adamh-hpricot 0.6.171 → 0.6.210

data/ext/hpricot_scan/hpricot_scan.rl

@@ -19,10 +19,10 @@ VALUE hpricot_css(VALUE, VALUE, VALUE, VALUE, VALUE);
 #define NO_WAY_SERIOUSLY "*** This should not happen, please send a bug report with the HTML you're parsing to why@whytheluckystiff.net.  So sorry!"
 
 static VALUE sym_xmldecl, sym_doctype, sym_procins, sym_stag, sym_etag, sym_emptytag, sym_comment,
-      sym_cdata, sym_text, sym_EMPTY;
+      sym_cdata, sym_text, sym_EMPTY, sym_CDATA;
 static VALUE mHpricot, rb_eHpricotParseError;
 static VALUE cBaseEle, cBogusETag, cCData, cComment, cDoc, cDocType, cElem, cETag, cText,
-      cXMLDecl, cProcIns;
+      cXMLDecl, cProcIns, symAllow, symDeny;
 static ID s_ElementContent;
 static ID s_downcase, s_new, s_parent, s_read, s_to_str;
 static ID iv_parent;
@@ -30,7 +30,7 @@ static VALUE reProcInsParse;
 
 typedef struct {
   int name;
-  VALUE tag, attr, etag, raw;
+  VALUE tag, attr, etag, raw, EC;
   VALUE parent, children;
 } hpricot_ele;
 
@@ -249,6 +249,7 @@ hpricot_ele_clear_raw(VALUE self)
   he->tag = tag; \
   he->attr = attr; \
   he->raw = Qnil; \
+  he->EC = ec; \
   he->etag = he->parent = he->children = Qnil; \
   if (raw != NULL && (sym == sym_emptytag || sym == sym_stag || sym == sym_etag || sym == sym_doctype)) { \
     he->raw = rb_str_new(raw, rawlen); \
@@ -262,7 +263,7 @@ hpricot_ele_alloc(VALUE klass)
   VALUE ele;
   hpricot_ele *he = ALLOC(hpricot_ele);
   he->name = 0;
-  he->tag = he->attr = he->raw = Qnil;
+  he->tag = he->attr = he->raw = he->EC = Qnil;
   he->etag = he->parent = he->children = Qnil;
   ele = Data_Wrap_Struct(klass, hpricot_ele_mark, hpricot_ele_free, he);
   return ele;
@@ -272,7 +273,7 @@ hpricot_ele_alloc(VALUE klass)
 // the swift, compact parser logic.  most of the complicated stuff is done
 // in the lexer.  this step just pairs up the start and end tags.
 //
-VALUE
+void
 rb_hpricot_token(hpricot_state *S, VALUE sym, VALUE tag, VALUE attr, char *raw, int rawlen, int taint)
 {
   VALUE ele, ec = Qnil;
@@ -281,6 +282,16 @@ rb_hpricot_token(hpricot_state *S, VALUE sym, VALUE tag, VALUE attr, char *raw,
   // in html mode, fix up start tags incorrectly formed as empty tags
   //
   if (!S->xml) {
+    hpricot_ele *last;
+    Data_Get_Struct(S->focus, hpricot_ele, last);
+    if (last->EC == sym_CDATA &&
+       (sym != sym_procins && sym != sym_comment && sym != sym_cdata && sym != sym_text) &&
+      !(sym == sym_etag && rb_str_hash(tag) == last->name))
+    {
+      sym = sym_text;
+      tag = rb_str_new(raw, rawlen);
+    }
+
     if (sym == sym_emptytag || sym == sym_stag || sym == sym_etag) {
       ec = rb_hash_aref(S->EC, tag);
       if (NIL_P(ec)) {
@@ -300,6 +311,37 @@ rb_hpricot_token(hpricot_state *S, VALUE sym, VALUE tag, VALUE attr, char *raw,
   if (sym == sym_emptytag || sym == sym_stag) {
     H_ELE(cElem);
     he->name = rb_str_hash(tag);
+
+    if (!S->xml) {
+      VALUE match = Qnil, e = S->focus;
+      while (e != S->doc)
+      {
+        hpricot_ele *hee;
+        Data_Get_Struct(e, hpricot_ele, hee);
+
+        if (TYPE(hee->EC) == T_HASH)
+        {
+          VALUE has = rb_hash_lookup(hee->EC, INT2NUM(he->name));
+          if (has != Qnil) {
+            if (has == Qtrue) {
+              if (match == Qnil)
+                match = e;
+            } else if (has == symAllow) {
+              match = S->focus;
+            } else if (has == symDeny) {
+              match = Qnil;
+            }
+          }
+        }
+
+        e = hee->parent;
+      }
+
+      if (match == Qnil)
+        match = S->focus;
+      S->focus = match;
+    }
+
     rb_hpricot_add(S->focus, ele);
 
     //
@@ -326,6 +368,8 @@ rb_hpricot_token(hpricot_state *S, VALUE sym, VALUE tag, VALUE attr, char *raw,
     // another optimization will be to improve this very simple
     // O(n) tag search, where n is the depth of the focused tag.
     //
+    // (see also: the search above for fixups)
+    //
     name = rb_str_hash(tag);
     while (e != S->doc)
     {
@@ -371,8 +415,10 @@ rb_hpricot_token(hpricot_state *S, VALUE sym, VALUE tag, VALUE attr, char *raw,
     VALUE match = rb_funcall(tag, rb_intern("match"), 1, reProcInsParse);
     tag = rb_reg_nth_match(1, match);
     attr = rb_reg_nth_match(2, match);
-    H_ELE(cProcIns);
-    rb_hpricot_add(S->focus, ele);
+    {
+        H_ELE(cProcIns);
+        rb_hpricot_add(S->focus, ele);
+    }
   } else if (sym == sym_text) {
     // TODO: add raw_string as well?
     if (!NIL_P(S->last) && RBASIC(S->last)->klass == cText) {
@@ -420,8 +466,8 @@ VALUE hpricot_scan(int argc, VALUE *argv, VALUE self)
 
   if (!rb_block_given_p())
   {
-    S = ALLOC(hpricot_state);
     hpricot_ele *he = ALLOC(hpricot_ele);
+    S = ALLOC(hpricot_state);
     MEMZERO(he, hpricot_ele, 1);
     he->tag = he->attr = he->etag = he->parent = he->children = Qnil;
     S->doc = Data_Wrap_Struct(cDoc, hpricot_ele_mark, hpricot_ele_free, he);
@@ -432,6 +478,7 @@ VALUE hpricot_scan(int argc, VALUE *argv, VALUE self)
     S->strict = OPT(opts, xhtml_strict);
     S->fixup = OPT(opts, fixup_tags);
     if (S->strict) S->fixup = 1;
+    rb_ivar_set(S->doc, rb_intern("@options"), opts);
 
     S->EC = rb_const_get(mHpricot, s_ElementContent);
   }
@@ -463,7 +510,7 @@ VALUE hpricot_scan(int argc, VALUE *argv, VALUE self)
        mark_aval_diff = mark_aval - buf;
 
        buffer_size += BUFSIZE;
-       buf = REALLOC_N(buf, char, buffer_size);
+       REALLOC_N(buf, char, buffer_size);
 
        space = buffer_size - have;
 
@@ -477,16 +524,17 @@ VALUE hpricot_scan(int argc, VALUE *argv, VALUE self)
 
     if ( rb_respond_to( port, s_read ) )
     {
-      str = rb_funcall( port, s_read, 1, INT2FIX(space) );
+      str = rb_funcall(port, s_read, 1, INT2FIX(space));
+      len = RSTRING_LEN(str);
+      memcpy(p, StringValuePtr(str), len);
     }
     else
     {
-      str = rb_str_substr( port, nread, space );
+      len = RSTRING_LEN(port) - nread;
+      if (len > space) len = space;
+      memcpy(p, StringValuePtr(port) + nread, len);
     }
 
-    StringValue(str);
-    memcpy( p, RSTRING_PTR(str), RSTRING_LEN(str) );
-    len = RSTRING_LEN(str);
     nread += len;
 
     /* If this is the last buffer, tack on an EOF. */
@@ -627,6 +675,8 @@ void Init_hpricot_scan()
   rb_define_method(cProcIns, "content=", hpricot_ele_set_attr, 1);
 
   s_ElementContent = rb_intern("ElementContent");
+  symAllow = ID2SYM(rb_intern("allow"));
+  symDeny = ID2SYM(rb_intern("deny"));
   s_downcase = rb_intern("downcase");
   s_new = rb_intern("new");
   s_parent = rb_intern("parent");
@@ -643,6 +693,7 @@ void Init_hpricot_scan()
   sym_cdata = ID2SYM(rb_intern("cdata"));
   sym_text = ID2SYM(rb_intern("text"));
   sym_EMPTY = ID2SYM(rb_intern("EMPTY"));
+  sym_CDATA = ID2SYM(rb_intern("CDATA"));
 
   rb_const_set(mHpricot, rb_intern("ProcInsParse"),
     reProcInsParse = rb_eval_string("/\\A<\\?(\\S+)\\s+(.+)/m"));

data/lib/hpricot/builder.rb

@@ -16,7 +16,7 @@ module Hpricot
     assigns.each do |k, v|
       ele.instance_variable_set("@#{k}", v)
     end
-    ele.instance_eval &blk
+    ele.instance_eval(&blk)
     ele
   end
 
@@ -38,14 +38,21 @@ module Hpricot
       @@default[option] = value
     end
 
+    def add_child ele
+      ele.parent = self
+      self.children ||= []
+      self.children << ele
+      ele
+    end
+
     # Write a +string+ to the HTML stream, making sure to escape it.
     def text!(string)
-      (self.children ||= []) << Text.new(string.fast_xs)
+      add_child Text.new(string.fast_xs)
     end
 
     # Write a +string+ to the HTML stream without escaping it.
     def text(string)
-      (self.children ||= []) << Text.new(string)
+      add_child Text.new(string)
       nil
     end
     alias_method :<<, :text
@@ -60,11 +67,11 @@ module Hpricot
               raise InvalidXhtmlError, "no element `#{tag}' for #{tagset.doctype}"
           elsif args.last.respond_to?(:to_hash)
               attrs = args.last.to_hash
-              
+
               if @tagset.forms.include?(tag) and attrs[:id]
                 attrs[:name] ||= attrs[:id]
               end
-              
+
               attrs.each do |k, v|
                   atname = k.to_s.downcase.intern
                   unless k =~ /:/ or @tagset.tagset[tag].include? atname
@@ -106,7 +113,7 @@ module Hpricot
         build(f, &block)
       end
 
-      (self.children ||= []) << f
+      add_child f
       f
     end
 
@@ -139,11 +146,11 @@ module Hpricot
     end
 
     def doctype(target, pub, sys)
-      (self.children ||= []) << DocType.new(target, pub, sys)
+      add_child DocType.new(target, pub, sys)
     end
 
     remove_method :head
-    
+
     # Builds a head tag.  Adds a <tt>meta</tt> tag inside with Content-Type
     # set to <tt>text/html; charset=utf-8</tt>.
     def head(*args, &block)
@@ -187,7 +194,7 @@ module Hpricot
     def initialize(builder, sym)
       @builder, @sym, @attrs = builder, sym, {}
     end
-    
+
     # Adds attributes to an element.  Bang methods set the :id attribute.
     # Other methods add to the :class attribute.
     def method_missing(id_or_class, *args, &block)
@@ -201,7 +208,7 @@ module Hpricot
         args.push(@attrs)
         return @builder.tag!(@sym, *args, &block)
       end
-      
+
       return self
     end
 

data/lib/hpricot/elements.rb

@@ -275,7 +275,7 @@ module Hpricot
             expr = $'
             m.compact!
             if m[0] == '@'
-                m[0] = "@#{m.slice!(2,1)}"
+                m[0] = "@#{m.slice!(2,1).join}"
             end
 
             if m[0] == '[' && m[1] =~ /^\d+$/
@@ -300,10 +300,10 @@ module Hpricot
                         args = m[1..-1]
                     end
                 end
-                i = -1
+                args << -1
                 nodes = Elements[*nodes.find_all do |x| 
-                                      i += 1
-                                      x.send(meth, *([*args] + [i])) ? truth : !truth
+                                      args[-1] += 1
+                                      x.send(meth, *args) ? truth : !truth
                                   end]
             end
         end
@@ -446,23 +446,23 @@ module Hpricot
       parent.containers.length == 1
     end
 
-    filter :parent do
+    filter :parent do |*a|
       containers.length > 0
     end
 
-    filter :empty do
+    filter :empty do |*a|
       containers.length == 0
     end
 
-    filter :root do
+    filter :root do |*a|
       self.is_a? Hpricot::Doc
     end
     
-    filter 'text' do
+    filter 'text' do |*a|
       self.text?
     end
 
-    filter 'comment' do
+    filter 'comment' do |*a|
       self.comment?
     end
 
@@ -495,7 +495,7 @@ module Hpricot
     end
 
     filter 'text()' do |val,i|
-      !self.inner_text.strip.empty?
+      self.children.grep(Hpricot::Text).detect { |x| x.content =~ /\S/ } if self.children
     end
 
     filter '@' do |attr,val,i|

data/lib/hpricot/htmlinfo.rb

@@ -473,9 +473,23 @@ module Hpricot
    "menu", "noframes", "noscript", "object", "ol", "p", "pre", "q", "s",
    "samp", "script", "select", "small", "span", "strike", "strong", "sub",
    "sup", "table", "textarea", "tt", "u", "ul", "var"]}
+  ElementContent.keys.each do |k|
+    v = ElementContent[k]
+    if v.is_a? Array
+      ElementContent[k] = v.inject({}) do |h, name|
+        h[name.hash] = true
+        h
+      end
+    end
+  end
 
   ElementInclusions =
 {"head"=>["link", "meta", "object", "script", "style"], "body"=>["del", "ins"]}
+  ElementInclusions.each do |k, v|
+    v.each do |name|
+      ElementContent[k][name.hash] = :allow
+    end
+  end
 
   ElementExclusions =
 {"button"=>
@@ -496,6 +510,11 @@ module Hpricot
    "h1", "h2", "h3", "h4", "h5", "h6", "hr", "isindex", "menu", "noframes",
    "noscript", "ol", "p", "pre", "table", "ul"],
  "label"=>["label"]}
+  ElementExclusions.each do |k, v|
+    v.each do |name|
+      ElementContent[k][name.hash] = :deny
+    end
+  end
 
   OmittedAttrName =
 {"h6"=>

data/lib/hpricot/parse.rb

@@ -26,7 +26,9 @@ module Hpricot
 
   def Hpricot.make(input = nil, opts = {}, &blk)
     if blk
-      Hpricot.build(&blk)
+      doc = Hpricot.build(&blk)
+      doc.instance_variable_set("@options", opts)
+      doc
     else
       Hpricot.scan(input, opts)
     end

data/lib/hpricot/tag.rb

@@ -12,6 +12,9 @@ module Hpricot
       Hpricot.make(input, @options, &blk).children
     end
     def altered!; end
+    def inspect_tree
+      children.map { |x| x.inspect_tree }.join if children
+    end
   end
 
   class BaseEle
@@ -29,6 +32,9 @@ module Hpricot
     def altered!
       clear_raw
     end
+    def inspect_tree(depth = 0)
+      %{#{" " * depth}} + self.class.name.split(/::/).last.downcase + "\n"
+    end
   end
 
   class Elem
@@ -43,6 +49,8 @@ module Hpricot
           hsh[k] = Hpricot.uxs(v)
           hsh
         end
+      else
+        {}
       end
     end
     def to_plain_text
@@ -87,6 +95,10 @@ module Hpricot
         end.join
       end
     end
+    def inspect_tree(depth = 0)
+      %{#{" " * depth}} + name + "\n" +
+        (children ? children.map { |x| x.inspect_tree(depth + 1) }.join : "")
+    end
   end
 
   class ETag
@@ -115,7 +127,7 @@ module Hpricot
     def output(out, opts = {})
       out <<
         if_output(opts) do
-          content
+          content.to_s
         end
     end
   end

data/test/files/boingboing.html

@@ -2263,4 +2263,4 @@ Why are so many drawings from earlier centuries so deliciously weird? Here are a
 -->
     </div>
 </body>
-</html>
+</html>

data/test/test_builder.rb

@@ -1,3 +1,4 @@
+# -*- coding: utf-8 -*-
 #!/usr/bin/env ruby
 
 require 'test/unit'
@@ -17,9 +18,9 @@ class TestBuilder < Test::Unit::TestCase
   end
 
   def test_latin1_entities
-    doc = Hpricot() { b "\200\225" }
+    doc = Hpricot() { b "ۥ" }
     assert_equal "<b>&#8364;&#8226;</b>", doc.to_html
-    assert_equal "\342\202\254\342\200\242", doc.at("text()").to_s
+    assert_equal "ۥ", doc.at("text()").to_s
   end
 
   def test_escaping_attrs
@@ -29,8 +30,7 @@ class TestBuilder < Test::Unit::TestCase
   end
 
   def test_korean_utf8_entities
-    # a = '한글'
-    a = "\xed\x95\x9c\xea\xb8\x80"
+    a = '한글'
     doc = Hpricot() { b a }
     assert_equal "<b>&#54620;&#44544;</b>", doc.to_html
   end

data/test/test_parser.rb

@@ -1,3 +1,4 @@
+# -*- coding: utf-8 -*-
 #!/usr/bin/env ruby
 
 require 'test/unit'
@@ -121,7 +122,7 @@ class TestParser < Test::Unit::TestCase
     assert_equal 60, @boingboing.search("h3").length
     assert_equal 59, @boingboing.search("h3[text()!='College kids reportedly taking more smart drugs']").length
     assert_equal 17, @boingboing.search("h3[text()$='s']").length
-    assert_equal 129, @boingboing.search("p[text()]").length
+    assert_equal 116, @boingboing.search("p[text()]").length
     assert_equal 211, @boingboing.search("p").length
   end
 
@@ -161,10 +162,10 @@ class TestParser < Test::Unit::TestCase
     assert_equal 60, @boingboing.search("/*/body//p[@class='posted']").length
     assert_equal 18, @boingboing.search("//script").length
     divs = @boingboing.search("//script/../div")
-    assert_equal 1,  divs.length
+    assert_equal 2,  divs.length
     imgs = @boingboing.search('//div/p/a/img')
-    assert_equal 15, imgs.length
-    assert_equal 17, @boingboing.search('//div').search('p/a/img').length
+    assert_equal 16, imgs.length
+    assert_equal 16, @boingboing.search('//div').search('p/a/img').length
     assert imgs.all? { |x| x.name == 'img' }
   end
 
@@ -172,10 +173,10 @@ class TestParser < Test::Unit::TestCase
     @boingboing = Hpricot.parse(TestFiles::BOINGBOING)
     assert_equal 2, @boingboing.search('//link[@rel="alternate"]').length
     p_imgs = @boingboing.search('//div/p[/a/img]')
-    assert_equal 15, p_imgs.length
+    assert_equal 16, p_imgs.length
     assert p_imgs.all? { |x| x.name == 'p' }
     p_imgs = @boingboing.search('//div/p[a/img]')
-    assert_equal 18, p_imgs.length
+    assert_equal 16, p_imgs.length
     assert p_imgs.all? { |x| x.name == 'p' }
     assert_equal 1, @boingboing.search('//input[@checked]').length
   end
@@ -218,7 +219,7 @@ class TestParser < Test::Unit::TestCase
   def test_many_paths
     @boingboing = Hpricot.parse(TestFiles::BOINGBOING)
     assert_equal 62, @boingboing.search('p.posted, link[@rel="alternate"]').length
-    assert_equal 20, @boingboing.search('//div/p[a/img]|//link[@rel="alternate"]').length
+    assert_equal 18, @boingboing.search('//div/p[a/img]|//link[@rel="alternate"]').length
   end
 
   def test_stacked_search
@@ -391,10 +392,18 @@ class TestParser < Test::Unit::TestCase
   end
 
   def test_uxs_handles_numeric_values
-    assert_equal "\303\251", Hpricot.uxs('&#233;')
+    if String.method_defined? :encoding
+      assert_equal "é", Hpricot.uxs('&#233;')
+    else
+      assert_equal "\303\251", Hpricot.uxs('&#233;')
+    end
   end
 
   def test_uxs_handles_entities
-    assert_equal "\303\251", Hpricot.uxs('&eacute;')
+    if String.method_defined? :encoding
+      assert_equal "é", Hpricot.uxs('&eacute;')
+    else
+      assert_equal "\303\251", Hpricot.uxs('&eacute;')
+    end
   end
 end

data/test/test_preserved.rb

@@ -1,3 +1,4 @@
+# -*- coding: utf-8 -*-
 #!/usr/bin/env ruby
 
 require 'test/unit'
@@ -9,7 +10,10 @@ class TestPreserved < Test::Unit::TestCase
     doc = Hpricot(str)
     yield doc if block_given?
     str2 = doc.to_original_html
-    [*str].zip([*str2]).each do |s1, s2|
+    if RUBY_VERSION =~ /^1.9/
+      str2.force_encoding('UTF-8')
+    end
+    str.lines.zip(str2.lines).each do |s1, s2|
       assert_equal s1, s2
     end
   end
@@ -40,7 +44,7 @@ class TestPreserved < Test::Unit::TestCase
 
   def test_escaping_of_contents
     doc = Hpricot(TestFiles::BOINGBOING)
-    assert_equal "Fukuda\342\200\231s Automatic Door opens around your body as you pass through it. The idea is to save energy and keep the room clean.", doc.at("img[@alt='200606131240']").next.to_s.strip
+    assert_equal "Fukuda’s Automatic Door opens around your body as you pass through it. The idea is to save energy and keep the room clean.", doc.at("img[@alt='200606131240']").next.to_s.strip
   end
 
   def test_files

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification 
 name: adamh-hpricot
 version: !ruby/object:Gem::Version 
-  version: 0.6.171
+  version: 0.6.210
 platform: ruby
 authors: 
 - why the lucky stiff
@@ -9,7 +9,7 @@ autorequire:
 bindir: bin
 cert_chain: []
 
-date: 2008-12-11 00:00:00 -08:00
+date: 2009-02-07 00:00:00 -08:00
 default_executable: 
 dependencies: []