in reply to Break string into array

It's basically the same problem as recursive algorithm for nested data structures

Update: Actually, it's a bit more different than I thought because you said you can't have both OR and AND in the same list. 

# make_parser.pl

use strict;
use warnings;

use Parse::RecDescent qw( );

my $grammar = <<'__EOI__';

    {
        use strict;
        use warnings;

        sub dequote {
            my ($s) = @_;
            $s =~ s/^"//;
            $s =~ s/"\z//;
            $s =~ s/\\(.)/$1/sg;
            return $s;
        }
    }

    parse     : list /\Z/ { $item[1] }

    list      : term list_[ $item[1] ]

    list_     : "AND" <commit> and_list
                { [ $item[1] => $arg[0], @{$item[3]} ] }

              | "OR"  <commit> or_list
                { [ $item[1] => $arg[0], @{$item[3]} ] }

              | { $arg[0] }

    and_list  : term and_list_
                { [ $item[1], @{$item[2]} ] }

    and_list_ : "AND" <commit> term and_list_
                { [ $item[3], @{$item[4]} ] }

              | { [] }

    or_list   : term or_list_
                { [ $item[1], @{$item[2]} ] }

    or_list_  : "OR" <commit> term or_list_
                { [ $item[3], @{$item[4]} ] }

              | { [] }

    term      : IDENT
              | STRING
              | '(' list ')' { $item[2] }

    IDENT     : /\w+/
    STRING    : /"(?:\\.|[^\\])*"/s { dequote($item[1]) }

__EOI__

Parse::RecDescent->Precompile($grammar, 'Parser')
    or die("Bad grammar\n");

[download]
# test.pl

use strict;
use warnings;

use Data::Dumper qw( Dumper );
use Parser       qw( );

my $text = 'dogs OR cats OR "flying fish" OR (shrimp AND squid)';

my $parser = Parser->new();

print(Dumper($parser->parse($text)));

[download]
>perl make_parser.pl

>perl test.pl
$VAR1 = [
          'OR',
          'dogs',
          'cats',
          'flying fish',
          [
            'AND',
            'shrimp',
            'squid'
          ]
        ];

[download]

Update: Small changes to grammar to eliminate backtracking.

Replies are listed 'Best First'.
Re^2: Break string into array
by Anonymous Monk on Sep 18, 2009 at 08:48 UTC
    Here is how that might look with Regexp::Grammars 
    #!/usr/bin/perl --

use strict;
use warnings;

my $s = q[dogs OR cats OR "flying fish" OR (shrimp AND squid)];

my $parser = do {
    use Regexp::Grammars;
    qr{
# <logfile: - >
<[TERM]>*
<rule: TERM>  <OP> | <MATCH=IDENT> | <MATCH=STRING> | <LIST>
<rule: STRING> "([^"]+?)"
<rule: OP> AND|OR
<rule: IDENT> \w+
<rule: LIST> \( <[TERM]>* \)
    }xs
};

if($s =~ $parser){
    my(%rash) = %/;#bah for scite lexer /#
    undef %/;# bah for scite lexer /#

    use Data::Dumper();
    print Data::Dumper->new([\%rash])->Indent(1)->Useqq(1)->Dump,"\n";
    
    kek(\%rash); # kill empty key
    print Data::Dumper->new([\%rash])->Indent(1)->Useqq(1)->Dump,"\n";
    my $rash = reorder_terms(\%rash); # consumes %rash
    print Data::Dumper->new([$rash])->Indent(1)->Useqq(1)->Dump,"\n";
}

sub reorder_terms {
    my( $ref ) = @_;
    if( $$ref{TERM}){
        my @term;
        my @op;
        for my $t( @{$$ref{TERM}} ){
            if( ref $t ){
                if( $$t{OP} ){
                    push @op, delete $$t{OP};
                }elsif( $$t{LIST} ){
                    push @term, reorder_terms(delete $$t{LIST} );
                }else{
                    die "uh oh, no OP or LIST key";
                }
            } else {
                push @term, $t;
            }
        }
        undef %$ref;
        #return [@op, @term ];
        return [$op[0], @term ];
    }

    die "uh oh, no TERM key";
}


sub kek {
    my ($ref) = @_;
    my $typ = ref $ref;

    if( $typ eq 'HASH'){
        delete $$ref{""};
        for my $val( values %$ref){
            ref $val and kek($val);
        }
    }
    if( $typ eq 'ARRAY'){
        for my $val( @$ref){
            ref $val and kek($val);
        }
    }
    return;
}

__END__
$VAR1 = {
  "" => "dogs OR cats OR \"flying fish\" OR (shrimp AND squid)",
  "TERM" => [
    "dogs",
    {
      "" => " OR",
      "OP" => "OR"
    },
    "cats",
    {
      "" => " OR",
      "OP" => "OR"
    },
    "\"flying fish\"",
    {
      "" => " OR",
      "OP" => "OR"
    },
    {
      "" => " (shrimp AND squid)",
      "LIST" => {
        "" => "(shrimp AND squid)",
        "TERM" => [
          "shrimp",
          {
            "" => " AND",
            "OP" => "AND"
          },
          "squid"
        ]
      }
    }
  ]
};

$VAR1 = {
  "TERM" => [
    "dogs",
    {
      "OP" => "OR"
    },
    "cats",
    {
      "OP" => "OR"
    },
    "\"flying fish\"",
    {
      "OP" => "OR"
    },
    {
      "LIST" => {
        "TERM" => [
          "shrimp",
          {
            "OP" => "AND"
          },
          "squid"
        ]
      }
    }
  ]
};

$VAR1 = [
  "OR",
  "dogs",
  "cats",
  "\"flying fish\"",
  [
    "AND",
    "shrimp",
    "squid"
  ]
];

    [download]
    Uncomment <logfile: - > for some debug. See also KinoSearch::Docs::Cookbook::CustomQueryParser, Text::Query.
[reply]
[d/l]
[select]

      Not only does your solution allow 

      my $s = q[dogs OR cats AND "flying fish" OR (shrimp AND squid)];

      [download]
      it parses to the same as 
      my $s = q[dogs OR cats OR "flying fish" OR (shrimp AND squid)];

      [download]

      And it also allows 

      my $s = q[OR dogs OR cats OR "flying fish" OR (shrimp AND squid)];

      [download]

      Finally, quoted strings are left quoted. A parser shouldn't return literals. If you want to differentiate between quoted and unquoted terms, you'll need to add to the parse tree. 

      $VAR1 = [
  "OR",
  [ term   => "dogs"        ],
  [ term   => "cats"        ],
  [ phrase => "flying fish" ],
  [
    "AND",
    [ term => "shrimp" ],
    [ term => "squid"  ],
  ]
];

      [download]
[reply]
[d/l]
[select]
        That is all true, but it does satisfy the OPs requirement/example :)
[reply]

In Section Seekers of Perl Wisdom