File Coverage

blib/lib/HTML/Embellish.pm

Criterion	Covered	Total	%
statement	127	134	94.7
branch	51	62	82.2
condition	20	30	66.6
subroutine	11	11	100.0
pod	3	4	75.0
total	212	241	87.9

#---------------------------------------------------------------------
package HTML::Embellish;
#
# Copyright 2010 Christopher J. Madsen
#
# Author: Christopher J. Madsen
# Created: October 8, 2006
#
# This program is free software; you can redistribute it and/or modify
# it under the same terms as Perl itself.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See either the
# GNU General Public License or the Artistic License for more details.
#
# ABSTRACT: Typographically enhance HTML trees
#---------------------------------------------------------------------

use 5.008; # Need good Unicode support; Perl 5.10 recommended but 5.8 may work
use warnings;
use strict;
use Carp qw(croak);

use Exporter 5.57 'import';     # exported import method

###open(LOG, '>:utf8', 'em.log');

#=====================================================================
# Package Global Variables:

our $VERSION = '1.000';
# This file is part of HTML-Embellish 1.000 (February 1, 2014)

our @EXPORT = qw(embellish);

my $mdash = chr(0x2014);
my $lsquo = chr(0x2018);
my $rsquo = chr(0x2019);
my $ldquo = chr(0x201C);
my $rdquo = chr(0x201D);
my $hellip = chr(0x2026);

my $notQuote = qq/[^\"$ldquo$rdquo]/;
my $balancedQuoteString = qr/(?: (?>[^ \t\n\r\pP]+)
                               | (?= [ \t\n\r\pP])$notQuote
                               | $ldquo (?>$notQuote*) $rdquo )*/x;

#=====================================================================
# Constants:
#---------------------------------------------------------------------

BEGIN
{
  my $i = 0;
  for (qw(textRefs fixQuotes fixDashes fixEllipses fixEllipseSpace fixHellip
          totalFields)) {
    ## no critic (ProhibitStringyEval)
    eval "sub $_ () { $i }";
    ++$i;
  }
} # end BEGIN

#=====================================================================
# Exported functions:
#---------------------------------------------------------------------
sub embellish
{
  my $html = shift @_;

  croak "First parameter of embellish must be an HTML::Element"
      unless ref $html and $html->can('content_refs_list');

  my $e = HTML::Embellish->new(@_);
  $e->process($html);
} # end embellish

#=====================================================================
# Class Methods:
#---------------------------------------------------------------------
sub new
{
  my $class = shift;
  croak "Odd number of parameters passed to HTML::Embellish->new" if @_ % 2;
  my %parms = @_;

  my $self = [ (undef) x totalFields ];
  bless $self, $class;

  my $def = (exists $parms{default} ? $parms{default} : 1);

  $self->[textRefs]    = undef;
  $self->[fixDashes]   = (exists $parms{dashes}   ? $parms{dashes}   : $def);
  $self->[fixEllipses] = (exists $parms{ellipses} ? $parms{ellipses} : $def);
  $self->[fixQuotes]   = (exists $parms{quotes}   ? $parms{quotes}   : $def);

  $self->[fixHellip]       = (exists $parms{hellip}
                              ? $parms{hellip} : $self->[fixEllipses]);
  $self->[fixEllipseSpace] = (exists $parms{space_ellipses}
                              ? $parms{space_ellipses} : $self->[fixEllipses]);

  return $self;
} # end new

#---------------------------------------------------------------------
# Convert quotes & apostrophes into curly quotes:
#
# Input:
#   self:  The HTML::Embellish object
#   refs:  Arrayref of stringrefs to the text of this paragraph

sub processTextRefs
{
  my ($self, $refs) = @_;

  local $_ = join('', map { $$_ } @$refs);
  utf8::upgrade($_);

  my $fixQuotes = $self->[fixQuotes];
  if ($fixQuotes) {
    s/\("/($ldquo/g;
    s/"\)/$rdquo)/g;

    s/^([\xA0\s]*)"/$1$ldquo/;
    s/(?<=[\s\pZ])"(?=[^\s\pZ])/$ldquo/g;
    s/(?<=\pP)"(?=\w)/$ldquo/g;
    s/(?<=[ \t\n\r])"(?=\xA0)/$ldquo/g;

    s/"[\xA0\s]*$/$rdquo/;
    s/(?
    s/(?<=\w)"(?=\pP)/$rdquo/g;
    s/(?<=\xA0)"(?=[ \t\n\r]|[\s\xA0]+$)/$rdquo/g;
    s/(?<=[,;.!?])"(?=[-$mdash])/$rdquo/go;

    s/'(?=(?:cause|cept|d|e[mr]?e?|fraidy?|gainst|im|m|n|nam|nothers?|nuff|re?|rithmetic|s|scuse|spects?|t|til|tisn?|tw(?:asn?|ere?n?|ould\w*)|ud|uns?)\b|\d\d\W?s|\d\d(?!\w))/$rsquo/ig;

    s/'([ \xA0]?$rdquo)/$rsquo$1/go;

    s/`/$lsquo/g;
    s/^'/$lsquo/;
    s/(?<=[\s\pZ])'(?=[^\s\pZ])/$lsquo/g;
    s/(?<=\pP)(?
    s/(?<=[ \t\n\r])'(?=\xA0)/$lsquo/g;

    s/'/$rsquo/g;

    s/(?
    s/(${rsquo}[\xA0\s]+)"(?!\PZ)/$1$rdquo/go;

    if (/"/) {
      1 while s/^($balancedQuoteString (?![\"$ldquo$rdquo])[ \t\n\r\pP]) "
               /$1$ldquo/xo
          or  s/^($balancedQuoteString $ldquo $notQuote*) "/$1$rdquo/xo;
    } # end if straight quotes remaining in string

    #s/(?<=\p{IsPunct})"(?=\p{IsAlpha})/$ldquo/go;
    s/(?<=[[:punct:]])"(?=[[:alpha:]])/$ldquo/go;

    s/${ldquo}\s([$lsquo$rsquo])/$ldquo\xA0$1/go;
    s/${rsquo}\s$rdquo/$rsquo\xA0$rdquo/go;
  } # end if fixQuotes

  if ($self->[fixEllipses]) {
    s/( [\"$ldquo$lsquo] \.(?:\xA0\.)+ ) \s /$1\xA0/xog;
    s/\s (?= \. (?:\xA0[.,!?])+ [$rdquo$rsquo\xA0\"]* $)/\xA0/xo;
  }

  # Return the text to where it came from:
  #   This only works because the replacement text is always
  #   the same length as the original.
  foreach my $r (@$refs) {
    $$r = substr($_, 0, length($$r), '');
    if ($fixQuotes) {
      # Since the replacement text isn't the same length,
      # these can't be done on the string as a whole:
      $$r =~ s/(?<=[$ldquo$rdquo])(?=[$lsquo$rsquo])/\xA0/go;
      $$r =~ s/(?<=[$lsquo$rsquo])(?=[$ldquo$rdquo])/\xA0/go;
      $$r =~ s/(?<=[$ldquo$lsquo])\xA0(?=\.\xA0\.)//go;
    } # end if fixQuotes
  } # end foreach @$refs
} # end processTextRefs

#---------------------------------------------------------------------
# Recursively process an HTML::Element tree:

sub process
{
  my ($self, $elt) = @_;

  croak "HTML::Embellish->process must be passed an HTML::Element"
      unless ref $elt and $elt->can('content_refs_list');

  return if $elt->is_empty;

  my $parentRefs;
  my $isP = ($elt->tag =~ /^(?: p | h\d | d[dt] | div | blockquote | title )$/x);

  if ($isP and ($self->[fixQuotes] or $self->[fixEllipses])) {
    $parentRefs = $self->[textRefs];
    $self->[textRefs] = []
  } # end if need to collect text refs

  $elt->normalize_content;
  my @content = $elt->content_refs_list;

  if ($self->[fixQuotes] and $self->[textRefs] and @content) {
    # A " that opens a tag can be assumed to be a left quote
    ${$content[ 0]} =~ s/^"/$ldquo/ unless ref ${$content[ 0]};
    # A " that ends a tag can be assumed to be a right quote
    ${$content[-1]} =~ s/"$/$rdquo/ unless ref ${$content[-1]};
  }

  foreach my $r (@content) {
    if (ref $$r) { # element node
      my $tag = $$r->tag;
      next if $tag =~ /^(?: ~comment | script | style )$/x;

      if ($self->[textRefs] and $tag eq 'br') {
        my $break = "\n";
        push @{$self->[textRefs]}, \$break;
      }
      $self->process($$r);
    } else { # text node
      # Convert -- to em-dash:
      utf8::upgrade($$r);
      if ($self->[fixDashes]) {
        $$r =~ s/(?
        $$r =~ s/(?
      } # end if fixDashes

      $$r =~ s/$hellip/.../go if $self->[fixHellip];

      # Fix ellipses:
      if ($self->[fixEllipses]) {
        $$r =~ s/(?
        $$r =~ s/(?
        $$r =~ s/(?<= \.) [^\PZ\x{200B}] (?=[.,?!])/\xA0/gx;
        $$r =~ s/(?:(?<=\w)|\A) (\.\xA0\.\xA0\.|\.\.\.)(?=[ \xA0\n\"\'?!$rsquo$rdquo])(?![ \xA0\n]+\w)/\xA0$1/go;
      } # end if fixEllipses

      if ($self->[fixEllipseSpace]) {
        $$r =~ s/(?<=\w) (\.(?:\xA0\.)+) (?=\w)/ $1 /gx;
        $$r =~ s/(?<=\w[!?,;]) (\.(?:\xA0\.)+) (?=\w)/ $1 /gx;
        $$r =~ s/( [\"$ldquo$lsquo] \.(?:\xA0\.)+ ) (?=\w) /$1\xA0/xog;
        $$r =~ s/(?<=\w) (\.\xA0\.\xA0\.) (?![\xA0\w])/\xA0$1/gx;

        if ($self->[textRefs] and @{$self->[textRefs]}) {
          $$r =~ s/^(\.(?:\xA0\.)+) (?=\w)/ $1 /gx
              if ${$self->[textRefs][-1]} =~ /\w[!?,;]?$/;

          ${$self->[textRefs][-1]} =~ s/(?<=\w)\xA0(\.\xA0\.\xA0\.)$/ $1 /
              if $$r =~ /^\w/;
        }
      } # end if fixEllipseSpace

      push @{$self->[textRefs]}, $r if $self->[textRefs];
    } # end else text node
  } # end foreach $r

  if ($isP and $self->[textRefs]) {
###    print LOG (map { utf8::is_utf8($$_) . "{$$_}" } @{ $self->[textRefs] }), "\n";
    $self->processTextRefs($self->[textRefs]);
    push @$parentRefs, @{$self->[textRefs]} if $parentRefs;
    $self->[textRefs] = $parentRefs;
  } # end if this was a paragraph-like element
} # end process

#=====================================================================
# Package Return Value:

1;

__END__