Data::Transform::Line - serialize and parse terminated records (lines)

  #!perl

  use POE qw(Wheel::FollowTail Filter::Line);

  POE::Session->create(
    inline_states => {
      _start => sub {
        $_[HEAP]{tailor} = POE::Wheel::FollowTail->new(
          Filename => "/var/log/system.log",
          InputEvent => "got_log_line",
          Filter => POE::Filter::Line->new(),
        );
      },
      got_log_line => sub {
        print "Log: $_[ARG0]\n";
      }
    }
  );

  POE::Kernel->run();
  exit;

  my $c_line_filter = Data::Transform::Line->new(
    InputLiteral => chr(0),
    OutputLiteral => chr(0),
  );

  my $c_line_filter = Data::Transform::Line->new(
    Literal => chr(0),
  );

  my $whichever_line_filter = Data::Transform::Line->new(
    InputLiteral => undef,
    OutputLiteral => "\n",
  );

  my $paragraph_filter = Data::Transform::Line->new(
    InputRegexp => "([\x0D\x0A]{2,})",
    OutputLiteral => "\n---\n",
  );

  /(\x0D\x0A?|\x0A\x0D?)/

  some stream dataCR
  LFother stream data

# vim: ts=2 sw=2 expandtab
package Data::Transform::Line;
use strict;

use Data::Transform;

use vars qw($VERSION @ISA);
$VERSION = '0.01';
@ISA = qw(Data::Transform);

use Carp qw(carp croak);

sub DEBUG () { 0 }

sub INPUT_BUFFER     () { 0 }
sub FRAMING_BUFFER   () { 1 }
sub INPUT_REGEXP     () { 2 }
sub OUTPUT_LITERAL   () { 3 }
sub AUTODETECT_STATE () { 4 }

sub AUTO_STATE_DONE   () { 0x00 }
sub AUTO_STATE_FIRST  () { 0x01 }
sub AUTO_STATE_SECOND () { 0x02 }

sub new {
  my $type = shift;

  croak "$type requires an even number of parameters" if @_ and @_ & 1;
  my %params = @_;

  croak "$type cannot have both Regexp and Literal line endings" if (
    defined $params{Regexp} and defined $params{Literal}
  );

  my ($input_regexp, $output_literal);
  my $autodetect = AUTO_STATE_DONE;

  # Literal newline for both incoming and outgoing.  Every other known
  # parameter conflicts with this one.
  if (defined $params{Literal}) {
    croak "A defined Literal must have a nonzero length"
      unless length($params{Literal});
    $input_regexp   = quotemeta $params{Literal};
    $output_literal = $params{Literal};
    if (  exists  $params{InputLiteral  } or # undef means something
          defined $params{InputRegexp   } or
          defined $params{OutputLiteral }    ) {
      croak "$type cannot have Literal with any other parameter";
    }

  } else { # Input and output are specified separately, then.

    # Input can be either a literal or a regexp.  The regexp may be
    # compiled or not; we don't rightly care at this point.
    if (exists $params{InputLiteral}) {
      $input_regexp = $params{InputLiteral};

      # InputLiteral is defined.  Turn it into a regexp and be done.
      # Otherwise we will autodetect it.
      if (defined($input_regexp) and length($input_regexp)) {
        $input_regexp = quotemeta $input_regexp;
      }
      else {
        $autodetect   = AUTO_STATE_FIRST;
        $input_regexp = '';
      }

      croak "$type cannot have both InputLiteral and InputRegexp"
        if defined $params{InputRegexp};
    }
    elsif (defined $params{InputRegexp}) {
      $input_regexp = $params{InputRegexp};
      # unreachable
      #croak "$type cannot have both InputLiteral and InputRegexp"
      #  if defined $params{InputLiteral};
    }
    else {
      $input_regexp = "(\\x0D\\x0A?|\\x0A\\x0D?)";
    }

    if (defined $params{OutputLiteral}) {
      $output_literal = $params{OutputLiteral};
    }
    else {
      $output_literal = "\x0D\x0A";
    }
  }

  delete @params{qw(Literal InputLiteral OutputLiteral InputRegexp)};
  carp("$type ignores unknown parameters: ", join(', ', sort keys %params))
    if scalar keys %params;

  my $self = bless [
    [],              # INPUT_BUFFER
    '',              # FRAMING_BUFFER
    $input_regexp,   # INPUT_REGEXP
    $output_literal, # OUTPUT_LITERAL
    $autodetect,     # AUTODETECT_STATE
  ], $type;

  DEBUG and warn join ':', @$self;

  $self;
}

sub clone {
   my $self = shift;

   my $new = bless [
      [],
      '',
      $self->[INPUT_REGEXP],
      $self->[OUTPUT_LITERAL],
      $self->[AUTODETECT_STATE],
   ];

   return bless $new, ref $self;
}

sub get_pending {
   my $self = shift;
   my @ret = @{$self->[INPUT_BUFFER]};
   if (length $self->[FRAMING_BUFFER]) {
      unshift @ret, $self->[FRAMING_BUFFER];
   }
   return @ret ? [ @ret ] : undef;
}

# get()           is inherited from Data::Transform.
# get_one_start() is inherited from Data::Transform.
# get_one()       is inherited from Data::Transform.

sub _handle_get_data {
  my ($self, $data) = @_;

   if (defined $data) {
      $self->[FRAMING_BUFFER] .= $data;
   }
  # Process as many newlines an we can find.
  LINE: while (1) {

    # Autodetect is done, or it never started.  Parse some buffer!
    unless ($self->[AUTODETECT_STATE]) {
      DEBUG and warn unpack 'H*', $self->[INPUT_REGEXP];
      last LINE
        unless $self->[FRAMING_BUFFER] =~ s/^(.*?)$self->[INPUT_REGEXP]//s;
      DEBUG and warn "got line: <<", unpack('H*', $1), ">>\n";

      return $1;
    }

    # Waiting for the first line ending.  Look for a generic newline.
    if ($self->[AUTODETECT_STATE] & AUTO_STATE_FIRST) {
      last LINE
        unless $self->[FRAMING_BUFFER] =~ s/^(.*?)(\x0D\x0A?|\x0A\x0D?)//;

      my $line = $1;

      # The newline can be complete under two conditions.  First: If
      # it's two characters.  Second: If there's more data in the
      # framing buffer.  Loop around in case there are more lines.
      if ( (length($2) == 2) or
           (length $self->[FRAMING_BUFFER])
         ) {
        DEBUG and warn "detected complete newline after line: <<$1>>\n";
        $self->[INPUT_REGEXP] = $2;
        $self->[AUTODETECT_STATE] = AUTO_STATE_DONE;
      }

      # The regexp has matched a potential partial newline.  Save it,
      # and move to the next state.  There is no more data in the
      # framing buffer, so we're done.
      else {
        DEBUG and warn "detected suspicious newline after line: <<$1>>\n";
        $self->[INPUT_REGEXP] = $2;
        $self->[AUTODETECT_STATE] = AUTO_STATE_SECOND;
      }

      return $line;
    }

    # Waiting for the second line beginning.  Bail out if we don't
    # have anything in the framing buffer.
    if ($self->[AUTODETECT_STATE] & AUTO_STATE_SECOND) {
      return unless length $self->[FRAMING_BUFFER];

      # Test the first character to see if it completes the previous
      # potentially partial newline.
      if (
        substr($self->[FRAMING_BUFFER], 0, 1) eq
        ( $self->[INPUT_REGEXP] eq "\x0D" ? "\x0A" : "\x0D" )
      ) {

        # Combine the first character with the previous newline, and
        # discard the newline from the buffer.  This is two statements
        # for backward compatibility.
        DEBUG and warn "completed newline after line: <<$1>>\n";
        $self->[INPUT_REGEXP] .= substr($self->[FRAMING_BUFFER], 0, 1);
        substr($self->[FRAMING_BUFFER], 0, 1) = '';
      }
      elsif (DEBUG) {
        warn "decided prior suspicious newline is okay\n";
      }

      # Regardless, whatever is in INPUT_REGEXP is now a complete
      # newline.  End autodetection, post-process the found newline,
      # and loop to see if there are other lines in the buffer.
      $self->[INPUT_REGEXP] = $self->[INPUT_REGEXP];
      $self->[AUTODETECT_STATE] = AUTO_STATE_DONE;
      next LINE;
    }

    die "consistency error: AUTODETECT_STATE = $self->[AUTODETECT_STATE]";
  }

  return;
}

# New behavior.  First translate system newlines ("\n") into whichever
# newlines are supposed to be sent.  Second, add a trailing newline if
# one doesn't already exist.  Since the referenced output list is
# supposed to contain one line per element, we also do a split and
# join.  Bleah. ... why isn't the code doing what the comment says?

sub _handle_put_data {
  my ($self, $line) = @_;

  return $line . $self->[OUTPUT_LITERAL];
}


1;

__END__