PPI::Token::Word - The generic "word" Token

  PPI::Token::Word
  isa PPI::Token
      isa PPI::Element

package PPI::Token::Word;

use strict;
use PPI::Token ();

use vars qw{$VERSION @ISA %OPERATOR %QUOTELIKE};
BEGIN {
	$VERSION = '1.215';
	@ISA     = 'PPI::Token';

	# Copy in OPERATOR from PPI::Token::Operator
	*OPERATOR  = *PPI::Token::Operator::OPERATOR;

	%QUOTELIKE = (
		'q'  => 'Quote::Literal',
		'qq' => 'Quote::Interpolate',
		'qx' => 'QuoteLike::Command',
		'qw' => 'QuoteLike::Words',
		'qr' => 'QuoteLike::Regexp',
		'm'  => 'Regexp::Match',
		's'  => 'Regexp::Substitute',
		'tr' => 'Regexp::Transliterate',
		'y'  => 'Regexp::Transliterate',
	);
}

sub literal {
	my $self = shift;
	my $word = $self->content;

	# Expand Foo'Bar to Foo::Bar
	$word =~ s/\'/::/g;

	return $word;
}

sub method_call {
	my $self = shift;

	my $previous = $self->sprevious_sibling;
	if (
		$previous
		and
		$previous->isa('PPI::Token::Operator')
		and
		$previous->content eq '->'
	) {
		return 1;
	}

	my $snext = $self->snext_sibling;
	return 0 unless $snext;

	if (
		$snext->isa('PPI::Structure::List')
		or
		$snext->isa('PPI::Token::Structure')
		or
		$snext->isa('PPI::Token::Operator')
		and (
			$snext->content eq ','
			or
			$snext->content eq '=>'
		)
	) {
		return 0;
	}

	if (
		$snext->isa('PPI::Token::Word')
		and
		$snext->content =~ m< \w :: \z >xms
	) {
		return 1;
	}

	return;
}

my %backoff = map { $_ => 1 } qw{
    eq ne ge le gt lt
    q qq qx qw qr m s tr y
    pack unpack
};

sub __TOKENIZER__on_char {
	my $class = shift;
	my $t     = shift;

	# Suck in till the end of the bareword
	my $rest = substr( $t->{line}, $t->{line_cursor} );
	if ( $rest =~ /^(\w+(?:(?:\'|::)\w+)*(?:::)?)/ ) {
		my $word = $1;
		# Special Case: If we accidentally treat eq'foo' like
		# the word "eq'foo", then just make 'eq' (or whatever
		# else is in the %backoff hash.
		if ( $word =~ /^(\w+)'/ && $backoff{$1} ) {
		    $word = $1;
		}
		$t->{token}->{content} .= $word;
		$t->{line_cursor} += length $word;

	}

	# We might be a subroutine attribute.
	my $tokens = $t->_previous_significant_tokens(1);
	if ( $tokens and $tokens->[0]->{_attribute} ) {
		$t->{class} = $t->{token}->set_class( 'Attribute' );
		return $t->{class}->__TOKENIZER__commit( $t );
	}

	# Check for a quote like operator
	my $word = $t->{token}->{content};
	if ( $QUOTELIKE{$word} and ! $class->__TOKENIZER__literal($t, $word, $tokens) ) {
		$t->{class} = $t->{token}->set_class( $QUOTELIKE{$word} );
		return $t->{class}->__TOKENIZER__on_char( $t );
	}

	# Or one of the word operators
	if ( $OPERATOR{$word} and ! $class->__TOKENIZER__literal($t, $word, $tokens) ) {
	 	$t->{class} = $t->{token}->set_class( 'Operator' );
 		return $t->_finalize_token->__TOKENIZER__on_char( $t );
	}

	# Unless this is a simple identifier, at this point
	# it has to be a normal bareword
	if ( $word =~ /\:/ ) {
		return $t->_finalize_token->__TOKENIZER__on_char( $t );
	}

	# If the NEXT character in the line is a colon, this
	# is a label.
	my $char = substr( $t->{line}, $t->{line_cursor}, 1 );
	if ( $char eq ':' ) {
		$t->{token}->{content} .= ':';
		$t->{line_cursor}++;
		$t->{class} = $t->{token}->set_class( 'Label' );

	# If not a label, '_' on its own is the magic filehandle
	} elsif ( $word eq '_' ) {
		$t->{class} = $t->{token}->set_class( 'Magic' );

	}

	# Finalise and process the character again
	$t->_finalize_token->__TOKENIZER__on_char( $t );
}



# We are committed to being a bareword.
# Or so we would like to believe.
sub __TOKENIZER__commit {
	my ($class, $t) = @_;

	# Our current position is the first character of the bareword.
	# Capture the bareword.
	my $rest = substr( $t->{line}, $t->{line_cursor} );
	unless ( $rest =~ /^((?!\d)\w+(?:(?:\'|::)\w+)*(?:::)?)/ ) {
		# Programmer error
		die "Fatal error... regex failed to match in '$rest' when expected";
	}

	# Special Case: If we accidentally treat eq'foo' like the word "eq'foo",
	# then unwind it and just make it 'eq' (or the other stringy comparitors)
	my $word = $1;
	if ( $word =~ /^(\w+)'/ && $backoff{$1} ) {
	    $word = $1;
	}

	# Advance the position one after the end of the bareword
	$t->{line_cursor} += length $word;

	# We might be a subroutine attribute.
	my $tokens = $t->_previous_significant_tokens(1);
	if ( $tokens and $tokens->[0]->{_attribute} ) {
		$t->_new_token( 'Attribute', $word );
		return ($t->{line_cursor} >= $t->{line_length}) ? 0
			: $t->{class}->__TOKENIZER__on_char($t);
	}

	# Check for the end of the file
	if ( $word eq '__END__' ) {
		# Create the token for the __END__ itself
		$t->_new_token( 'Separator', $1 );
		$t->_finalize_token;

		# Move into the End zone (heh)
		$t->{zone} = 'PPI::Token::End';

		# Add the rest of the line as a comment, and a whitespace newline
		# Anything after the __END__ on the line is "ignored". So we must
		# also ignore it, by turning it into a comment.
		$rest = substr( $t->{line}, $t->{line_cursor} );
		$t->{line_cursor} = length $t->{line};
		if ( $rest =~ /\n$/ ) {
			chomp $rest;
			$t->_new_token( 'Comment', $rest ) if length $rest;
			$t->_new_token( 'Whitespace', "\n" );
		} else {
			$t->_new_token( 'Comment', $rest ) if length $rest;
		}
		$t->_finalize_token;

		return 0;
	}

	# Check for the data section
	if ( $word eq '__DATA__' ) {
		# Create the token for the __DATA__ itself
		$t->_new_token( 'Separator', "$1" );
		$t->_finalize_token;

		# Move into the Data zone
		$t->{zone} = 'PPI::Token::Data';

		# Add the rest of the line as the Data token
		$rest = substr( $t->{line}, $t->{line_cursor} );
		$t->{line_cursor} = length $t->{line};
		if ( $rest =~ /\n$/ ) {
			chomp $rest;
			$t->_new_token( 'Comment', $rest ) if length $rest;
			$t->_new_token( 'Whitespace', "\n" );
		} else {
			$t->_new_token( 'Comment', $rest ) if length $rest;
		}
		$t->_finalize_token;

		return 0;
	}

	my $token_class;
	if ( $word =~ /\:/ ) {
		# Since its not a simple identifier...
		$token_class = 'Word';

	} elsif ( $class->__TOKENIZER__literal($t, $word, $tokens) ) {
		$token_class = 'Word';

	} elsif ( $QUOTELIKE{$word} ) {
		# Special Case: A Quote-like operator
		$t->_new_token( $QUOTELIKE{$word}, $word );
		return ($t->{line_cursor} >= $t->{line_length}) ? 0
			: $t->{class}->__TOKENIZER__on_char( $t );

	} elsif ( $OPERATOR{$word} ) {
		# Word operator
		$token_class = 'Operator';

	} else {
		# If the next character is a ':' then its a label...
		my $string = substr( $t->{line}, $t->{line_cursor} );
		if ( $string =~ /^(\s*:)(?!:)/ ) {
			if ( $tokens and $tokens->[0]->{content} eq 'sub' ) {
				# ... UNLESS its after 'sub' in which
				# case it is a sub name and an attribute
				# operator.
				# We COULD have checked this at the top
				# level of checks, but this would impose
				# an additional performance per-word
				# penalty, and every other case where the
				# attribute operator doesn't directly
				# touch the object name already works.
				$token_class = 'Word';
			} else {
				$word .= $1;
				$t->{line_cursor} += length($1);
				$token_class = 'Label';
			}
		} elsif ( $word eq '_' ) {
			$token_class = 'Magic';
		} else {
			$token_class = 'Word';
		}
	}

	# Create the new token and finalise
	$t->_new_token( $token_class, $word );
	if ( $t->{line_cursor} >= $t->{line_length} ) {
		# End of the line
		$t->_finalize_token;
		return 0;
	}
	$t->_finalize_token->__TOKENIZER__on_char($t);
}

# Is the word in a "forced" context, and thus cannot be either an
# operator or a quote-like thing. This version is only useful
# during tokenization.
sub __TOKENIZER__literal {
	my ($class, $t, $word, $tokens) = @_;

	# Is this a forced-word context?
	# i.e. Would normally be seen as an operator.
	unless ( $QUOTELIKE{$word} or $PPI::Token::Operator::OPERATOR{$word} ) {
		return '';
	}

	# Check the cases when we have previous tokens
	my $rest = substr( $t->{line}, $t->{line_cursor} );
	if ( $tokens ) {
		my $token = $tokens->[0] or return '';

		# We are forced if we are a method name
		return 1 if $token->{content} eq '->';

		# We are forced if we are a sub name
		return 1 if $token->isa('PPI::Token::Word') && $token->{content} eq 'sub';

		# If we are contained in a pair of curly braces,
		# we are probably a bareword hash key
		if ( $token->{content} eq '{' and $rest =~ /^\s*\}/ ) {
			return 1;
		}
	}

	# In addition, if the word is followed by => it is probably
	# also actually a word and not a regex.
	if ( $rest =~ /^\s*=>/ ) {
		return 1;
	}

	# Otherwise we probably arn't forced
	'';
}

1;