NAME
Marpa::R2::NAIF - Marpa named argument interface (NAIF)
Synopsis
use Marpa::R2;
my $grammar = Marpa::R2::Grammar->new(
{ start => 'Expression',
actions => 'My_Actions',
default_action => 'first_arg',
rules => [
{ lhs => 'Expression', rhs => [qw/Term/] },
{ lhs => 'Term', rhs => [qw/Factor/] },
{ lhs => 'Factor', rhs => [qw/Number/] },
{ lhs => 'Term', rhs => [qw/Term Add Term/], action => 'do_add' },
{ lhs => 'Factor',
rhs => [qw/Factor Multiply Factor/],
action => 'do_multiply'
},
],
}
);
$grammar->precompute();
my $recce = Marpa::R2::Recognizer->new( { grammar => $grammar } );
$recce->read( 'Number', 42 );
$recce->read('Multiply');
$recce->read( 'Number', 1 );
$recce->read('Add');
$recce->read( 'Number', 7 );
sub My_Actions::do_add {
my ( undef, $t1, undef, $t2 ) = @_;
return $t1 + $t2;
}
sub My_Actions::do_multiply {
my ( undef, $t1, undef, $t2 ) = @_;
return $t1 * $t2;
}
sub My_Actions::first_arg { shift; return shift; }
my $value_ref = $recce->value;
my $value = $value_ref ? ${$value_ref} : 'No Parse';About this document
This document contains a top-level overview of, and tutorial for, the named argument inteface (NAIF) for the Marpa parse engine. If you are a new to Marpa, you want to start with the tutorial for the the Scanless interface (SLIF) instead.
The NAIF is a middle level interface. It is more low level than the Scanless interface (SLIF), which uses a domain-specific language. But it is higher level, and provides more features, than the thin interface, which provides direct access to the underlying Libmarpa C library.
The two examples in this document show the typical flows of NAIF Marpa method calls. This document will use these examples to describe the basic features of Marpa in semi-tutorial fashion. More advanced features, and full reference details of all features, can be found in the other Marpa API documents.
The three phases
A parser needs to:
Accept a grammar.
Read input.
Return values from the parses, according to a semantics.
In Marpa these three tasks are, for the most part, distinct phases. Grammars are Marpa::R2::Grammar objects. The reading of input and the evaluation of the parse according to the semantics is performed by Marpa::R2::Recognizer objects.
Example 1: a simple calculator
The synopsis shows the code for a very simple calculator. It handles only addition and multiplication of integers. This section explains, line by line, how it works.
Marpa::R2::Grammar::new
my $grammar = Marpa::R2::Grammar->new(
{ start => 'Expression',
actions => 'My_Actions',
default_action => 'first_arg',
rules => [
{ lhs => 'Expression', rhs => [qw/Term/] },
{ lhs => 'Term', rhs => [qw/Factor/] },
{ lhs => 'Factor', rhs => [qw/Number/] },
{ lhs => 'Term', rhs => [qw/Term Add Term/], action => 'do_add' },
{ lhs => 'Factor',
rhs => [qw/Factor Multiply Factor/],
action => 'do_multiply'
},
],
}
);Marpa grammars are Marpa::R2::Grammar objects. They are created with the Marpa::R2::Grammar::new constructor. The arguments to Marpa::R2::Grammar::new are references to hashes of named arguments. In the key/value pairs of these hashes, the hash key is the name of the argument, and the hash value is the value of the named argument.
The start named argument
start => 'Expression',The start named argument is required. Its value is a string containing the name of the grammar's start symbol.
Named arguments for the semantics
actions => 'My_Actions',
default_action => 'first_arg',The actions and default_action named arguments specify semantics. Their argument values are strings, which acquire their semantics during evaluation.
Evaluation will be described later. Peeking ahead, actions provides the name of a Perl package where Marpa will look for its actions. The default_action named argument will be interpreted as an action name in that package. This action name will resolve to an action -- a Perl closure that implements semantics. The action specified by default_action is used as the action for rules with no action of their own.
The rules named argument
rules => [
{ lhs => 'Expression', rhs => [qw/Term/] },
{ lhs => 'Term', rhs => [qw/Factor/] },
{ lhs => 'Factor', rhs => [qw/Number/] },
{ lhs => 'Term', rhs => [qw/Term Add Term/], action => 'do_add' },
{ lhs => 'Factor',
rhs => [qw/Factor Multiply Factor/],
action => 'do_multiply'
},
],The value of the rules named argument is a reference to an array of rule descriptors. In this example, all the rule descriptors are in the "long" form -- they are references to hashes of rule properties. In each key/value pair of a rule descriptor hash, the key is the name of a rule property, and the hash value is the value of that rule property.
The lhs property
The value of the lhs rule property must be a string containing the name of the rule's left hand side symbol. Every Marpa rule must have a left hand side symbol.
The rhs property
The value of the rhs property is a reference to an array of strings containing names of the rule's right hand symbols, in order. This array may be zero length, in which case this is an empty rule -- a rule with no symbols on the right hand side. There are no empty rules in this example.
The action property
The value of the action rule property is a string. Peeking ahead, each action property string will be interpreted as an action name. This action name will be resolved to a Perl closure that implements the rule's semantics.
Marpa::R2::Grammar::precompute
$grammar->precompute();Before a Marpa grammar object can be used by a Marpa recognizer, it must be precomputed. Precomputation compiles data structures that the recognizer will need.
Marpa::R2::Recognizer::new
my $recce = Marpa::R2::Recognizer->new( { grammar => $grammar } );Marpa::R2::Recognizer::new creates a new recognizer. Its arguments are references to hashes of named arguments. In this example the only named argument is the required argument: "grammar". The value of the grammar named argument must be a precomputed Marpa grammar.
Marpa::R2::Recognizer::read
$recce->read( 'Number', 42 );
$recce->read('Multiply');
$recce->read( 'Number', 1 );
$recce->read('Add');
$recce->read( 'Number', 7 );The Marpa::R2::Recognizer::read method takes two arguments, a token name and a token value. The token name must be the name of a valid terminal symbol in the grammar. By default symbols are valid as terminal symbols, if and only if they do NOT occur on the LHS of any rule.
The token value must be a Perl scalar, but otherwise its form and semantics are entirely up to the application. If the token value is never used, it can be omitted. In the calculator example, the values of the "Add" and "Multiply" tokens are never used, and are allowed to default to an undefined value.
Marpa::R2::Recognizer::value
my $value_ref = $recce->value;
my $value = $value_ref ? ${$value_ref} : 'No Parse';The Marpa::R2::Recognizer::value method returns a reference to the parse result's value, if there was a parse result. If there was no parse result, Marpa::R2::Recognizer::value returns undef.
Resolving the semantics
The first thing Marpa::R2::Recognizer::value needs to do is to resolve the semantics. Resolving the semantics means mapping the action names into actions. Actions are Perl closures which directly implement semantics. In this example, the actions named argument is specified. actions is a Perl package name. Marpa will look for actions in that package.
actions => 'My_Actions',{ lhs => 'Factor', rhs => [qw/Factor Multiply Factor/], action => 'do_multiply' },For example, the action property for the above rule is "do_multiply" and the actions named argument to the grammar was "My_Actions". So Marpa looks for a closure whose fully qualified name is My_Actions::do_multiply, which it finds:
sub My_Actions::do_multiply {
my ( undef, $t1, undef, $t2 ) = @_;
return $t1 * $t2;
}Rules do not always have action properties. That is the case with these rules in this example:
{ lhs => 'Expression', rhs => [qw/Term/] },
{ lhs => 'Term', rhs => [qw/Factor/] },
{ lhs => 'Factor', rhs => [qw/Number/] },The rules in the above display have no action names. When a rule has no action name, Marpa will fall back to trying to use the default action, as described next.
default_action => 'first_arg',The default_action named argument is resolved in the same way as are the action properties of the rules. In this example, default_action is specified as "first_arg" and resolves to My_Actions::first_arg.
Actions
sub My_Actions::first_arg { shift; return shift; }sub My_Actions::do_add {
my ( undef, $t1, undef, $t2 ) = @_;
return $t1 + $t2;
}Value actions are Perl closures used as callbacks. Value actions are called when nodes in a parse tree are evaluated. A value action receives one or more arguments. The first argument to a value action is always a per-parse-tree object, which the callbacks can use as a scratchpad. In these examples, the per-parse-tree object is not used.
For a non-empty rule, the second and any subsequent arguments to the callback are the values, in lexical order, of the symbols on the right hand side of the rule. If the action is for an empty rule, the per-parse-tree object will be its only argument.
Every value action is expected to return a value. With one exception, this value is passed up to a parent node as an argument. The exception is the value for the start rule. The return value for the start rule becomes the parse result.
Rules with no action specified for them take their semantics from the default_action named argument. If there is no default action for a grammar, rules with no action specified for them return a Perl undef.
Example 2: an ambiguous parse
This is the same calculator as before, rewritten to be ambiguous. Rather than give multiplication precedence over addition, the rewritten calculator allows any order of operations. In this example, the actions (My_Actions::do_add, etc.) and the @tokens array remain the same as before.
Eliminating precedence makes the grammar shorter, but it also means there can be multiple parse trees, and that the different parse trees can have different parse results. In this application we decide, for each input, to return every one of the parse results.
use Marpa::R2;
my $ambiguous_grammar = Marpa::R2::Grammar->new(
{ start => 'E',
actions => 'My_Actions',
rules => [
[ 'E', [qw/E Add E/], 'do_add' ],
[ 'E', [qw/E Multiply E/], 'do_multiply' ],
[ 'E', [qw/Number/], ],
],
default_action => 'first_arg',
}
);
$ambiguous_grammar->precompute();
my $ambiguous_recce =
Marpa::R2::Recognizer->new( { grammar => $ambiguous_grammar } );
$ambiguous_recce->read( 'Number', 42 );
$ambiguous_recce->read('Multiply');
$ambiguous_recce->read( 'Number', 1 );
$ambiguous_recce->read('Add');
$ambiguous_recce->read( 'Number', 7 );
my @values = ();
while ( defined( my $ambiguous_value_ref = $ambiguous_recce->value() ) ) {
push @values, ${$ambiguous_value_ref};
}Short form rule descriptors
rules => [
[ 'E', [qw/E Add E/], 'do_add' ],
[ 'E', [qw/E Multiply E/], 'do_multiply' ],
[ 'E', [qw/Number/], ],
],The rule descriptors in the ambiguous example demonstrate the "short" or array form of rule descriptors. Array form rule descriptors are references to arrays. Here the elements are, in order, the lhs property, the rhs property, and the action property.
Marpa::R2::Recognizer::value
my @values = ();
while ( defined( my $ambiguous_value_ref = $ambiguous_recce->value() ) ) {
push @values, ${$ambiguous_value_ref};
}When called more than once, the Marpa::R2::Recognizer::value method iterates through the parse results. For each call, it returns a reference to the parse result. At the end of the iteration, after all parse results have been returned, Marpa::R2::Recognizer::value returns undef. If there were no parse results, Marpa::R2::Recognizer::value returns undef the first time that it is called.
Errors and exceptions
As a general rule, methods in the Marpa NAIF API do not return errors. When there are errors, Marpa NAIF API methods throw an exception.
Inheritance
Classes in the Marpa API are not designed to be inherited.
The Marpa:: namespace
The Marpa:: top-level namespace is reserved. For extensions to Marpa, one appropriate place is the MarpaX:: namespace. This practice helps avoid namespace collisions, and follows a CPAN standard, as exemplified by the DBIx:: LWPx:: and MooseX:: which are for extensions of, respectively, DBI, LWP and Moose.
Other documents
This document gives a semi-tutorial overview of the entire Marpa NAIF API. For full details on Marpa's grammar objects and their methods, see the Marpa::R2::Grammar document. For full details on Marpa's recognizer objects and their methods, see the Marpa::R2::Recognizer document.
Marpa::R2::Vocabulary is intended as a quick refresher in parsing terminology, emphasizing how the standard terms are used in the Marpa context. Marpa's standard semantics are fully described in the Marpa::R2::Semantics document. Techniques for tracing and for debugging your Marpa grammars are described in the Marpa::R2::Tracing document and the Marpa::R2::Progress document. For those with a theoretical bent, my sources, and other useful references, are described in Marpa::R2::Advanced::Bibliography.
Support
Marpa::R2 comes without warranty. Support is provided on a volunteer basis through the standard mechanisms for CPAN modules. The Support document has details.
Copyright and License
Copyright 2013 Jeffrey Kegler
This file is part of Marpa::R2. Marpa::R2 is free software: you can
redistribute it and/or modify it under the terms of the GNU Lesser
General Public License as published by the Free Software Foundation,
either version 3 of the License, or (at your option) any later version.
Marpa::R2 is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser
General Public License along with Marpa::R2. If not, see
http://www.gnu.org/licenses/.