Make identical text nodes reference identical

 Note that this assumes that the tree is not manipulated
 afterwards. Since the code assumes this already (or breaks when the
 tree is manipulated, eg. by breaking _rank numbers), I don't think
 that is a big deal.

diff --git a/lib/HTML/TreeBuilder/XPath.pm b/lib/HTML/TreeBuilder/XPath.pm
index ccc5f6b..aeda292 100644 (file)
--- a/lib/HTML/TreeBuilder/XPath.pm
+++ b/lib/HTML/TreeBuilder/XPath.pm
@@ -116,20 +116,13 @@ sub find_xpath           { my( $elt, $path)= @_; return xp->find(
 sub matches              { my( $elt, $path)= @_; return xp->matches( $elt, $path, $elt);        }
 sub set_namespace        { my $elt= shift; xp->new->set_namespace( @_); }
 
-{
     my %roots;
-    sub DESTROY {
-        my $self = shift;
-        delete $roots{$self};
-        $self->SUPER::DESTROY if $self->can("SUPER::DESTROY");
-    }
 
     sub getParentNode
     { my $elt= shift;
       return $elt->{_parent} if $elt->{_parent};
       return $roots{$elt} //= bless { _root => $elt }, 'HTML::TreeBuilder::XPath::Root';
     }
-}
 
 sub getRootNode
   { my $elt= shift;
@@ -207,21 +200,35 @@ sub string_value
 # called on a parent, with a child as second argument and its rank as third
 # returns the child if it is already an element, or
 # a new HTML::TreeBuilder::XPath::Text element if it is a plain string
+    my %text_children;
+
+sub DESTROY {
+    my $self = shift;
+    delete $text_children{$self};
+    delete $roots{$self};
+    $self->SUPER::DESTROY if $self->can("SUPER::DESTROY");
+}
+
 sub _child_as_object
   { my( $elt, $elt_or_text, $rank)= @_;
     return undef unless( defined $elt_or_text);
     if( ! ref $elt_or_text)
       { # $elt_or_text is a string, turn it into a TextNode object
-        $elt_or_text= bless { _content => $elt_or_text, _parent => $elt, }, 
-                            'HTML::TreeBuilder::XPath::TextNode'
-                      ;
-        Scalar::Util::weaken($elt_or_text->{_parent});
+          if(!defined $text_children{$elt}{$rank} ||
+             ($elt_or_text= $text_children{$elt}{$rank}, $elt_or_text->{_rank} != $rank || $elt_or_text->{_parent} != $elt)) {
+              $elt_or_text = ($text_children{$elt}{$rank} =
+                              bless { _content => $elt_or_text, _parent => $elt, },
+                              'HTML::TreeBuilder::XPath::TextNode')
+                  ;
+              Scalar::Util::weaken($elt_or_text->{_parent});
+          }
       }
     if( ref $rank) { warn "rank is a ", ref( $rank), " elt_or_text is a ", ref( $elt_or_text); } 
     $elt_or_text->{_rank}= $rank; # used for sorting;
     return $elt_or_text;
   }
 
+
 sub toString { return shift->as_XML( @_); }
 
 # produces better looking XML

diff --git a/t/HTML-TreeBuilder-XPath.t b/t/HTML-TreeBuilder-XPath.t
index cdf78b3..f81ac3a 100644 (file)
--- a/t/HTML-TreeBuilder-XPath.t
+++ b/t/HTML-TreeBuilder-XPath.t
@@ -3,7 +3,7 @@
 
 #########################
 
-use Test::More tests => 31;
+use Test::More tests => 34;
 BEGIN { use_ok('HTML::TreeBuilder::XPath') };
 
 #########################
@@ -37,6 +37,10 @@ is( $html->findvalue( '//*[@id="foo"]/@id|//*[@id="foo"]/@class'), 'myspanfoo',
 is( $html->findvalue( '//*[@id="foo"]/@class|//*[@id="foo"]/@id'), 'myspanfoo', '2 atts on same element (unsorted)');
 
 is( $html->findvalue( 'count(/|/)' ), 1, 'count roots');
+is( $html->findvalue( 'count(/html/body/h1/text()|/html/body/h1/text())' ), 1, 'count identical text nodes');
+is( $html->findvalue( 'count(id("bq")/@id|id("bq")/@id)' ), 1, 'count identical attribute nodes');
+
+is( $html->findvalue( '//text()[contains(., "folks")]/preceding-sibling::*[1]' ), 'all', 'previous sibling on a text node');
 
 is( $html->findvalue( '//b'), 'boldall', '2 texts');
 is( join( '|', $html->findvalues( '//b')), 'bold|all', '2 texts with findvalues');