[comment {-*- tcl -*- doctools manpage}]
[manpage_begin grammar::aycock n 1.0]
[keywords ambiguous]
[keywords aycock]
[keywords earley]
[keywords grammar]
[keywords horspool]
[keywords parser]
[keywords parsing]
[keywords transducer]
[copyright "2006 by Kevin B. Kenny <kennykb@acm.org>
Redistribution permitted under the terms of the Open\
Publication License <http://www.opencontent.org/openpub/>"]
[moddesc "Aycock-Horspool-Earley parser generator for Tcl"]
[titledesc "Aycock-Horspool-Earley parser generator for Tcl"]
[category "Grammars and finite automata"]
[require Tcl 8.5]
[require grammar::aycock [opt 1.0]]
[description]
[para]
The [package grammar::aycock] package
implements a parser generator for the class of parsers described
in John Aycock and R. Nigel Horspool. Practical Earley Parsing.
[emph "The Computer Journal,"] [emph 45](6):620-630, 2002.
[uri http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.12.4254]
[section "PROCEDURES"]

The [package grammar::aycock] package exports the single procedure:

[list_begin definitions]
[call [cmd ::aycock::parser] [arg grammar] [opt [option -verbose]]]

Generates a parser for the given [arg grammar], and returns its
name.  If the optional [option -verbose] flag is given, dumps verbose
information relating to the generated parser to the standard output.
The returned parser is an object that accepts commands as shown in
[sectref {OBJECT COMMAND}] below.

[list_end]

[section "OBJECT COMMAND"]

[list_begin definitions]
[call [arg parserName] [method parse] [arg symList] [arg valList] [opt [arg clientData]]]

Invokes a parser returned from [cmd ::aycock::parser]. [arg symList] is
a list of grammar symbols representing the terminals in an input
string, and [arg valList] is a list of their semantic values. The
result is the semantic value of the entire string when parsed.

[call [arg parserName] [method destroy]]

Destroys a parser constructed by [cmd ::aycock::parser].

[call [arg parserName] [method terminals]]

Returns a list of terminal symbols that may be presented in the
[arg symList] argument to the [method parse] object command.

[call [arg parserName] [method nonterminals]]

Returns a list of nonterminal symbols that were defined in the
parser's grammar.

[call [arg parserName] [method save]]

Returns a Tcl script that will reconstruct the parser without
needing all the mechanism of the parser generator at run time.
The reconstructed parser depends on a set of commands in the
package [package grammar::aycock::runtime],
which is also automatically loaded
when the [package grammar::aycock] package is loaded.

[list_end]

[section "DESCRIPTION"]

The [cmd grammar::aycock::parser] command accepts a grammar expressed as
a Tcl list.  The list must be structured as the concatenation of a set
of [term rule]s. Each [term rule] comprises a variable number of
elements in the list:

[list_begin itemized]

[item] The name of the nonterminal symbol that the rule reduces.

[item] The literal string, [const "::="]

[item] Zero or more names of terminal or nonterminal symbols that
comprise the right-hand-side of the rule.

[item] Finally, a Tcl script to execute when the rule is reduced.
Within the given script, a variable called [var _] contains a list of
the semantic values of the symbols on the right-hand side. The value
returned by the script is expected to be the semantic value of the
left-hand side.  If the [arg clientData] parameter was passed to the
[method parse] method, it is available in a variable called
[var clientData].  It is permissible for the script to be the empty
string.  In this case, the semantic value of the rule will be the same
as the semantic value of the first symbol on the right-hand side.  If
the right-hand side is also empty, the semantic value will be the
empty string.

[list_end]

Parsing is done with an Earley parser, which is not terribly efficient
in speed or memory consumption, but which deals effectively with
ambiguous grammars.  For this reason, the [package grammar::aycock] package is
perhaps best adapted to natural-language processing or the parsing of
extraordinarily complex languages in which ambiguity can be tolerated.

[section EXAMPLE]

The following code demonstrates a trivial desk calculator, admitting
only [const +], [const *] and parentheses as its operators.  It also
shows the format in which the lexical analyzer is expected to present
terminal symbols to the parser.

[example {
set p [aycock::parser {
    start ::= E {}
    E ::= E + T {expr {[lindex $_ 0] + [lindex $_ 2]}}
    E ::= T {}
    T ::= T * F {expr {[lindex $_ 0] * [lindex $_ 2]}}
    T ::= F {}
    F ::= NUMBER {}
    F ::= ( E ) {lindex $_ 1}
}]
puts [$p parse {(  NUMBER +  NUMBER )  *  ( NUMBER +  NUMBER ) } \
	       {{} 2      {} 3      {} {} {} 7     {} 1      {}}]
$p destroy
}]

The example, when run, prints [const 40].

[section KEYWORDS]

Aycock, Earley, Horspool, parser, compiler
[manpage_end]