Provided by: libsyntax-keyword-match-perl_0.13-2build3_amd64 
NAME
"Syntax::Keyword::Match" - a "match/case" syntax for perl
SYNOPSIS
use v5.14;
use Syntax::Keyword::Match;
my $n = ...;
match($n : ==) {
case(1) { say "It's one" }
case(2) { say "It's two" }
case(3) { say "It's three" }
case(4), case(5)
{ say "It's four or five" }
case if($n < 10)
{ say "It's less than ten" }
default { say "It's something else" }
}
DESCRIPTION
This module provides a syntax plugin that implements a control-flow block called "match/case", which
executes at most one of a choice of different blocks depending on the value of its controlling
expression.
This is similar to C's "switch/case" syntax (copied into many other languages), or syntax provided by
Switch::Plain.
This is an initial, experimental implementation. Furthermore, it is built as a non-trivial example use-
case on top of XS::Parse::Keyword, which is also experimental. No API or compatibility guarantees are
made at this time.
Experimental Features
Some of the features of this module are currently marked as experimental (even within the context that
the module itself is experimental). They will provoke warnings in the "experimental" category, unless
silenced.
use Syntax::Keyword::Match qw( match :experimental(dispatch) );
use Syntax::Keyword::Match qw( match :experimental ); # all of the above
KEYWORDS
match
match( EXPR : OP ) {
...
}
A "match" statement provides the controlling expression, comparison operator, and sequence of "case"
statements for a match operation. The expression is evaluated to yield a scalar value, which is then
compared, using the comparison operator, against each of the "case" labels in the order they are written,
topmost first. If a match is found then the body of the labelled block is executed. If no label matches
but a "default" block is present, that will be executed instead. After a single inner block has been
executed, no further tests are performed and execution continues from the statement following the "match"
statement.
The braces following the "match" block must only contain "case" or "default" statements. Arbitrary code
is not supported here.
Even though a "match" statement is a full statement and not an expression, it can still yield a value if
it appears as the final statment in its containing "sub" or "do" block. For example:
my $result = do {
match( $topic : == ) {
case(1) { ... }
}
};
If the controlling expression introduces a new variable, that variable will be visible within any of the
"case" blocks, and will go out of scope after the "match" statement finishes. This may be useful for
temporarily storing the result of a more complex expression.
match( my $x = some_function_call() : == ) {
case ...
}
Comparison Operators
The comparison operator must be either "eq" (to compare cases as strings) or "==" (to compare them as
numbers), or "=~" (to compare cases using regexps).
Since version 0.11 on any Perl release, or previous versions on Perl releases 5.32 onwards, the "isa"
operator is also supported, allowing dispatch based on what type of object the controlling expression
gives.
match( $obj : isa ) {
case(A::Package) { ... }
case(Another::Package) { ... }
}
Remember that comparisons are made in the order they are written, from the top downwards. Therefore, if
you list a derived class as well as a base class, make sure to put the derived class before the base
class, or instances of that type will also match the base class "case" block and the derived one will
never match.
class TheBase {}
class Derived :isa(TheBase) {}
match( $obj : isa ) {
case(TheBase) { ... }
case(Derived) {
# This case will never match as the one above will always happen first
}
}
Since version 0.08 the operator syntax is parsed using XS::Parse::Infix, meaning that custom infix
operators can be recognised, even on versions of perl that do not support the full "PL_infix_plugin"
mechanism.
case
case(VAL) { STATEMENTS... }
case(VAL), case(VAL), ... { STATEMENTS... }
A "case" statement must only appear inside the braces of a "match". It provides a block of code to run if
the controlling expression's value matches the value given in the "case" statement, according to the
comparison operator.
Multiple "case" statements are permitted for a single block. A value matching any of them will run the
code inside the block.
If the value is a non-constant expression, such as a variable or function call, it will be evaluated as
part of performing the comparison every time the "match" statement is executed. For best performance it
is advised to extract values that won't need computing again into a variable or "use constant" that can
be calculated just once at program startup; for example:
use constant CONDITION => a_function("with", "arguments");
match( $var : eq ) {
case(CONDITION) { ... }
...
}
The :experimental(dispatch) feature selects a more efficient handling of sequences of multiple "case"
blocks with constant expressions. This handling is implemented with a custom operator that will entirely
confuse modules like "B::Deparse" or optree inspectors like coverage tools so is not selected by default,
but can be enabled for extra performance in critical sections.
case if
case if(EXPR) { STATEMENTS... }
case(VAL), case if(EXPR) { STATEMENTS... }
A "case" statement may also be written "case if" with a boolean predicate expression in parentheses. This
inserts a direct boolean test into the comparison logic, allowing for other logical tests that aren't
easily expressed as uses of the comparison operator. As "case if" is an alternative to a regular "case",
they can be combined on a single code block if required.
For example, when testing an inequality in a selection of numerical "==" tests, or a single regexp test
among some string "eq" tests.
match( $num : == ) {
case(0) { ... }
case(1), case(2) { ... }
case if($num < 5) { ... }
}
match( $str : eq ) {
case("abc") { ... }
case("def") { ... }
case if($str =~ m/g/) { ... }
}
By default the match value is not assigned into a variable that is visible to "case if" expressions, but
if needed a new lexical can be constructed by using a regular "my" assignment.
match( my $v = some_expression() : eq ) {
case if($v =~ m/pattern/) { ... }
}
default
A "default" statement must only appear inside the braces of a "match". If present, it must be the final
choice, and there must only be one of them. It provides a block of code to run if the controlling
expression's value did not match any of the given "case" labels.
COMPARISONS
As this syntax is fairly similar to a few other ideas, the following comparisons may be useful.
Core perl's given/when syntax
Compared to core perl's "given/when" syntax (available with "use feature 'switch'"), this syntax is
initially visually very similar but actually behaves very differently. Core's "given/when" uses the
smartmatch ("~~") operator for its comparisons, which is complex, subtle, and hard to use correctly -
doubly-so when comparisons against values stored in variables rather than literal constants are involved.
It can be unpredictable whether string or numerical comparison are being used, for example. By
comparison, this module requires the programmer to specify the comparison operator. The choice of string
or numerical comparison is given in the source code - there can be no ambiguity.
Additionally, the "isa" operator is also permitted, which has no equivalent ability in smartmatch.
Also, the "given/when" syntax permits mixed code within a "given" block which is run unconditionally, or
at least, until the first successful "when" statement is encountered. The syntax provided by this module
requires that the only code inside a "match" block be a sequence of "case" statements. No other code is
permitted.
Switch::Plain
Like this module, Switch::Plain also provides a syntax where the programmer specifies whether the
comparison is made using stringy or numerical semantics. "Switch::Plain" also permits additional
conditions to be placed on "case" blocks, whereas this module does not.
Additionally, the "isa" operator is also permitted, which has no equivalent ability in "Switch::Plain".
C's switch/case
The C programming language provides a similar sort of syntax, using keywords named "switch" and "case".
One key difference between that and the syntax provided for Perl by this module is that in C the "case"
labels really are just labels. The "switch" part of the statement effectively acts as a sort of computed
"goto". This often leads to bugs caused by forgetting to put a "break" at the end of a sequence of
statements before the next "case" label; a situation called "fallthrough". Such a mistake is impossible
with this module, because every "case" is provided by a block. Once execution has finished with the
block, the entire "match" statement is finished. There is no possibility of accidental fallthrough.
C's syntax only permits compiletime constants for "case" labels, whereas this module will also allow the
result of any runtime expression.
Code written in C will perform identically even if any of the "case" labels and associated code are moved
around into a different order. The syntax provided by this module notionally performs all of its tests in
the order they are written in, and any changes of that order might cause a different result.
TODO
This is clearly an early experimental work. There are many features to add, and design decisions to make.
Rather than attempt to list them all here it would be best to check the RT bug queue at
<https://rt.cpan.org/Dist/Display.html?Name=Syntax-Keyword-Match>
AUTHOR
Paul Evans <leonerd@leonerd.org.uk>