        case 20  { print "teens\n"; }     # if 20 < $age
                        case 30  { print "twenties\n"; }  # if 30 < $age
                        else     { print "history\n"; }
                    }
                }
                print "must be punctuation\n" case /\W/;  # if $_ ~= /\W/
        }

Note that switch es can be nested within case (or any other) blocks, and a series of case statements can try different types of matches -- hash membership, pattern match, array intersection, simple equality, etc. -- against the same switch value.

The use of intersection tests against an array reference is particularly useful for aggregating integral cases:

        sub classify_digit
        {
                switch ($_[0]) { case 0            { return 'zero' }
                                 case [2,4,6,8]    { return 'even' }
                                 case [1,3,4,7,9]  { return 'odd' }
                                 case /[A-F]/i     { return 'hex' }
                               }
        }

Allowing fall-through

Fall-though (trying another case after one has already succeeded) is usually a Bad Idea in a switch statement. However, this is Perl, not a police state, so there is a way to do it, if you must.

If a case block executes an untargetted next, control is immediately transferred to the statement after the case statement (i.e. usually another case), rather than out of the surrounding switch block.

For example:

        switch ($val) {
                case 1      { handle_num_1(); next }    # and try next case...
                case "1"    { handle_str_1(); next }    # and try next case...
                case [0..9] { handle_num_any(); }       # and we're done
                case /\d/   { handle_dig_any(); next }  # and try next case...
                case /.*/   { handle_str_any(); next }  # and try next case...
        }

If $val held the number 1 , the above switch block would call the first three handle_... subroutines, jumping to the next case test each time it encountered a next. After the thrid case block was executed, control would jump to the end of the enclosing switch block.

On the other hand, if $val held 10 , then only the last two handle_... subroutines would be called.

Note that this mechanism allows the notion of conditional fall-through. For example:

        switch ($val) {
                case [0..9] { handle_num_any(); next if $val < 7; }
                case /\d/   { handle_dig_any(); }
        }

If an untargetted last statement is executed in a case block, this immediately transfers control out of the enclosing switch block (in other words, there is an implicit last at the end of each normal case block). Thus the previous example could also have been written:

        switch ($val) {
                case [0..9] { handle_num_any(); last if $val >= 7; next; }
                case /\d/   { handle_dig_any(); }
        }

Automating fall-through

In situations where case fall-through should be the norm, rather than an exception, an endless succession of terminal nexts is tedious and ugly. Hence, it is possible to reverse the default behaviour by specifying the string "fallthrough" when importing the module. For example, the following code is equivalent to the first example in "Allowing fall-through":

        use Switch 'fallthrough';
        switch ($val) {
                case 1      { handle_num_1(); }
                case "1"    { handle_str_1(); }
                case [0..9] { handle_num_any(); last }
                case /\d/   { handle_dig_any(); }
                case /.*/   { handle_str_any(); }
        }

Note the explicit use of a last to preserve the non-fall-through behaviour of the third case.

Alternative syntax

Perl 6 will provide a built-in switch statement with essentially the same semantics as those offered by Switch.pm, but with a different pair of keywords. In Perl 6 switch will be spelled given , and case will be pronounced when . In addition, the when statement will not require switch or case values to be parenthesized.

This future syntax is also (largely) available via the Switch.pm module, by importing it with the argument "Perl6" . For example:

        use Switch 'Perl6';
        given ($val) {
                when 1       { handle_num_1(); }
                when ($str1) { handle_str_1(); }
                when [0..9]  { handle_num_any(); last }
                when /\d/    { handle_dig_any(); }
                when /.*/    { handle_str_any(); }
                default      { handle anything else; }
        }

Note that scalars still need to be parenthesized, since they would be ambiguous in Perl 5.

Note too that you can mix and match both syntaxes by importing the module with:

	use Switch 'Perl5', 'Perl6';

Higher-order Operations

One situation in which switch and case do not provide a good substitute for a cascaded if, is where a switch value needs to be tested against a series of conditions. For example:

        sub beverage {
            switch (shift) {
                case sub { $_[0] < 10 }  { return 'milk' }
                case sub { $_[0] < 20 }  { return 'coke' }
                case sub { $_[0] < 30 }  { return 'beer' }
                case sub { $_[0] < 40 }  { return 'wine' }
                case sub { $_[0] < 50 }  { return 'malt' }
                case sub { $_[0] < 60 }  { return 'Moet' }
                else                     { return 'milk' }
            }
        }

The need to specify each condition as a subroutine block is tiresome. To overcome this, when importing Switch.pm, a special "placeholder" subroutine named __ [sic] may also be imported. This subroutine converts (almost) any expression in which it appears to a reference to a higher-order function. That is, the expression:

        use Switch '__';
        __ < 2 + __

is equivalent to:

        sub { $_[0] < 2 + $_[1] }

With __ , the previous ugly case statements can be rewritten:

        case  __ < 10  { return 'milk' }
        case  __ < 20  { return 'coke' }
        case  __ < 30  { return 'beer' }
        case  __ < 40  { return 'wine' }
        case  __ < 50  { return 'malt' }
        case  __ < 60  { return 'Moet' }
        else           { return 'milk' }

The __ subroutine makes extensive use of operator overloading to perform its magic. All operations involving __ are overloaded to produce an anonymous subroutine that implements a lazy version of the original operation.

The only problem is that operator overloading does not allow the boolean operators && and || to be overloaded. So a case statement like this:

        case  0 <= __ && __ < 10  { return 'digit' }  

doesn't act as expected, because when it is executed, it constructs two higher order subroutines and then treats the two resulting references as arguments to &&:

        sub { 0 <= $_[0] } && sub { $_[0] < 10 }

This boolean expression is inevitably true, since both references are non-false. Fortunately, the overloaded 'bool' operator catches this situation and flags it as a error.

DEPENDENCIES

The module is implemented using Filter::Util::Call and Text::Balanced and requires both these modules to be installed.

AUTHOR

Damian Conway (damian@conway.org). The maintainer of this module is now Rafael Garcia-Suarez (rgarciasuarez@free.fr).

BUGS

There are undoubtedly serious bugs lurking somewhere in code this funky :-) Bug reports and other feedback are most welcome.

LIMITATIONS

Due to the heuristic nature of Switch.pm's source parsing, the presence of regexes specified with raw ?...? delimiters may cause mysterious errors. The workaround is to use m?...? instead.

Due to the way source filters work in Perl, you can't use Switch inside an string eval.

If your source file is longer then 1 million characters and you have a switch statement that crosses the 1 million (or 2 million, etc.) character boundary you will get mysterious errors. The workaround is to use smaller source files.

COPYRIGHT

    Copyright (c) 1997-2003, Damian Conway. All Rights Reserved.
    This module is free software. It may be used, redistributed
        and/or modified under the same terms as Perl itself.