NAME

Text::Query::Advanced - Match text against Boolean expression

SYNOPSIS

    use Text::Query::Advanced;
    
    # Constructor
    $query = Text::Query::Advanced->new([QSTRING] [OPTIONS]);

    # Methods
    $query->prepare(QSTRING [OPTIONS]);
    $query->match([TARGET]);
    $query->matchscalar([TARGET]);

    # Methods that can be overridden to produce custom query trees, etc.

    $query->build_final_expression(Q1);
    $query->build_expression(Q1,Q2);
    $query->build_expression_finish(Q1);
    $query->build_conj(Q1,Q2,F);
    $query->build_near(Q1,Q2);
    $query->build_concat(Q1,Q2);
    $query->build_negation(Q1);
    $query->build_literal(Q1);

DESCRIPTION

This module provides an object that matches a string or list of strings against a Boolean query expression similar to an AltaVista "advanced query". Elements of the query expression may be regular expressions or literal text.

Query expressions are compiled into an internal form (currently, a regular expression making use of most of the tricks listed in Recipe 6.17 of _The Perl Cookbook_) when a new object is created or the prepare method is called; they are not recompiled on each match.

The class provided by this module may be subclassed to produce query processors that match against input other than literal strings, e.g. indices.

Query expressions consist of literal strings (or regexps) joined by the following operators, in order of precedence from lowest to highest:

OR, |

AND, &

NEAR

NOT, !

Operator names are not case-sensitive. Note that if you want to use a | in a regexp, you need to backwhack it to keep it from being seen as a query operator. Sub-expressions may be quoted in single or double quotes to match "and," "or," or "not" literally and may be grouped in parentheses ((, )) to alter the precedence of evaluation.

A parenthesized sub-expression may also be concatenated with other sub- expressions to match sequences: (Perl or Python) interpreter would match either "Perl interpreter" or "Python interpreter". Concatenation has a precedence higher than NOT but lower than AND. Juxtaposition of simple words has the highest precedence of all.

EXAMPLES

  use Text::Query::Advanced;
  my $q=new Text::Query::Advanced('hello and world');
  die "bad query expression" if not defined $q;
  print if $q->match;
  ...
  $q->prepare('goodbye or adios or ta ta',-litspace=>1,-case=>1);
  #requires single space between the two ta's
  if ($q->match($line)) {
  #doesn't match "Goodbye"
  ...
  $q->prepare('"and" or "or"');
  #quoting operators for literal match
  ...
  $q->prepare('\\bintegrate\\b',-regexp=>1);
  #won't match "disintegrated"

CONSTRUCTOR

new ([QSTRING] [OPTIONS])

This is the constructor for a new Text::Query object. If a QSTRING is given it will be compiled to internal form.

OPTIONS are passed in a hash like fashion, using key and value pairs. Possible options are:

-case - If true, do case-sensitive match.

-litspace - If true, match spaces (except between operators) in QSTRING literally. If false, match spaces as \s+.

-near - Sets the number of words that can occur between two expressions and still satisfy the NEAR operator. The default is 10.

-regexp - If true, treat patterns in QSTRING as regular expressions rather than literal text.

-whole - If true, match whole words only, not substrings of words.

The constructor will return undef if a QSTRING was supplied and had illegal syntax.

METHODS

prepare (QSTRING [OPTIONS])

Compiles the query expression in QSTRING to internal form and sets any options (same as in the constructor). prepare may be used to change the query expression and options for an existing query object. If OPTIONS are omitted, any options set by a previous call to the constructor or prepare remain in effect.

This method returns a reference to the query object if the syntax of the expression was legal, or undef if not.

match ([TARGET])

If TARGET is a scalar, match returns a true value if the string specified by TARGET matches the query object's query expression. If TARGET is not given, the match is made against $_.

If TARGET is an array, match returns a (possibly empty) list of all matching elements. If the elements of the array are references to sub- arrays, the match is done against the first element of each sub-array. This allows arbitrary information (e.g. filenames) to be associated with each string to match.

If TARGET is a reference to an array, match returns a reference to a (possibly empty) list of all matching elements.

matchscalar ([TARGET])

Behaves just like MATCH when TARGET is a scalar or is not given. Slightly faster than MATCH under these circumstances.

CODE-GENERATION METHODS

The following methods are used to generate regexps based on query elements. They may be overridden to generate other forms of matching code, such as trees to be used by a custom version of match that evaluates index lists or the like.

All these methods return a scalar corresponding to the code that performs the specified options. As supplied, they return regexp strings, but overridden methods could return objects, array references, etc.

Parameters Q1 and Q2 are the same type of scalar as the return values.

build_final_expression(Q1)

Does any final processing to generate code to match a top-level expression. As supplied, optionally adds case-insensitivity code and then uses qr// to compile the regexp. The return value will be stored in the object's matchexp field. It is NOT necessarily of a type that can be passed to the other code-generation methods.

build_expression(Q1,Q2)

Generate code to match Q1 OR Q2

build_expression_finish(Q1)

Generate any code needed to enclose an expression. As supplied, encloses the generated regexp in non-capturing parentheses.

build_conj(Q1,Q2,F)

Generate code needed to match Q1 AND Q2. F will be true if this is the first time this method is called in a sequence of several conjunctions (the supplied method uses this to factor a common ^ out of the generated sub- expressions, which greatly speeds up matching).

build_near(Q1,Q2)

Generate code needed to match Q1 NEAR Q2.

build_concat(Q1,Q2)

Generate code needed to match Q1 immediately followed by Q2.

build_negation(Q1)

Generate code needed to match NOT Q1.

build_literal(Q1)

Generate code to match Q1 as a literal.

AUTHOR

Eric Bohlman (ebohlman@netcom.com)

CREDITS

The parse_tokens routine was adapted from the parse_line routine in Text::Parsewords.

COPYRIGHT

Copyright (c) 1998-1999 Eric Bohlman. All rights reserved. This program is free software; you can redistribute and/or modify it under the same terms as Perl itself.