File Coverage

lib/Graph/Regexp.pm

Criterion	Covered	Total	%
statement	188	224	83.9
branch	58	98	59.1
condition	18	38	47.3
subroutine	15	19	78.9
pod	7	7	100.0
total	286	386	74.0

############################################################################
# Generate flowcharts from Regexp debug dumpes
#

package Graph::Regexp;

require 5.008001;
use Graph::Easy;
use Graph::Easy::Base;

$VERSION = 0.05;
@ISA = qw/Graph::Easy::Base/;

use strict;

# Perl 5.8.8, might be different for 5.10?
use constant MAX_MATCHES => 32767;

#############################################################################
#############################################################################

sub _init
  {
  my ($self, $args) = @_;

  $self->{options} = {};
  $self->{debug} = $args->{debug} || 0;
  $self->reset();
  $self;
  }

sub option
  {
  my $self = shift;
  $self->{options}->{$_[0]};
  }

sub graph
  {
  # decompose regexp dump and return as Graph::Easy object

  # allow Graph::Regexp->graph() calling style
  my $class = 'Graph::Regexp';
  $class = shift if @_ == 2; $class = ref($class) if ref($class);
  my $code = shift;

  my $self = $class->new();
  $self->reset();
  $self->parse($code);

  $self->{graph};			# return the Graph::Easy object
  }

sub as_graph
  {
  # return the internal Graph::Easy object
  my $self = shift;

  $self->{graph};
  }

sub as_ascii
  {
  # return the graph as ASCII
  my $self = shift;

  $self->{graph}->as_ascii();
  }

BEGIN
  {
  # make an alias for decompose
  *decompose = \&parse;
  }

sub parse
  {
  my ($self, $doc) = @_;

  $self->reset();				# clear data

  $self->_croak("Expected SCALAR ref, but got " . ref($doc))
   if ref($doc) && ref($doc) ne 'SCALAR';

  $self->_croak("Got filename '$doc', but can't read it: $!")
   if !ref($doc) && !-f $doc;

  # XXX TODO: filenames

  $self->_parse($$doc);

  $self;
  }

sub reset
  {
  # reset the internal structure
  my $self = shift;

  delete $self->{fail};
  delete $self->{success};
  $self->{graph} = Graph::Easy->new();

  $self->{stack} = [];
  $self->{entries} = {};

  $self;
  }

sub graph_label
  {
  # get/set the label of the graph
  my ($self) = shift;

  my $g = $self->{graph};
  if (@_ > 0)
    {
    $g->set_attribute('label',$_[0]);
    }
  $g->label();
  }

#############################################################################
#############################################################################
# main parse routine, recursive

sub _setup_nodeclass
  {
  # add the attributes for one node class
  my ($self, $class, $title, $label) = @_;

  my $g = $self->{graph};

  $g->set_attribute("node.$class", 'title', $title);
  $g->set_attribute("node.$class", 'label', $label);
  }

sub _parse
  {
  # take the regexp string and decompose it into a tree, then turn this into
  # a graph.
  my ($self, $text) = @_;

  my $g = $self->{graph};

  # add the start node
  my $root = $g->add_node('0');
  $g->set_attribute('root','0');	# the first node is the root
  $root->set_attribute('label','START');
  $root->set_attribute('class','start');

  # add the final fail and success nodes
  $self->{fail} = $g->add_node('FAIL');
  $self->{success} = $g->add_node('SUCCESS');
  $self->{fail}->set_attribute('class','fail');
  $self->{success}->set_attribute('class','success');

  # this is a hack to workaround that Graph::Easy has not yet "end => '0'" for edges
  $self->{fail}->set_attribute('origin','SUCCESS');
  $self->{fail}->set_attribute('offset','0,2');

  $g->set_attribute('node.nothing', 'label', "\\''");
  $g->set_attribute('node.nothing', 'title', "Nothing (always matches)");

  # Special nodes:
  #  ^ (BOL)
  #  $ (EOL)
  #  \z (EOS)
  #  \Z (SEOL)
  #  \A (SBOL)
  #  \b \B (BOUND, NBOUND)
  #  \d \D (DIGIT, NDIGIT)
  #  \w \W (ALNUM, NALNUM)

  $self->_setup_nodeclass('bol',   'BOL (Begin Of Line)', '^');
  $self->_setup_nodeclass('eol',   'EOL (End Of Line)', '$');
  $self->_setup_nodeclass('eos',   'EOS (End Of String)', '\\z');
  $self->_setup_nodeclass('seol',  'SEOL (String end or End Of Line)', '\\Z');
  $self->_setup_nodeclass('sbol',  'SBOL (String begin or Begin Of Line)', '\\A');
  $self->_setup_nodeclass('bound',   'BOUND (Boundary)', '\\b');
  $self->_setup_nodeclass('nbound',  'NBOUND (Non-boundary)', '\\B');
  $self->_setup_nodeclass('digit',   'DIGIT (Digit)', '\\d');
  $self->_setup_nodeclass('ndigit',  'NDIGIT (Non-digit)', '\\D');
  $self->_setup_nodeclass('alnum',   'ALNUM (Alphanumeric)', '\\w');
  $self->_setup_nodeclass('nalnum',  'NALNUM (Non-alphanumeric)', '\\W');

  $g->set_attributes('node.fail', { fill => 'darkred', color => 'white' } );
  $g->set_attributes('node.success', { fill => 'darkgreen', color => 'white' } );

  $g->set_attributes('edge.match', {
	'label' => 'match',
	'color' => 'darkgreen'
	} );
  $g->set_attributes('edge.always', {
	'label' => 'always',
	} );
  $g->set_attributes('edge.fail', {
	'label' => 'fail',
	'color' => 'darkred'
	} );

#  The general family of this object. These are any of:
#   alnum, anchor, anyof, anyof_char, anyof_class, anyof_range,
#   assertion, bol, branch, close, clump, digit, exact, flags, group, groupp,
#   minmod, open, prop, sol, eol, seol, sbol, quant, ref, reg_any,
#   star, plus ...

  # first we parse the following text:

#   1: OPEN1(3)
#   3:   BRANCH(6)
#   4:     EXACT (9)
#   6:   BRANCH(9)
#   7:     EXACT (9)
#   9: CLOSE1(11)
#  11: EXACT (13)
#  13: PLUS(16)
#  14:   EXACT (0)
#  16: EXACT <1>(18)

  # into entries like:

  #  { id => 1, level => 0, type => "open", next => 3, id => 1, }

  # to preserve the entries in their original order
  my $stack = $self->{stack};
  # to quickly find entries by their id
  my $entries = $self->{entries};

  $text =~ s/[\r\n]\z//;

  print STDERR "# Input: \n# '$text'\n" if $self->{debug};

  my @lines = split /\n/, $text; my $index = 0;
  for my $line (@lines)
    {
    # ignore all other lines
    next unless $line =~ /^\s+(\d+):(\s+)[A-Z]/;

    print STDERR "# Parsing line: '$line'\n" if $self->{debug} > 1;

    # level: ' ' => 0, '   ' => 1 etc
    my $entry = { level => (length($2)-1) / 2, id => $1 };

    # "7: EXACT (9)" => "EXACT (9)"
    $line =~ s/^\s+\d+:\s+//;

    # OPEN1(3)  or OPEN1 (3)
    if ($line =~ /^([A-Z][A-Z0-9]+)\s*\((\d+)\)/)
      {
      $entry->{class} = lc($1);
      $entry->{next} = $2;
      $entry->{exact} = '';
      }
    # EXACT (16) or EXACT  (16)
    elsif ($line =~ /^([A-Z][A-Z0-9-]+)(\s*<(.+)>)?\s*\((\d+)\)/)
      {
      $entry->{class} = lc($1);
      my $t = $3;
      $entry->{next} = $4;
      $t =~ s/(\$|\@|\\)/\\$1/g;		# quote $, @ and \
      $entry->{exact} = "\\\"$t\\\"";
      $entry->{title} = "EXACT <$t>";
      }
    # TRIE-EXACT [bo](9)
    elsif ($line =~ /^TRIE-EXACT\s*(\[([^\]]+)\])\s*?\((\d+)\)/)
      {
      $entry->{class} = 'trie';
      $entry->{title} = "TRIE-EXACT <$1>";
      $entry->{exact} = "$1";
      $entry->{next} = $2;
      }
    # ANYOF[ab](8)
    elsif ($line =~ /^([A-Z][A-Z0-9-]+)\s*(\[([^\]]+)\])\s*?\((\d+)\)/)
      {
      $entry->{class} = lc($1);
      if ($entry->{class} eq 'anyof')
	{
        $entry->{exact} = "[$3]";
	}
      elsif ($entry->{class} eq 'nothing')
	{
        $entry->{exact} = "[$3]";
	}
      else
	{
        $entry->{exact} = "\"$3\"";
        }
      $entry->{title} = "EXACT <$3>";
      $entry->{next} = $4;
      }
    # CURLY {0,1}(22) or CURLY {0,1} (22)
    elsif ($line =~ /^([A-Z][A-Z0-9]+)\s*\{(\d+),(\d+)\}\s*\((\d+)\)/)
      {
      $entry->{class} = lc($1);
      $entry->{next} = $4;
      $entry->{min} = $2;
      $entry->{max} = $3;
      $entry->{exact} = "\{$entry->{min},$entry->{max}\}";
      }
    # CURLYM[1] {0,1}(22) or CURLY {0,1} (22) or CURLYX[1] {1,2}(22)
    elsif ($line =~ /^([A-Z][A-Z0-9]+)\[[^]]\]\s*\{(\d+),(\d+)\}\s*\((\d+)\)/)
      {
      $entry->{class} = lc($1);
      $entry->{next} = $4;
      $entry->{min} = $2;
      $entry->{max} = $3;
      $entry->{exact} = "\{$entry->{min},$entry->{max}\}";
      # make curlym, curly and curlyx all "curly"
      $entry->{class} = 'curly' if $entry->{class} =~ /^curly/;
      }
    # PLUS (22)
    elsif ($line =~ /^PLUS\s*\((\d+)\)/)
      {
      $entry->{class} = 'plus';
      $entry->{next} = $1;
      $entry->{min} = 1;
      $entry->{max} = MAX_MATCHES;
      $entry->{exact} = "\{$entry->{min},$entry->{max}\}";
      }
    $entry->{class} =~ s/[0-9]//g;	# OPEN1 => open
    $entry->{index} = $index++;

    push @$stack, $entry;
    $entries->{ $entry->{id} } = $entry;

    next if $entry->{class} =~ /(open|close|branch|end|succeed|curly|minmod|plus|star|whilem)/;

    # add the nodes right away
    # print STDERR "# adding node for $line\n";

    my $n = $g->add_node($entry->{id});
    $n->set_attribute('label', $entry->{exact}) if $entry->{exact} ne '';
    $n->set_attribute('class', $entry->{class});
    $n->set_attribute('title', $entry->{title}) if $entry->{title};

    $entry->{node} = $n;
    }

  # empty text => matches always
  if (keys %$entries == 0)
    {
    my $edge = $g->add_edge( $root, $self->{success});
    $edge->set_attribute('class','always');
    return $self;
    }

  # Now we take the stack of entries and transform it into a graph by
  # connecting all the nodes with "match" and "fail" edges.

  # Notes:

  #  Each tried (sub)expression in the regexp has exactly two outcomes:
  #  'match' or 'fail'.
  #  If a expression consists of more than on part than it is handled
  #  like an "and" (first and second part must match).
  #  F.i. in "[ab]foo", if [ab] matches, it goes to try "foo", If it
  #  it fails, it goes one level up. Likewise for "foo", match goes
  #  on to the next part and fail goes up.
  #  If we are already at level 0, the entire expression fails.

  #  Branches try each subexpression in order, that is if one subexpression
  #  fails, it goes to the next branch. If any of them matches, it goes
  #  on to the next part, and if all of them fail, it goes up.

  # /just(another|perl)hacker/ will result in:

#   1: EXACT (3)
#   3: OPEN1(5)
#   5:   BRANCH(9)
#   6:     EXACT (12)
#   9:   BRANCH(12)
#  10:     EXACT (12)
#  12: CLOSE1(14)
#  14: EXACT (17)
#  17: END(0)

  # [ just ] - match -> [ another ] - match -> [ hacker ] - match -> [ success ]
  #   |                   |                       ^   |
  #   | fail              | fail                  |   |
  #   |                   |                       |   | fail
  #   |                 [ perl ]    - match ------|   |
  #   |                   |                           |
  #   |                   | fail                      |
  #   -------------------------------------------------------------> [ fail ]

  # XXX TODO: each OPEN/CLOSE pair should result in a subgroup. This is not
  #           yet possible since Graph::Easy doesn't allow nesting yet.

  # connect the root node to the first part
  my $next = $self->_find_node($stack->[0]);
  my $edge = $g->add_edge( $root, $next);

  # The "NOTHING" node has no predecessor and needs to be weeded out:
  #
  #  1: CURLYM[1] {0,32767}(15)
  #  5:   BRANCH(8)
  #  6:     EXACT (13)
  #  8:   BRANCH(11)
  #  9:     EXACT (13)
  # 13:   SUCCEED(0)
  # 14:   NOTHING(15)
  # 15: END(0)

  ###########################################################################
  ###########################################################################
  # main conversion loop

  # the entry/part we are trying
  my $i = 0;
  while ($i < @$stack)
    {
    my $entry = $stack->[$i];

    next unless exists $entry->{node};

    if ($entry->{class} eq 'nothing' && $entry->{node}->predecessors() == 0)
      {
      # a nothing node with no incoming connection, filter it out
      $g->del_node($entry->{node});
      next;
      }

    # the "match" egde goes to the next part
    my $next = $self->_find_next($entry);

    my $n = $next; $n = $self->{success} unless defined $n;

    my $edge = $g->add_edge( $entry->{node}, $n);
    $edge->set_attribute('class','match');

    if ($n == $self->{success})
      {
      $edge->set_attribute('end','back,0');
      }

    # nothing nodes do not have a fail edge, they match always
    if ( ($entry->{class} eq 'nothing') ||
         (defined $entry->{min} && $entry->{min} == 0) )
      {
      $edge->set_attribute('class','always');
      next;
      }

    # generate the fail edge:

    # if the next node is $self->{success}, then fail must be $self->{fail}
    my $fail = $self->{fail};
    # walked over end?
    if (!defined $next)
      {
      $fail = $self->_find_next_branching($entry);
      }
    # otherwise, find the next branching part
    elsif ($next != $self->{success})
      {
      $fail = $self->_find_next_branching($entry);
      }

    $edge = $g->add_edge( $entry->{node}, $fail);
    $edge->set_attribute('class','fail');
    $edge->set_attribute('end','back,0');

    } continue { $i++; }

  # if there are no incoming edges to fail, the regexp always matches (like //):
  $g->del_node($self->{fail}) if scalar $self->{fail}->incoming() == 0;

  $self;
  }

sub _find_next_branching
  {
  # Given an entry on the stack, go backwards to find the
  # last branch, then skip to the next part in that branch.
  # If there is no next part, try one level higher, until
  # we are at the upper-most level.
  my ($self, $entry) = @_;

  # Example:

  # starting with 14: EXACT (19)

#   1: EXACT <0>(3)
#   3: OPEN1(5)
#   5:   BRANCH(8)
#   6:     EXACT (35)
#   8:   BRANCH(32)
#   9:     EXACT (11)
#  11:     OPEN2(13)
#  13:       BRANCH(16)			1 # look at next(16) is it a branch?
					  # yes it is, so go forward to it
#  14:         EXACT (19)		0 # find 13: BRANCH(16)
#  16:       BRANCH(19)			2 # skip forward
#  17:         EXACT (19)		3 # return this
#  19:     CLOSE2(21)
#  21:     ANYOF[i](35)
#  32:   BRANCH(35)
#  33:     EXACT (35)
#  35: CLOSE1(37)
#  37: EXACT (39)
#  39: END(0)

  # starting with 17: EXACT (19)

#   1: EXACT <0>(3)
#   3: OPEN1(5)
#   5:   BRANCH(8)
#   6:     EXACT (35)
#   8:   BRANCH(32)			2 # look at next(32) is it a branch?
					  # yes it is, so go forward to it
#   9:     EXACT (11)
#  11:     OPEN2(13)
#  13:       BRANCH(16)
#  14:         EXACT (19)
#  16:       BRANCH(19)			1 # look at next(16) is it a branch?
					  # no, 19 is not, so find 8: BRANCH(32)
#  17:         EXACT (19)		0 # find 16: BRANCH(19)
#  19:     CLOSE2(21)
#  21:     ANYOF[i](35)
#  32:   BRANCH(35)
#  33:     EXACT (35)		3 # return this:
#  35: CLOSE1(37)
#  37: EXACT (39)
#  39: END(0)

  print STDERR "# find next branch for $entry->{id}\n" if $self->{debug};

  my $entries = $self->{entries};
  do {
    # find branch one level up
    my $branch = $self->_find_previous_branch($entry);

    print STDERR "#  prev branch for $entry->{id} should be at $branch->{id}\n"
	if $self->{debug} && $branch && defined $branch->{id};

    # no branch above us, fail completely
    return $self->{fail} unless defined $branch;

    # skip to next part
    $entry = $entries->{ $branch->{next} };

    print STDERR "# next branch should be at $entry->{id} ($entry->{class})\n"
	if $self->{debug};

    return $self->{fail} if $entry && $entry->{class} eq 'end';

    # loop ends if there is a next part in the current branch
    } while ($entry->{class} ne 'branch');

  # skip over the branch, open etc to the first real part
  $entry = $self->_find_node($entry);

  print STDERR "# next branch is at $entry->{id}\n"
	if $self->{debug};

  $entry;
  }

sub _find_previous_branch
  {
  # Given an entry on the stack, go backwards to find the
  # last branch.
  my ($self, $entry) = @_;

  my $entries = $self->{entries};
  my $stack = $self->{stack};

  my $index = $entry->{index};

 print STDERR "# Finding prev branch for entry $entry->{id}\n"
	if $self->{debug};

  # the branch must be this level or lower
  my $level = $entry->{level};

  # go backwards until we find a BRANCH
  while ($index > 0)
    {
    $index--;
    my $e = $stack->[$index];

    print STDERR "#  Found $entry->{id} ($level vs $e->{level}\n"
	if $self->{debug} && $entry && $entry->{class} eq 'branch';

    return $e if $e->{class} eq 'branch' && $e->{level} <= $level;
    }
  # the part we looked at is in the upper-most level, so there is
  # no next branch part we can skip to, meaning we fail completely.
  return;
  }

sub _find_node
  {
  # Given an entry on the stack, skip to next entry if the current
  # isnt a node itself.
  my ($self, $entry) = @_;

  # Example:

#   3: OPEN1(5)				# open => skip, go to next
#   5:   BRANCH(9)			# branch => skip, go to next
#   6:     EXACT (12)		# return this

#   1: EXACT (3)			# return this
#   3: OPEN1(5)
#   5:   BRANCH(9)
#   6:     EXACT (12)

  print STDERR "#  find node for entry $entry->{id}\n"
	if $self->{debug};

  my $entries = $self->{entries};
  my $stack = $self->{stack};
  while (!exists $entry->{node})
    {
    print STDERR "#  at entry $entry->{id}\n"
	if $self->{debug};

    if ($entry->{class} =~ /^(open|branch|plus|star|curly)/)
      {
      $entry = $stack->[ $entry->{index} + 1 ];
      }
    else
      {
      $entry = $entries->{ $entry->{next} };
      }
    return $self->{success} unless ref $entry;		# walked over end
    }

  $entry->{node};
  }

sub _find_next
  {
  # Given an entry on the stack, find the next entry.
  my ($self, $entry) = @_;

  # Example:

#   1: EXACT (3)			# go to 3
#   3: OPEN1(5)				# open => skip, go to next
#   5:   BRANCH(9)			# branch => skip, go to next
#   6:     EXACT (12)		# return this

  print STDERR "# Skipping ahead for $entry->{id}:\n"
	if $self->{debug};
  my $entries = $self->{entries};
  my $stack = $self->{stack};
  do
    {
    print STDERR "#  at entry $entry->{id}\n"
	if $self->{debug};

    if ($entry->{class} =~ /^(open|branch|plus|star|curly)/)
      {
      $entry = $stack->[ $entry->{index} + 1 ];
      }
    else
      {
      $entry = $entries->{ $entry->{next} };
      }
    return unless ref $entry;		# walked over end

    print STDERR "#   next $entry->{id}\n"
	if $self->{debug} && ref($entry);
    } while (!exists $entry->{node});

  print STDERR "# return $entry->{id}\n"
	if $self->{debug};

  $entry->{node};
  }

1;
__END__