Big refactoring into multiple files.

8 files changed, 1410 insertions, 1367 deletions

diff --git a/BeautifulSoup.py b/BeautifulSoup.py
index 4d40c34..0999c07 100644
--- a/BeautifulSoup.py
+++ b/BeautifulSoup.py
@@ -79,1335 +79,26 @@ __version__ = "4.0.0"
 __copyright__ = "Copyright (c) 2004-2009 Leonard Richardson"
 __license__ = "New-style BSD"
 
-import markupbase
-import types
-import re
-from HTMLParser import HTMLParser, HTMLParseError
-try:
-    from htmlentitydefs import name2codepoint
-except ImportError:
-    name2codepoint = {}
-try:
-    set
-except NameError:
-    from sets import Set as set
-from dammit import UnicodeDammit
-
-#These hacks make Beautiful Soup able to parse XML with namespaces
-markupbase._declname_match = re.compile(r'[a-zA-Z][-_.:a-zA-Z0-9]*\s*').match
-
-DEFAULT_OUTPUT_ENCODING = "utf-8"
-
-# First, the classes that represent markup elements.
-
-class Entities:
-    """Knows about XML entities."""
-
-    HTML_ENTITIES = "html"
-    XML_ENTITIES = "xml"
-    XHTML_ENTITIES = "xhtml"
-
-    def _invert(h):
-        "Cheap function to invert a hash."
-        i = {}
-        for k,v in h.items():
-            i[v] = k
-        return i
-
-    XML_ENTITIES_TO_SPECIAL_CHARS = { "apos" : "'",
-                                      "quot" : '"',
-                                      "amp" : "&",
-                                      "lt" : "<",
-                                      "gt" : ">" }
-
-    XML_SPECIAL_CHARS_TO_ENTITIES = _invert(XML_ENTITIES_TO_SPECIAL_CHARS)
-
-
-class PageElement:
-    """Contains the navigational information for some part of the page
-    (either a tag or a piece of text)"""
-
-    def setup(self, parent=None, previous=None):
-        """Sets up the initial relations between this element and
-        other elements."""
-        self.parent = parent
-        self.previous = previous
-        self.next = None
-        self.previousSibling = None
-        self.nextSibling = None
-        if self.parent and self.parent.contents:
-            self.previousSibling = self.parent.contents[-1]
-            self.previousSibling.nextSibling = self
-
-    def replaceWith(self, replaceWith):
-        oldParent = self.parent
-        myIndex = self.parent.contents.index(self)
-        if hasattr(replaceWith, 'parent') and replaceWith.parent == self.parent:
-            # We're replacing this element with one of its siblings.
-            index = self.parent.contents.index(replaceWith)
-            if index and index < myIndex:
-                # Furthermore, it comes before this element. That
-                # means that when we extract it, the index of this
-                # element will change.
-                myIndex = myIndex - 1
-        self.extract()
-        oldParent.insert(myIndex, replaceWith)
-
-    def extract(self):
-        """Destructively rips this element out of the tree."""
-        if self.parent:
-            try:
-                self.parent.contents.remove(self)
-            except ValueError:
-                pass
-
-        #Find the two elements that would be next to each other if
-        #this element (and any children) hadn't been parsed. Connect
-        #the two.
-        lastChild = self._lastRecursiveChild()
-        nextElement = lastChild.next
-
-        if self.previous:
-            self.previous.next = nextElement
-        if nextElement:
-            nextElement.previous = self.previous
-        self.previous = None
-        lastChild.next = None
-
-        self.parent = None
-        if self.previousSibling:
-            self.previousSibling.nextSibling = self.nextSibling
-        if self.nextSibling:
-            self.nextSibling.previousSibling = self.previousSibling
-        self.previousSibling = self.nextSibling = None
-        return self
-
-    def _lastRecursiveChild(self):
-        "Finds the last element beneath this object to be parsed."
-        lastChild = self
-        while hasattr(lastChild, 'contents') and lastChild.contents:
-            lastChild = lastChild.contents[-1]
-        return lastChild
-
-    def insert(self, position, newChild):
-        if (isinstance(newChild, basestring)
-            or isinstance(newChild, unicode)) \
-            and not isinstance(newChild, NavigableString):
-            newChild = NavigableString(newChild)
-
-        position =  min(position, len(self.contents))
-        if hasattr(newChild, 'parent') and newChild.parent != None:
-            # We're 'inserting' an element that's already one
-            # of this object's children.
-            if newChild.parent == self:
-                index = self.find(newChild)
-                if index and index < position:
-                    # Furthermore we're moving it further down the
-                    # list of this object's children. That means that
-                    # when we extract this element, our target index
-                    # will jump down one.
-                    position = position - 1
-            newChild.extract()
-
-        newChild.parent = self
-        previousChild = None
-        if position == 0:
-            newChild.previousSibling = None
-            newChild.previous = self
-        else:
-            previousChild = self.contents[position-1]
-            newChild.previousSibling = previousChild
-            newChild.previousSibling.nextSibling = newChild
-            newChild.previous = previousChild._lastRecursiveChild()
-        if newChild.previous:
-            newChild.previous.next = newChild
-
-        newChildsLastElement = newChild._lastRecursiveChild()
-
-        if position >= len(self.contents):
-            newChild.nextSibling = None
-
-            parent = self
-            parentsNextSibling = None
-            while not parentsNextSibling:
-                parentsNextSibling = parent.nextSibling
-                parent = parent.parent
-                if not parent: # This is the last element in the document.
-                    break
-            if parentsNextSibling:
-                newChildsLastElement.next = parentsNextSibling
-            else:
-                newChildsLastElement.next = None
-        else:
-            nextChild = self.contents[position]
-            newChild.nextSibling = nextChild
-            if newChild.nextSibling:
-                newChild.nextSibling.previousSibling = newChild
-            newChildsLastElement.next = nextChild
-
-        if newChildsLastElement.next:
-            newChildsLastElement.next.previous = newChildsLastElement
-        self.contents.insert(position, newChild)
-
-    def append(self, tag):
-        """Appends the given tag to the contents of this tag."""
-        self.insert(len(self.contents), tag)
-
-    def findNext(self, name=None, attrs={}, text=None, **kwargs):
-        """Returns the first item that matches the given criteria and
-        appears after this Tag in the document."""
-        return self._findOne(self.findAllNext, name, attrs, text, **kwargs)
-
-    def findAllNext(self, name=None, attrs={}, text=None, limit=None,
-                    **kwargs):
-        """Returns all items that match the given criteria and appear
-        after this Tag in the document."""
-        return self._findAll(name, attrs, text, limit, self.nextGenerator,
-                             **kwargs)
-
-    def findNextSibling(self, name=None, attrs={}, text=None, **kwargs):
-        """Returns the closest sibling to this Tag that matches the
-        given criteria and appears after this Tag in the document."""
-        return self._findOne(self.findNextSiblings, name, attrs, text,
-                             **kwargs)
-
-    def findNextSiblings(self, name=None, attrs={}, text=None, limit=None,
-                         **kwargs):
-        """Returns the siblings of this Tag that match the given
-        criteria and appear after this Tag in the document."""
-        return self._findAll(name, attrs, text, limit,
-                             self.nextSiblingGenerator, **kwargs)
-    fetchNextSiblings = findNextSiblings # Compatibility with pre-3.x
-
-    def findPrevious(self, name=None, attrs={}, text=None, **kwargs):
-        """Returns the first item that matches the given criteria and
-        appears before this Tag in the document."""
-        return self._findOne(self.findAllPrevious, name, attrs, text, **kwargs)
-
-    def findAllPrevious(self, name=None, attrs={}, text=None, limit=None,
-                        **kwargs):
-        """Returns all items that match the given criteria and appear
-        before this Tag in the document."""
-        return self._findAll(name, attrs, text, limit, self.previousGenerator,
-                           **kwargs)
-    fetchPrevious = findAllPrevious # Compatibility with pre-3.x
-
-    def findPreviousSibling(self, name=None, attrs={}, text=None, **kwargs):
-        """Returns the closest sibling to this Tag that matches the
-        given criteria and appears before this Tag in the document."""
-        return self._findOne(self.findPreviousSiblings, name, attrs, text,
-                             **kwargs)
-
-    def findPreviousSiblings(self, name=None, attrs={}, text=None,
-                             limit=None, **kwargs):
-        """Returns the siblings of this Tag that match the given
-        criteria and appear before this Tag in the document."""
-        return self._findAll(name, attrs, text, limit,
-                             self.previousSiblingGenerator, **kwargs)
-    fetchPreviousSiblings = findPreviousSiblings # Compatibility with pre-3.x
-
-    def findParent(self, name=None, attrs={}, **kwargs):
-        """Returns the closest parent of this Tag that matches the given
-        criteria."""
-        # NOTE: We can't use _findOne because findParents takes a different
-        # set of arguments.
-        r = None
-        l = self.findParents(name, attrs, 1)
-        if l:
-            r = l[0]
-        return r
-
-    def findParents(self, name=None, attrs={}, limit=None, **kwargs):
-        """Returns the parents of this Tag that match the given
-        criteria."""
-
-        return self._findAll(name, attrs, None, limit, self.parentGenerator,
-                             **kwargs)
-    fetchParents = findParents # Compatibility with pre-3.x
-
-    #These methods do the real heavy lifting.
-
-    def _findOne(self, method, name, attrs, text, **kwargs):
-        r = None
-        l = method(name, attrs, text, 1, **kwargs)
-        if l:
-            r = l[0]
-        return r
-
-    def _findAll(self, name, attrs, text, limit, generator, **kwargs):
-        "Iterates over a generator looking for things that match."
-
-        if isinstance(name, SoupStrainer):
-            strainer = name
-        else:
-            # Build a SoupStrainer
-            strainer = SoupStrainer(name, attrs, text, **kwargs)
-        results = ResultSet(strainer)
-        g = generator()
-        while True:
-            try:
-                i = g.next()
-            except StopIteration:
-                break
-            if i:
-                found = strainer.search(i)
-                if found:
-                    results.append(found)
-                    if limit and len(results) >= limit:
-                        break
-        return results
-
-    #These Generators can be used to navigate starting from both
-    #NavigableStrings and Tags.
-    def nextGenerator(self):
-        i = self
-        while i:
-            i = i.next
-            yield i
-
-    def nextSiblingGenerator(self):
-        i = self
-        while i:
-            i = i.nextSibling
-            yield i
-
-    def previousGenerator(self):
-        i = self
-        while i:
-            i = i.previous
-            yield i
-
-    def previousSiblingGenerator(self):
-        i = self
-        while i:
-            i = i.previousSibling
-            yield i
-
-    def parentGenerator(self):
-        i = self
-        while i:
-            i = i.parent
-            yield i
-
-    # Utility methods
-    def substituteEncoding(self, str, encoding=None):
-        encoding = encoding or "utf-8"
-        return str.replace("%SOUP-ENCODING%", encoding)
-
-    def toEncoding(self, s, encoding=None):
-        """Encodes an object to a string in some encoding, or to Unicode.
-        ."""
-        if isinstance(s, unicode):
-            if encoding:
-                s = s.encode(encoding)
-        elif isinstance(s, str):
-            if encoding:
-                s = s.encode(encoding)
-            else:
-                s = unicode(s)
-        else:
-            if encoding:
-                s  = self.toEncoding(str(s), encoding)
-            else:
-                s = unicode(s)
-        return s
-
-class NavigableString(unicode, PageElement):
-
-    def __new__(cls, value):
-        """Create a new NavigableString.
-
-        When unpickling a NavigableString, this method is called with
-        the string in DEFAULT_OUTPUT_ENCODING. That encoding needs to be
-        passed in to the superclass's __new__ or the superclass won't know
-        how to handle non-ASCII characters.
-        """
-        if isinstance(value, unicode):
-            return unicode.__new__(cls, value)
-        return unicode.__new__(cls, value, DEFAULT_OUTPUT_ENCODING)
-
-    def __getnewargs__(self):
-        return (unicode(self),)
-
-    def __getattr__(self, attr):
-        """text.string gives you text. This is for backwards
-        compatibility for Navigable*String, but for CData* it lets you
-        get the string without the CData wrapper."""
-        if attr == 'string':
-            return self
-        else:
-            raise AttributeError, "'%s' object has no attribute '%s'" % (self.__class__.__name__, attr)
-
-    def encode(self, encoding=DEFAULT_OUTPUT_ENCODING):
-        return self.decode().encode(encoding)
-
-    def decodeGivenEventualEncoding(self, eventualEncoding):
-        return self
-
-class CData(NavigableString):
-
-    def decodeGivenEventualEncoding(self, eventualEncoding):
-        return u'<![CDATA[' + self + u']]>'
-
-class ProcessingInstruction(NavigableString):
-
-    def decodeGivenEventualEncoding(self, eventualEncoding):
-        output = self
-        if u'%SOUP-ENCODING%' in output:
-            output = self.substituteEncoding(output, eventualEncoding)
-        return u'<?' + output + u'?>'
-
-class Comment(NavigableString):
-    def decodeGivenEventualEncoding(self, eventualEncoding):
-        return u'<!--' + self + u'-->'
-
-class Declaration(NavigableString):
-    def decodeGivenEventualEncoding(self, eventualEncoding):
-        return u'<!' + self + u'>'
-
-class Tag(PageElement, Entities):
-
-    """Represents a found HTML tag with its attributes and contents."""
-
-    def _convertEntities(self, builder, match):
-        """Used in a call to re.sub to replace HTML, XML, and numeric
-        entities with the appropriate Unicode characters. If HTML
-        entities are being converted, any unrecognized entities are
-        escaped."""
-        x = match.group(1)
-        if builder.convertHTMLEntities and x in name2codepoint:
-            return unichr(name2codepoint[x])
-        elif x in self.XML_ENTITIES_TO_SPECIAL_CHARS:
-            if builder.convertXMLEntities:
-                return self.XML_ENTITIES_TO_SPECIAL_CHARS[x]
-            else:
-                return u'&%s;' % x
-        elif len(x) > 0 and x[0] == '#':
-            # Handle numeric entities
-            if len(x) > 1 and x[1] == 'x':
-                return unichr(int(x[2:], 16))
-            else:
-                return unichr(int(x[1:]))
-
-        elif self.escapeUnrecognizedEntities:
-            return u'&amp;%s;' % x
-        else:
-            return u'&%s;' % x
-
-    def __init__(self, parser, builder, name, attrs=None, parent=None,
-                 previous=None):
-        "Basic constructor."
-
-        # We don't actually store the parser object: that lets extracted
-        # chunks be garbage-collected
-        self.parserClass = parser.__class__
-        self.name = name
-        self.isSelfClosing = builder.isSelfClosingTag(name)
-        if attrs == None:
-            attrs = []
-        if isinstance(attrs, types.DictType):
-            self.attrMap = attrs
-        self.attrs = attrs
-        self.contents = []
-        self.setup(parent, previous)
-        self.hidden = False
-        self.containsSubstitutions = False
-        self.escapeUnrecognizedEntities = parser.escapeUnrecognizedEntities
-
-        # Convert any HTML, XML, or numeric entities in the attribute values.
-        convert_one = lambda x: self._convertEntities(parser.builder, x)
-        def convert(kval):
-            k, val = kval
-            if val is None:
-                return kval
-            return (k, re.sub("&(#\d+|#x[0-9a-fA-F]+|\w+);", convert_one, val))
-        if isinstance(attrs, types.DictType):
-            self.attrs = [convert(kv) for kv in attrs.items()]
-        else:
-            self.attrs = map(convert, attrs)
-
-    def get(self, key, default=None):
-        """Returns the value of the 'key' attribute for the tag, or
-        the value given for 'default' if it doesn't have that
-        attribute."""
-        return self._getAttrMap().get(key, default)
-
-    def has_key(self, key):
-        return self._getAttrMap().has_key(key)
-
-    def __getitem__(self, key):
-        """tag[key] returns the value of the 'key' attribute for the tag,
-        and throws an exception if it's not there."""
-        return self._getAttrMap()[key]
-
-    def __iter__(self):
-        "Iterating over a tag iterates over its contents."
-        return iter(self.contents)
-
-    def __len__(self):
-        "The length of a tag is the length of its list of contents."
-        return len(self.contents)
-
-    def __contains__(self, x):
-        return x in self.contents
-
-    def __nonzero__(self):
-        "A tag is non-None even if it has no contents."
-        return True
-
-    def __setitem__(self, key, value):
-        """Setting tag[key] sets the value of the 'key' attribute for the
-        tag."""
-        self._getAttrMap()
-        self.attrMap[key] = value
-        found = False
-        for i in range(0, len(self.attrs)):
-            if self.attrs[i][0] == key:
-                self.attrs[i] = (key, value)
-                found = True
-        if not found:
-            self.attrs.append((key, value))
-        self._getAttrMap()[key] = value
-
-    def __delitem__(self, key):
-        "Deleting tag[key] deletes all 'key' attributes for the tag."
-        for item in self.attrs:
-            if item[0] == key:
-                self.attrs.remove(item)
-                #We don't break because bad HTML can define the same
-                #attribute multiple times.
-            self._getAttrMap()
-            if self.attrMap.has_key(key):
-                del self.attrMap[key]
-
-    def __call__(self, *args, **kwargs):
-        """Calling a tag like a function is the same as calling its
-        findAll() method. Eg. tag('a') returns a list of all the A tags
-        found within this tag."""
-        return apply(self.findAll, args, kwargs)
-
-    def __getattr__(self, tag):
-        #print "Getattr %s.%s" % (self.__class__, tag)
-        if len(tag) > 3 and tag.rfind('Tag') == len(tag)-3:
-            return self.find(tag[:-3])
-        elif tag.find('__') != 0:
-            return self.find(tag)
-        raise AttributeError, "'%s' object has no attribute '%s'" % (self.__class__, tag)
-
-    def __eq__(self, other):
-        """Returns true iff this tag has the same name, the same attributes,
-        and the same contents (recursively) as the given tag.
-
-        NOTE: right now this will return false if two tags have the
-        same attributes in a different order. Should this be fixed?"""
-        if not hasattr(other, 'name') or not hasattr(other, 'attrs') or not hasattr(other, 'contents') or self.name != other.name or self.attrs != other.attrs or len(self) != len(other):
-            return False
-        for i in range(0, len(self.contents)):
-            if self.contents[i] != other.contents[i]:
-                return False
-        return True
-
-    def __ne__(self, other):
-        """Returns true iff this tag is not identical to the other tag,
-        as defined in __eq__."""
-        return not self == other
-
-    def __repr__(self, encoding=DEFAULT_OUTPUT_ENCODING):
-        """Renders this tag as a string."""
-        return self.decode(eventualEncoding=encoding)
-
-    BARE_AMPERSAND_OR_BRACKET = re.compile("([<>]|"
-                                           + "&(?!#\d+;|#x[0-9a-fA-F]+;|\w+;)"
-                                           + ")")
-
-    def _sub_entity(self, x):
-        """Used with a regular expression to substitute the
-        appropriate XML entity for an XML special character."""
-        return "&" + self.XML_SPECIAL_CHARS_TO_ENTITIES[x.group(0)[0]] + ";"
-
-    def __unicode__(self):
-        return self.decode()
-
-    def __str__(self):
-        return self.encode()
-
-    def encode(self, encoding=DEFAULT_OUTPUT_ENCODING,
-               prettyPrint=False, indentLevel=0):
-        return self.decode(prettyPrint, indentLevel, encoding).encode(encoding)
-
-    def decode(self, prettyPrint=False, indentLevel=0,
-               eventualEncoding=DEFAULT_OUTPUT_ENCODING):
-        """Returns a string or Unicode representation of this tag and
-        its contents. To get Unicode, pass None for encoding."""
-
-        attrs = []
-        if self.attrs:
-            for key, val in self.attrs:
-                fmt = '%s="%s"'
-                if isString(val):
-                    if (self.containsSubstitutions
-                        and eventualEncoding is not None
-                        and '%SOUP-ENCODING%' in val):
-                        val = self.substituteEncoding(val, eventualEncoding)
-
-                    # The attribute value either:
-                    #
-                    # * Contains no embedded double quotes or single quotes.
-                    #   No problem: we enclose it in double quotes.
-                    # * Contains embedded single quotes. No problem:
-                    #   double quotes work here too.
-                    # * Contains embedded double quotes. No problem:
-                    #   we enclose it in single quotes.
-                    # * Embeds both single _and_ double quotes. This
-                    #   can't happen naturally, but it can happen if
-                    #   you modify an attribute value after parsing
-                    #   the document. Now we have a bit of a
-                    #   problem. We solve it by enclosing the
-                    #   attribute in single quotes, and escaping any
-                    #   embedded single quotes to XML entities.
-                    if '"' in val:
-                        fmt = "%s='%s'"
-                        if "'" in val:
-                            # TODO: replace with apos when
-                            # appropriate.
-                            val = val.replace("'", "&squot;")
-
-                    # Now we're okay w/r/t quotes. But the attribute
-                    # value might also contain angle brackets, or
-                    # ampersands that aren't part of entities. We need
-                    # to escape those to XML entities too.
-                    val = self.BARE_AMPERSAND_OR_BRACKET.sub(self._sub_entity, val)
-                if val is None:
-                    # Handle boolean attributes.
-                    decoded = key
-                else:
-                    decoded = fmt % (key, val)
-                attrs.append(decoded)
-        close = ''
-        closeTag = ''
-        if self.isSelfClosing:
-            close = ' /'
-        else:
-            closeTag = '</%s>' % self.name
-
-        indentTag, indentContents = 0, 0
-        if prettyPrint:
-            indentTag = indentLevel
-            space = (' ' * (indentTag-1))
-            indentContents = indentTag + 1
-        contents = self.decodeContents(prettyPrint, indentContents,
-                                       eventualEncoding)
-        if self.hidden:
-            s = contents
-        else:
-            s = []
-            attributeString = ''
-            if attrs:
-                attributeString = ' ' + ' '.join(attrs)
-            if prettyPrint:
-                s.append(space)
-            s.append('<%s%s%s>' % (self.name, attributeString, close))
-            if prettyPrint:
-                s.append("\n")
-            s.append(contents)
-            if prettyPrint and contents and contents[-1] != "\n":
-                s.append("\n")
-            if prettyPrint and closeTag:
-                s.append(space)
-            s.append(closeTag)
-            if prettyPrint and closeTag and self.nextSibling:
-                s.append("\n")
-            s = ''.join(s)
-        return s
-
-    def decompose(self):
-        """Recursively destroys the contents of this tree."""
-        contents = [i for i in self.contents]
-        for i in contents:
-            if isinstance(i, Tag):
-                i.decompose()
-            else:
-                i.extract()
-        self.extract()
-
-    def prettify(self, encoding=DEFAULT_OUTPUT_ENCODING):
-        return self.encode(encoding, True)
-
-    def encodeContents(self, encoding=DEFAULT_OUTPUT_ENCODING,
-                       prettyPrint=False, indentLevel=0):
-        return self.decodeContents(prettyPrint, indentLevel).encode(encoding)
-
-    def decodeContents(self, prettyPrint=False, indentLevel=0,
-                       eventualEncoding=DEFAULT_OUTPUT_ENCODING):
-        """Renders the contents of this tag as a string in the given
-        encoding. If encoding is None, returns a Unicode string.."""
-        s=[]
-        for c in self:
-            text = None
-            if isinstance(c, NavigableString):
-                text = c.decodeGivenEventualEncoding(eventualEncoding)
-            elif isinstance(c, Tag):
-                s.append(c.decode(prettyPrint, indentLevel, eventualEncoding))
-            if text and prettyPrint:
-                text = text.strip()
-            if text:
-                if prettyPrint:
-                    s.append(" " * (indentLevel-1))
-                s.append(text)
-                if prettyPrint:
-                    s.append("\n")
-        return ''.join(s)
-
-    #Soup methods
-
-    def find(self, name=None, attrs={}, recursive=True, text=None,
-             **kwargs):
-        """Return only the first child of this Tag matching the given
-        criteria."""
-        r = None
-        l = self.findAll(name, attrs, recursive, text, 1, **kwargs)
-        if l:
-            r = l[0]
-        return r
-    findChild = find
-
-    def findAll(self, name=None, attrs={}, recursive=True, text=None,
-                limit=None, **kwargs):
-        """Extracts a list of Tag objects that match the given
-        criteria.  You can specify the name of the Tag and any
-        attributes you want the Tag to have.
-
-        The value of a key-value pair in the 'attrs' map can be a
-        string, a list of strings, a regular expression object, or a
-        callable that takes a string and returns whether or not the
-        string matches for some custom definition of 'matches'. The
-        same is true of the tag name."""
-        generator = self.recursiveChildGenerator
-        if not recursive:
-            generator = self.childGenerator
-        return self._findAll(name, attrs, text, limit, generator, **kwargs)
-    findChildren = findAll
-
-    # Pre-3.x compatibility methods. Will go away in 4.0.
-    first = find
-    fetch = findAll
-
-    def fetchText(self, text=None, recursive=True, limit=None):
-        return self.findAll(text=text, recursive=recursive, limit=limit)
-
-    def firstText(self, text=None, recursive=True):
-        return self.find(text=text, recursive=recursive)
-
-    # 3.x compatibility methods. Will go away in 4.0.
-    def renderContents(self, encoding=DEFAULT_OUTPUT_ENCODING,
-                       prettyPrint=False, indentLevel=0):
-        if encoding is None:
-            return self.decodeContents(prettyPrint, indentLevel, encoding)
-        else:
-            return self.encodeContents(encoding, prettyPrint, indentLevel)
-
-
-    #Private methods
-
-    def _getAttrMap(self):
-        """Initializes a map representation of this tag's attributes,
-        if not already initialized."""
-        if not getattr(self, 'attrMap'):
-            self.attrMap = {}
-            for (key, value) in self.attrs:
-                self.attrMap[key] = value
-        return self.attrMap
-
-    #Generator methods
-    def childGenerator(self):
-        for i in range(0, len(self.contents)):
-            yield self.contents[i]
-        raise StopIteration
-
-    def recursiveChildGenerator(self):
-        if not len(self.contents):
-            raise StopIteration
-        stopNode = self._lastRecursiveChild().next
-        current = self.contents[0]
-        while current is not stopNode:
-            yield current
-            current = current.next
-
-# Next, a couple classes to represent queries and their results.
-class SoupStrainer:
-    """Encapsulates a number of ways of matching a markup element (tag or
-    text)."""
-
-    def __init__(self, name=None, attrs={}, text=None, **kwargs):
-        self.name = name
-        if isString(attrs):
-            kwargs['class'] = attrs
-            attrs = None
-        if kwargs:
-            if attrs:
-                attrs = attrs.copy()
-                attrs.update(kwargs)
-            else:
-                attrs = kwargs
-        self.attrs = attrs
-        self.text = text
-
-    def __str__(self):
-        if self.text:
-            return self.text
-        else:
-            return "%s|%s" % (self.name, self.attrs)
-
-    def searchTag(self, markupName=None, markupAttrs={}):
-        found = None
-        markup = None
-        if isinstance(markupName, Tag):
-            markup = markupName
-            markupAttrs = markup
-        callFunctionWithTagData = callable(self.name) \
-                                and not isinstance(markupName, Tag)
-
-        if (not self.name) \
-               or callFunctionWithTagData \
-               or (markup and self._matches(markup, self.name)) \
-               or (not markup and self._matches(markupName, self.name)):
-            if callFunctionWithTagData:
-                match = self.name(markupName, markupAttrs)
-            else:
-                match = True
-                markupAttrMap = None
-                for attr, matchAgainst in self.attrs.items():
-                    if not markupAttrMap:
-                         if hasattr(markupAttrs, 'get'):
-                            markupAttrMap = markupAttrs
-                         else:
-                            markupAttrMap = {}
-                            for k,v in markupAttrs:
-                                markupAttrMap[k] = v
-                    attrValue = markupAttrMap.get(attr)
-                    if not self._matches(attrValue, matchAgainst):
-                        match = False
-                        break
-            if match:
-                if markup:
-                    found = markup
-                else:
-                    found = markupName
-        return found
-
-    def search(self, markup):
-        #print 'looking for %s in %s' % (self, markup)
-        found = None
-        # If given a list of items, scan it for a text element that
-        # matches.
-        if isList(markup) and not isinstance(markup, Tag):
-            for element in markup:
-                if isinstance(element, NavigableString) \
-                       and self.search(element):
-                    found = element
-                    break
-        # If it's a Tag, make sure its name or attributes match.
-        # Don't bother with Tags if we're searching for text.
-        elif isinstance(markup, Tag):
-            if not self.text:
-                found = self.searchTag(markup)
-        # If it's text, make sure the text matches.
-        elif isinstance(markup, NavigableString) or \
-                 isString(markup):
-            if self._matches(markup, self.text):
-                found = markup
-        else:
-            raise Exception, "I don't know how to match against a %s" \
-                  % markup.__class__
-        return found
-
-    def _matches(self, markup, matchAgainst):
-        #print "Matching %s against %s" % (markup, matchAgainst)
-        result = False
-        if matchAgainst == True and type(matchAgainst) == types.BooleanType:
-            result = markup != None
-        elif callable(matchAgainst):
-            result = matchAgainst(markup)
-        else:
-            #Custom match methods take the tag as an argument, but all
-            #other ways of matching match the tag name as a string.
-            if isinstance(markup, Tag):
-                markup = markup.name
-            if markup is not None and not isString(markup):
-                markup = unicode(markup)
-            #Now we know that chunk is either a string, or None.
-            if hasattr(matchAgainst, 'match'):
-                # It's a regexp object.
-                result = markup and matchAgainst.search(markup)
-            elif (isList(matchAgainst)
-                  and (markup is not None or not isString(matchAgainst))):
-                result = markup in matchAgainst
-            elif hasattr(matchAgainst, 'items'):
-                result = markup.has_key(matchAgainst)
-            elif matchAgainst and isString(markup):
-                if isinstance(markup, unicode):
-                    matchAgainst = unicode(matchAgainst)
-                else:
-                    matchAgainst = str(matchAgainst)
-
-            if not result:
-                result = matchAgainst == markup
-        return result
-
-class ResultSet(list):
-    """A ResultSet is just a list that keeps track of the SoupStrainer
-    that created it."""
-    def __init__(self, source):
-        list.__init__([])
-        self.source = source
-
-# Now, some helper functions.
-
-def isList(l):
-    """Convenience method that works with all 2.x versions of Python
-    to determine whether or not something is listlike."""
-    return ((hasattr(l, '__iter__') and not isString(l))
-            or (type(l) in (types.ListType, types.TupleType)))
-
-def isString(s):
-    """Convenience method that works with all 2.x versions of Python
-    to determine whether or not something is stringlike."""
-    try:
-        return isinstance(s, unicode) or isinstance(s, basestring)
-    except NameError:
-        return isinstance(s, str)
-
-def buildSet(args):
-    """Turns a list or a string into a set."""
-    if isinstance(args, str):
-        return set([args])
-    return set(args)
-
-def buildTagMap(default, *args):
-    """Turns a list of maps, lists, or scalars into a single map.
-    Used to build the nestable_tags and reset_nesting_tags maps out of
-    lists and partial maps."""
-    built = {}
-    for portion in args:
-        if hasattr(portion, 'items'):
-            #It's a map. Merge it.
-            for k,v in portion.items():
-                built[k] = v
-        elif isList(portion) and not isString(portion):
-            #It's a list. Map each item to the default.
-            for k in portion:
-                built[k] = default
-        else:
-            #It's a scalar. Map it to the default.
-            built[portion] = default
-    return built
-
-# Now, the parser classes.
-
-class TreeBuilder(Entities):
-
-    smartQuotesTo = Entities.XML_ENTITIES
-    preserve_whitespace_tags = set()
-    quote_tags = set()
-    self_closing_tags = set()
-    assume_html = False
-
-    def __init__(self):
-        self.soup = None
-
-    def isSelfClosingTag(self, name):
-        return name in self.self_closing_tags
+__all__ = ['BeautifulSoup',
 
-    def reset(self):
-        pass
-
-    def feed(self):
-        pass
-
-
-class HTMLParserXMLTreeBuilder(HTMLParser, TreeBuilder):
-
-    """
-    This class defines a basic tree builder based on Python's built-in
-    HTMLParser. The tree builder knows nothing about tag
-    behavior except for the following:
-
-      You can't close a tag without closing all the tags it encloses.
-      That is, "<foo><bar></foo>" actually means
-      "<foo><bar></bar></foo>".
-
-    [Another possible explanation is "<foo><bar /></foo>", but unless
-    you specify 'bar' in self_closing_tags, this class will never use
-    that explanation.]
-
-    This class is useful for parsing XML or made-up markup languages,
-    or when BeautifulSoup makes an assumption counter to what you were
-    expecting.
+           # Stuff imported from other packages
+           'Entities',
+           'HTMLParserXMLTreeBuilder',
+           'HTMLParserTreeBuilder',
+           'HTMLTreeBuilder',
+           'XMLTreeBuilder',
 
+           'BeautifulStoneSoup',
+           'ICantBelieveItsBeautifulSoup']
 
-    HTMLParser will process most bad HTML, and the BeautifulSoup class
-    has some tricks for dealing with some HTML that kills HTMLParser,
-    but Beautiful Soup can nonetheless choke or lose data if your data
-    uses self-closing tags or declarations incorrectly.
-
-    This class uses regexes to sanitize input, avoiding the vast
-    majority of these problems. If the problems don't apply to you,
-    pass in False for markupMassage, and you'll get better
-    performance.
-
-    The default parser massage techniques fix the two most common
-    instances of invalid HTML that choke HTMLParser:
-
-        <br/> (No space between name of closing tag and tag close)
-        <! --Comment--> (Extraneous whitespace in declaration)
-
-    You can pass in a custom list of (RE object, replace method)
-    tuples to get HTMLParserXMLTreeBuilder to scrub your input the way you
-    want.
-    """
-    reset_nesting_tags = {}
-    nestable_tags = {}
-
-    MARKUP_MASSAGE = [(re.compile('(<[^<>]*)/>'),
-                       lambda x: x.group(1) + ' />'),
-                      (re.compile('<!\s+([^<>]*)>'),
-                       lambda x: '<!' + x.group(1) + '>')
-                      ]
-
-    def __init__(self, convertEntities=None, markupMassage=True,
-                 selfClosingTags=None,
-                 smartQuotesTo=Entities.XML_ENTITIES):
-        HTMLParser.__init__(self)
-        self.soup = None
-        self.convertEntities = convertEntities
-        self.instanceSelfClosingTags = buildSet(selfClosingTags or [])
-        self.markupMassage = markupMassage
-        self.smartQuotesTo = smartQuotesTo
-        self.quoteStack = []
-
-        # Set the rules for how we'll deal with the entities we
-        # encounter
-        if self.convertEntities:
-            # It doesn't make sense to convert encoded characters to
-            # entities even while you're converting entities to Unicode.
-            # Just convert it all to Unicode.
-            self.smartQuotesTo = None
-            if convertEntities == self.HTML_ENTITIES:
-                self.convertXMLEntities = False
-                self.convertHTMLEntities = True
-                self.escapeUnrecognizedEntities = True
-            elif convertEntities == self.XHTML_ENTITIES:
-                self.convertXMLEntities = True
-                self.convertHTMLEntities = True
-                self.escapeUnrecognizedEntities = False
-            elif convertEntities == self.XML_ENTITIES:
-                self.convertXMLEntities = True
-                self.convertHTMLEntities = False
-                self.escapeUnrecognizedEntities = False
-        else:
-            self.convertXMLEntities = False
-            self.convertHTMLEntities = False
-            self.escapeUnrecognizedEntities = False
-
-    def feed(self, markup):
-        if markup is not None:
-            if self.markupMassage:
-                if not isList(self.markupMassage):
-                    self.markupMassage = self.MARKUP_MASSAGE
-                for fix, m in self.markupMassage:
-                    markup = fix.sub(m, markup)
-                # TODO: We get rid of markupMassage so that the
-                # soup object can be deepcopied later on. Some
-                # Python installations can't copy regexes. If anyone
-                # was relying on the existence of markupMassage, this
-                # might cause problems.
-                # XXX: This might not be necessary now that we've moved
-                # the massage code into the builder.
-                #del(self.markupMassage)
-        HTMLParser.feed(self, markup)
-
-    def isSelfClosingTag(self, name):
-        """Returns true iff the given string is the name of a
-        self-closing tag according to this parser."""
-        return (name in self.self_closing_tags
-                or name in self.instanceSelfClosingTags)
-
-    def handle_starttag(self, name, attrs):
-        if len(self.quoteStack) > 0:
-            #This is not a real tag.
-            #print "<%s> is not real!" % name
-            attrs = ''.join(map(lambda(x, y): ' %s="%s"' % (x, y), attrs))
-            self.handle_data('<%s%s>' % (name, attrs))
-            return
-        if not self.isSelfClosingTag(name):
-            self.soup.endData()
-            self._smartPop(name)
-        tag = self.soup.handle_starttag(name, attrs)
-        if tag is None:
-            # The tag was filtered out by the SoupStrainer
-            return
-        if name in self.quote_tags:
-            #print "Beginning quote (%s)" % name
-            self.quoteStack.append(name)
-            self.literal = 1
-        if self.isSelfClosingTag(name):
-            self.soup.handle_endtag(name)
-
-    def handle_endtag(self, name):
-        if self.quoteStack and self.quoteStack[-1] != name:
-            #This is not a real end tag.
-            #print "</%s> is not real!" % name
-            self.handle_data('</%s>' % name)
-            return
-        self.soup.handle_endtag(name)
-        if self.quoteStack and self.quoteStack[-1] == name:
-            self.quoteStack.pop()
-            self.literal = (len(self.quoteStack) > 0)
-
-    def handle_data(self, content):
-        #print "Handling data " + content
-        self.soup.handle_data(content)
-
-    def handle_pi(self, text):
-        """Handle a processing instruction as a ProcessingInstruction
-        object, possibly one with a %SOUP-ENCODING% slot into which an
-        encoding will be plugged later."""
-        if text[:3] == "xml":
-            text = u"xml version='1.0' encoding='%SOUP-ENCODING%'"
-        self._toStringSubclass(text, ProcessingInstruction)
-
-    def handle_comment(self, text):
-        "Handle comments as Comment objects."
-        self._toStringSubclass(text, Comment)
-
-    def handle_charref(self, ref):
-        "Handle character references as data."
-        if self.convertEntities:
-            data = unichr(int(ref))
-        else:
-            data = '&#%s;' % ref
-        self.handle_data(data)
-
-    def handle_entityref(self, ref):
-        """Handle entity references as data, possibly converting known
-        HTML and/or XML entity references to the corresponding Unicode
-        characters."""
-        data = None
-        if self.convertHTMLEntities:
-            try:
-                data = unichr(name2codepoint[ref])
-            except KeyError:
-                pass
-
-        if not data and self.convertXMLEntities:
-                data = self.XML_ENTITIES_TO_SPECIAL_CHARS.get(ref)
-
-        if not data and self.convertHTMLEntities and \
-            not self.XML_ENTITIES_TO_SPECIAL_CHARS.get(ref):
-                # TODO: We've got a problem here. We're told this is
-                # an entity reference, but it's not an XML entity
-                # reference or an HTML entity reference. Nonetheless,
-                # the logical thing to do is to pass it through as an
-                # unrecognized entity reference.
-                #
-                # Except: when the input is "&carol;" this function
-                # will be called with input "carol". When the input is
-                # "AT&T", this function will be called with input
-                # "T". We have no way of knowing whether a semicolon
-                # was present originally, so we don't know whether
-                # this is an unknown entity or just a misplaced
-                # ampersand.
-                #
-                # The more common case is a misplaced ampersand, so I
-                # escape the ampersand and omit the trailing semicolon.
-                data = "&amp;%s" % ref
-        if not data:
-            # This case is different from the one above, because we
-            # haven't already gone through a supposedly comprehensive
-            # mapping of entities to Unicode characters. We might not
-            # have gone through any mapping at all. So the chances are
-            # very high that this is a real entity, and not a
-            # misplaced ampersand.
-            data = "&%s;" % ref
-        self.handle_data(data)
-
-    def handle_decl(self, data):
-        "Handle DOCTYPEs and the like as Declaration objects."
-        self._toStringSubclass(data, Declaration)
-
-    def _toStringSubclass(self, text, subclass):
-        """Adds a certain piece of text to the tree as a NavigableString
-        subclass."""
-        self.soup.endData()
-        self.handle_data(text)
-        self.soup.endData(subclass)
-
-    def _smartPop(self, name):
-
-        """We need to pop up to the previous tag of this type, unless
-        one of this tag's nesting reset triggers comes between this
-        tag and the previous tag of this type, OR unless this tag is a
-        generic nesting trigger and another generic nesting trigger
-        comes between this tag and the previous tag of this type.
-
-        Examples:
-         <p>Foo<b>Bar *<p>* should pop to 'p', not 'b'.
-         <p>Foo<table>Bar *<p>* should pop to 'table', not 'p'.
-         <p>Foo<table><tr>Bar *<p>* should pop to 'tr', not 'p'.
-
-         <li><ul><li> *<li>* should pop to 'ul', not the first 'li'.
-         <tr><table><tr> *<tr>* should pop to 'table', not the first 'tr'
-         <td><tr><td> *<td>* should pop to 'tr', not the first 'td'
-        """
+import re
 
-        nestingResetTriggers = self.nestable_tags.get(name)
-        isNestable = nestingResetTriggers != None
-        isResetNesting = self.reset_nesting_tags.has_key(name)
-        popTo = None
-        inclusive = True
-        for i in range(len(self.soup.tagStack)-1, 0, -1):
-            p = self.soup.tagStack[i]
-            if (not p or p.name == name) and not isNestable:
-                #Non-nestable tags get popped to the top or to their
-                #last occurance.
-                popTo = name
-                break
-            if (nestingResetTriggers != None
-                and p.name in nestingResetTriggers) \
-                or (nestingResetTriggers == None and isResetNesting
-                    and self.reset_nesting_tags.has_key(p.name)):
-
-                #If we encounter one of the nesting reset triggers
-                #peculiar to this tag, or we encounter another tag
-                #that causes nesting to reset, pop up to but not
-                #including that tag.
-                popTo = p.name
-                inclusive = False
-                break
-            p = p.parent
-        if popTo:
-            self.soup._popToTag(popTo, inclusive)
-
-    def parse_declaration(self, i):
-        """Treat a bogus SGML declaration as raw data. Treat a CDATA
-        declaration as a CData object."""
-        j = None
-        if self.rawdata[i:i+9] == '<![CDATA[':
-             k = self.rawdata.find(']]>', i)
-             if k == -1:
-                 k = len(self.rawdata)
-             data = self.rawdata[i+9:k]
-             j = k+3
-             self._toStringSubclass(data, CData)
-        else:
-            try:
-                j = HTMLParser.parse_declaration(self, i)
-            except HTMLParseError:
-                toHandle = self.rawdata[i:]
-                self.handle_data(toHandle)
-                j = i + len(toHandle)
-        return j
-
-
-class HTMLParserTreeBuilder(HTMLParserXMLTreeBuilder):
-    """This builder knows the following facts about HTML:
-
-    * Some tags have no closing tag and should be interpreted as being
-      closed as soon as they are encountered.
-
-    * The text inside some tags (ie. 'script') may contain tags which
-      are not really part of the document and which should be parsed
-      as text, not tags. If you want to parse the text as tags, you can
-      always fetch it and parse it explicitly.
-
-    * Tag nesting rules:
-
-      Most tags can't be nested at all. For instance, the occurance of
-      a <p> tag should implicitly close the previous <p> tag.
-
-       <p>Para1<p>Para2
-        should be transformed into:
-       <p>Para1</p><p>Para2
-
-      Some tags can be nested arbitrarily. For instance, the occurance
-      of a <blockquote> tag should _not_ implicitly close the previous
-      <blockquote> tag.
-
-       Alice said: <blockquote>Bob said: <blockquote>Blah
-        should NOT be transformed into:
-       Alice said: <blockquote>Bob said: </blockquote><blockquote>Blah
-
-      Some tags can be nested, but the nesting is reset by the
-      interposition of other tags. For instance, a <tr> tag should
-      implicitly close the previous <tr> tag within the same <table>,
-      but not close a <tr> tag in another table.
-
-       <table><tr>Blah<tr>Blah
-        should be transformed into:
-       <table><tr>Blah</tr><tr>Blah
-        but,
-       <tr>Blah<table><tr>Blah
-        should NOT be transformed into
-       <tr>Blah<table></tr><tr>Blah
-
-    Differing assumptions about tag nesting rules are a major source
-    of problems with the BeautifulSoup class. If BeautifulSoup is not
-    treating as nestable a tag your page author treats as nestable,
-    try subclassing this tree builder or using another parser's tree
-    builder."""
-
-    assume_html = True
-    preserve_whitespace_tags = set(['pre', 'textarea'])
-    quote_tags = set(['script', 'textarea'])
-    self_closing_tags = set(['br' , 'hr', 'input', 'img', 'meta',
-                             'spacer', 'link', 'frame', 'base'])
-
-    #According to the HTML standard, each of these inline tags can
-    #contain another tag of the same type. Furthermore, it's common
-    #to actually use these tags this way.
-    nestable_inline_tags = ['span', 'font', 'q', 'object', 'bdo', 'sub', 'sup',
-                            'center']
-
-    #According to the HTML standard, these block tags can contain
-    #another tag of the same type. Furthermore, it's common
-    #to actually use these tags this way.
-    nestable_block_tags = ['blockquote', 'div', 'fieldset', 'ins', 'del']
-
-    #Lists can contain other lists, but there are restrictions.
-    nestable_list_tags = { 'ol' : [],
-                           'ul' : [],
-                           'li' : ['ul', 'ol'],
-                           'dl' : [],
-                           'dd' : ['dl'],
-                           'dt' : ['dl'] }
-
-    #Tables can contain other tables, but there are restrictions.
-    nestable_table_tags = {'table' : [],
-                           'tr' : ['table', 'tbody', 'tfoot', 'thead'],
-                           'td' : ['tr'],
-                           'th' : ['tr'],
-                           'thead' : ['table'],
-                           'tbody' : ['table'],
-                           'tfoot' : ['table'],
-                           }
-
-    non_nestable_block_tags = ['address', 'form', 'p', 'pre']
-
-    #If one of these tags is encountered, all tags up to the next tag of
-    #this type are popped.
-    reset_nesting_tags = buildTagMap(None, nestable_block_tags, 'noscript',
-                                     non_nestable_block_tags,
-                                     nestable_list_tags,
-                                     nestable_table_tags)
-
-    nestable_tags = buildTagMap([], nestable_inline_tags, nestable_block_tags,
-                                nestable_list_tags, nestable_table_tags)
-
-
-    def __init__(self, *args, **kwargs):
-        if not kwargs.has_key('smartQuotesTo'):
-            kwargs['smartQuotesTo'] = self.HTML_ENTITIES
-        HTMLParserXMLTreeBuilder.__init__(self, *args, **kwargs)
+from util import isList, isString, buildSet
+from dammit import UnicodeDammit
+from builder import (
+    HTMLParserXMLTreeBuilder, HTMLParserTreeBuilder, HTMLTreeBuilder,
+    ICantBelieveItsValidHTMLTreeBuilder, XMLTreeBuilder)
+from element import Entities, NavigableString, Tag
 
 
 class BeautifulStoneSoup(Tag):
@@ -1530,7 +221,7 @@ class BeautifulStoneSoup(Tag):
         if self.currentData:
             currentData = u''.join(self.currentData)
             if (currentData.translate(self.STRIP_ASCII_SPACES) == '' and
-                not set([tag.name for tag in self.tagStack]).intersection(
+                not buildSet([tag.name for tag in self.tagStack]).intersection(
                     self.builder.preserve_whitespace_tags)):
                 if '\n' in currentData:
                     currentData = '\n'
@@ -1665,42 +356,7 @@ class StopParsing(Exception):
     pass
 
 
-class ICantBelieveItsValidHTMLBuilder(HTMLParserTreeBuilder):
-    i_cant_believe_theyre_nestable_inline_tags = \
-     ['em', 'big', 'i', 'small', 'tt', 'abbr', 'acronym', 'strong',
-      'cite', 'code', 'dfn', 'kbd', 'samp', 'strong', 'var', 'b',
-      'big']
-
-    i_cant_believe_theyre_nestable_block_tags = ['noscript']
-
-    nestable_tags = buildTagMap([], HTMLParserTreeBuilder.nestable_tags,
-                                i_cant_believe_theyre_nestable_block_tags,
-                                i_cant_believe_theyre_nestable_inline_tags)
-
-
 class ICantBelieveItsBeautifulSoup(BeautifulStoneSoup):
-    """The BeautifulSoup class is oriented towards skipping over
-    common HTML errors like unclosed tags. However, sometimes it makes
-    errors of its own. For instance, consider this fragment:
-
-     <b>Foo<b>Bar</b></b>
-
-    This is perfectly valid (if bizarre) HTML. However, the
-    BeautifulSoup class will implicitly close the first b tag when it
-    encounters the second 'b'. It will think the author wrote
-    "<b>Foo<b>Bar", and didn't close the first 'b' tag, because
-    there's no real-world reason to bold something that's already
-    bold. When it encounters '</b></b>' it will close two more 'b'
-    tags, for a grand total of three tags closed instead of two. This
-    can throw off the rest of your document structure. The same is
-    true of a number of other tags, listed below.
-
-    It's much more common for someone to forget to close a 'b' tag
-    than to actually use nested 'b' tags, and the BeautifulSoup class
-    handles the common case. This class handles the not-co-common
-    case: where you can't believe someone wrote what they did, but
-    it's valid HTML and BeautifulSoup screwed up by assuming it
-    wouldn't be."""
     def _defaultBuilder(self):
         return ICantBelieveItsValidHTMLBuilder()
 
diff --git a/BeautifulSoupTests.py b/BeautifulSoupTests.py
index 721020d..7b29c94 100644
--- a/BeautifulSoupTests.py
+++ b/BeautifulSoupTests.py
@@ -5,8 +5,12 @@ These tests make sure the Beautiful Soup works as it should. If you
 find a bug in Beautiful Soup, the best way to express it is as a test
 case like this that fails."""
 
+import re
 import unittest
 from BeautifulSoup import *
+from element import CData, Comment, Declaration, SoupStrainer, Tag
+from builder import ICantBelieveItsValidHTMLTreeBuilder
+from dammit import UnicodeDammit
 
 class SoupTest(unittest.TestCase):
 
@@ -836,7 +840,7 @@ class AlternateBuilders(SoupTest):
     """Test alternate builders."""
 
     def testICantBelieveItsValidHTML(self):
-        builder = ICantBelieveItsValidHTMLBuilder()
+        builder = ICantBelieveItsValidHTMLTreeBuilder()
         markup = "<b>Foo<b>Bar</b></b>"
 
         soup = BeautifulSoup(markup)
diff --git a/builder.py b/builder.py
new file mode 100644
index 0000000..aeae1e8
--- /dev/null
+++ b/builder.py
@@ -0,0 +1,484 @@
+import markupbase
+import re
+from util import buildSet, isList, isString
+# element has taken care of import weirdness, so import name2codepoint
+# from there to avoid duplicating the weirdness.
+from element import name2codepoint
+from element import (
+    CData, Comment, Declaration, Entities, ProcessingInstruction)
+from HTMLParser import HTMLParser, HTMLParseError
+
+__all__ = ['TreeBuilder',
+           'HTMLParserXMLTreeBuilder',
+           'HTMLParserTreeBuilder',
+           'XMLTreeBuilder',
+           'HTMLTreeBuilder',
+           'ICantBelieveItsValidHTMLTreeBuilder']
+
+#This hack makes the HTMLParser-based tree builders able to parse XML
+#with namespaces.
+markupbase._declname_match = re.compile(r'[a-zA-Z][-_.:a-zA-Z0-9]*\s*').match
+
+def buildTagMap(default, *args):
+    """Turns a list of maps, lists, or scalars into a single map.
+    Used to build the nestable_tags and reset_nesting_tags maps out of
+    lists and partial maps."""
+    built = {}
+    for portion in args:
+        if hasattr(portion, 'items'):
+            #It's a map. Merge it.
+            for k,v in portion.items():
+                built[k] = v
+        elif isList(portion) and not isString(portion):
+            #It's a list. Map each item to the default.
+            for k in portion:
+                built[k] = default
+        else:
+            #It's a scalar. Map it to the default.
+            built[portion] = default
+    return built
+
+
+class TreeBuilder(Entities):
+
+    smartQuotesTo = Entities.XML_ENTITIES
+    preserve_whitespace_tags = buildSet()
+    quote_tags = buildSet()
+    self_closing_tags = buildSet()
+    assume_html = False
+
+    def __init__(self):
+        self.soup = None
+
+    def isSelfClosingTag(self, name):
+        return name in self.self_closing_tags
+
+    def reset(self):
+        pass
+
+    def feed(self):
+        pass
+
+
+class HTMLParserXMLTreeBuilder(HTMLParser, TreeBuilder):
+
+    """
+    This class defines a basic tree builder based on Python's built-in
+    HTMLParser. The tree builder knows nothing about tag
+    behavior except for the following:
+
+      You can't close a tag without closing all the tags it encloses.
+      That is, "<foo><bar></foo>" actually means
+      "<foo><bar></bar></foo>".
+
+    [Another possible explanation is "<foo><bar /></foo>", but unless
+    you specify 'bar' in self_closing_tags, this class will never use
+    that explanation.]
+
+    This class is useful for parsing XML or made-up markup languages,
+    or when BeautifulSoup makes an assumption counter to what you were
+    expecting.
+
+
+    HTMLParser will process most bad HTML, and the BeautifulSoup class
+    has some tricks for dealing with some HTML that kills HTMLParser,
+    but Beautiful Soup can nonetheless choke or lose data if your data
+    uses self-closing tags or declarations incorrectly.
+
+    This class uses regexes to sanitize input, avoiding the vast
+    majority of these problems. If the problems don't apply to you,
+    pass in False for markupMassage, and you'll get better
+    performance.
+
+    The default parser massage techniques fix the two most common
+    instances of invalid HTML that choke HTMLParser:
+
+        <br/> (No space between name of closing tag and tag close)
+        <! --Comment--> (Extraneous whitespace in declaration)
+
+    You can pass in a custom list of (RE object, replace method)
+    tuples to get HTMLParserXMLTreeBuilder to scrub your input the way you
+    want.
+    """
+    reset_nesting_tags = {}
+    nestable_tags = {}
+
+    MARKUP_MASSAGE = [(re.compile('(<[^<>]*)/>'),
+                       lambda x: x.group(1) + ' />'),
+                      (re.compile('<!\s+([^<>]*)>'),
+                       lambda x: '<!' + x.group(1) + '>')
+                      ]
+
+    def __init__(self, convertEntities=None, markupMassage=True,
+                 selfClosingTags=None,
+                 smartQuotesTo=Entities.XML_ENTITIES):
+        HTMLParser.__init__(self)
+        self.soup = None
+        self.convertEntities = convertEntities
+        self.instanceSelfClosingTags = buildSet(selfClosingTags or [])
+        self.markupMassage = markupMassage
+        self.smartQuotesTo = smartQuotesTo
+        self.quoteStack = []
+
+        # Set the rules for how we'll deal with the entities we
+        # encounter
+        if self.convertEntities:
+            # It doesn't make sense to convert encoded characters to
+            # entities even while you're converting entities to Unicode.
+            # Just convert it all to Unicode.
+            self.smartQuotesTo = None
+            if convertEntities == self.HTML_ENTITIES:
+                self.convertXMLEntities = False
+                self.convertHTMLEntities = True
+                self.escapeUnrecognizedEntities = True
+            elif convertEntities == self.XHTML_ENTITIES:
+                self.convertXMLEntities = True
+                self.convertHTMLEntities = True
+                self.escapeUnrecognizedEntities = False
+            elif convertEntities == self.XML_ENTITIES:
+                self.convertXMLEntities = True
+                self.convertHTMLEntities = False
+                self.escapeUnrecognizedEntities = False
+        else:
+            self.convertXMLEntities = False
+            self.convertHTMLEntities = False
+            self.escapeUnrecognizedEntities = False
+
+    def feed(self, markup):
+        if markup is not None:
+            if self.markupMassage:
+                if not isList(self.markupMassage):
+                    self.markupMassage = self.MARKUP_MASSAGE
+                for fix, m in self.markupMassage:
+                    markup = fix.sub(m, markup)
+                # TODO: We get rid of markupMassage so that the
+                # soup object can be deepcopied later on. Some
+                # Python installations can't copy regexes. If anyone
+                # was relying on the existence of markupMassage, this
+                # might cause problems.
+                # XXX: This might not be necessary now that we've moved
+                # the massage code into the builder.
+                #del(self.markupMassage)
+        HTMLParser.feed(self, markup)
+
+    def isSelfClosingTag(self, name):
+        """Returns true iff the given string is the name of a
+        self-closing tag according to this parser."""
+        return (name in self.self_closing_tags
+                or name in self.instanceSelfClosingTags)
+
+    def handle_starttag(self, name, attrs):
+        if len(self.quoteStack) > 0:
+            #This is not a real tag.
+            #print "<%s> is not real!" % name
+            attrs = ''.join(map(lambda(x, y): ' %s="%s"' % (x, y), attrs))
+            self.handle_data('<%s%s>' % (name, attrs))
+            return
+        if not self.isSelfClosingTag(name):
+            self.soup.endData()
+            self._smartPop(name)
+        tag = self.soup.handle_starttag(name, attrs)
+        if tag is None:
+            # The tag was filtered out by the SoupStrainer
+            return
+        if name in self.quote_tags:
+            #print "Beginning quote (%s)" % name
+            self.quoteStack.append(name)
+            self.literal = 1
+        if self.isSelfClosingTag(name):
+            self.soup.handle_endtag(name)
+
+    def handle_endtag(self, name):
+        if self.quoteStack and self.quoteStack[-1] != name:
+            #This is not a real end tag.
+            #print "</%s> is not real!" % name
+            self.handle_data('</%s>' % name)
+            return
+        self.soup.handle_endtag(name)
+        if self.quoteStack and self.quoteStack[-1] == name:
+            self.quoteStack.pop()
+            self.literal = (len(self.quoteStack) > 0)
+
+    def handle_data(self, content):
+        #print "Handling data " + content
+        self.soup.handle_data(content)
+
+    def handle_pi(self, text):
+        """Handle a processing instruction as a ProcessingInstruction
+        object, possibly one with a %SOUP-ENCODING% slot into which an
+        encoding will be plugged later."""
+        if text[:3] == "xml":
+            text = u"xml version='1.0' encoding='%SOUP-ENCODING%'"
+        self._toStringSubclass(text, ProcessingInstruction)
+
+    def handle_comment(self, text):
+        "Handle comments as Comment objects."
+        self._toStringSubclass(text, Comment)
+
+    def handle_charref(self, ref):
+        "Handle character references as data."
+        if self.convertEntities:
+            data = unichr(int(ref))
+        else:
+            data = '&#%s;' % ref
+        self.handle_data(data)
+
+    def handle_entityref(self, ref):
+        """Handle entity references as data, possibly converting known
+        HTML and/or XML entity references to the corresponding Unicode
+        characters."""
+        data = None
+        if self.convertHTMLEntities:
+            try:
+                data = unichr(name2codepoint[ref])
+            except KeyError:
+                pass
+
+        if not data and self.convertXMLEntities:
+                data = self.XML_ENTITIES_TO_SPECIAL_CHARS.get(ref)
+
+        if not data and self.convertHTMLEntities and \
+            not self.XML_ENTITIES_TO_SPECIAL_CHARS.get(ref):
+                # TODO: We've got a problem here. We're told this is
+                # an entity reference, but it's not an XML entity
+                # reference or an HTML entity reference. Nonetheless,
+                # the logical thing to do is to pass it through as an
+                # unrecognized entity reference.
+                #
+                # Except: when the input is "&carol;" this function
+                # will be called with input "carol". When the input is
+                # "AT&T", this function will be called with input
+                # "T". We have no way of knowing whether a semicolon
+                # was present originally, so we don't know whether
+                # this is an unknown entity or just a misplaced
+                # ampersand.
+                #
+                # The more common case is a misplaced ampersand, so I
+                # escape the ampersand and omit the trailing semicolon.
+                data = "&amp;%s" % ref
+        if not data:
+            # This case is different from the one above, because we
+            # haven't already gone through a supposedly comprehensive
+            # mapping of entities to Unicode characters. We might not
+            # have gone through any mapping at all. So the chances are
+            # very high that this is a real entity, and not a
+            # misplaced ampersand.
+            data = "&%s;" % ref
+        self.handle_data(data)
+
+    def handle_decl(self, data):
+        "Handle DOCTYPEs and the like as Declaration objects."
+        self._toStringSubclass(data, Declaration)
+
+    def _toStringSubclass(self, text, subclass):
+        """Adds a certain piece of text to the tree as a NavigableString
+        subclass."""
+        self.soup.endData()
+        self.handle_data(text)
+        self.soup.endData(subclass)
+
+    def _smartPop(self, name):
+
+        """We need to pop up to the previous tag of this type, unless
+        one of this tag's nesting reset triggers comes between this
+        tag and the previous tag of this type, OR unless this tag is a
+        generic nesting trigger and another generic nesting trigger
+        comes between this tag and the previous tag of this type.
+
+        Examples:
+         <p>Foo<b>Bar *<p>* should pop to 'p', not 'b'.
+         <p>Foo<table>Bar *<p>* should pop to 'table', not 'p'.
+         <p>Foo<table><tr>Bar *<p>* should pop to 'tr', not 'p'.
+
+         <li><ul><li> *<li>* should pop to 'ul', not the first 'li'.
+         <tr><table><tr> *<tr>* should pop to 'table', not the first 'tr'
+         <td><tr><td> *<td>* should pop to 'tr', not the first 'td'
+        """
+
+        nestingResetTriggers = self.nestable_tags.get(name)
+        isNestable = nestingResetTriggers != None
+        isResetNesting = self.reset_nesting_tags.has_key(name)
+        popTo = None
+        inclusive = True
+        for i in range(len(self.soup.tagStack)-1, 0, -1):
+            p = self.soup.tagStack[i]
+            if (not p or p.name == name) and not isNestable:
+                #Non-nestable tags get popped to the top or to their
+                #last occurance.
+                popTo = name
+                break
+            if (nestingResetTriggers != None
+                and p.name in nestingResetTriggers) \
+                or (nestingResetTriggers == None and isResetNesting
+                    and self.reset_nesting_tags.has_key(p.name)):
+
+                #If we encounter one of the nesting reset triggers
+                #peculiar to this tag, or we encounter another tag
+                #that causes nesting to reset, pop up to but not
+                #including that tag.
+                popTo = p.name
+                inclusive = False
+                break
+            p = p.parent
+        if popTo:
+            self.soup._popToTag(popTo, inclusive)
+
+    def parse_declaration(self, i):
+        """Treat a bogus SGML declaration as raw data. Treat a CDATA
+        declaration as a CData object."""
+        j = None
+        if self.rawdata[i:i+9] == '<![CDATA[':
+             k = self.rawdata.find(']]>', i)
+             if k == -1:
+                 k = len(self.rawdata)
+             data = self.rawdata[i+9:k]
+             j = k+3
+             self._toStringSubclass(data, CData)
+        else:
+            try:
+                j = HTMLParser.parse_declaration(self, i)
+            except HTMLParseError:
+                toHandle = self.rawdata[i:]
+                self.handle_data(toHandle)
+                j = i + len(toHandle)
+        return j
+
+
+class HTMLParserTreeBuilder(HTMLParserXMLTreeBuilder):
+    """This builder knows the following facts about HTML:
+
+    * Some tags have no closing tag and should be interpreted as being
+      closed as soon as they are encountered.
+
+    * The text inside some tags (ie. 'script') may contain tags which
+      are not really part of the document and which should be parsed
+      as text, not tags. If you want to parse the text as tags, you can
+      always fetch it and parse it explicitly.
+
+    * Tag nesting rules:
+
+      Most tags can't be nested at all. For instance, the occurance of
+      a <p> tag should implicitly close the previous <p> tag.
+
+       <p>Para1<p>Para2
+        should be transformed into:
+       <p>Para1</p><p>Para2
+
+      Some tags can be nested arbitrarily. For instance, the occurance
+      of a <blockquote> tag should _not_ implicitly close the previous
+      <blockquote> tag.
+
+       Alice said: <blockquote>Bob said: <blockquote>Blah
+        should NOT be transformed into:
+       Alice said: <blockquote>Bob said: </blockquote><blockquote>Blah
+
+      Some tags can be nested, but the nesting is reset by the
+      interposition of other tags. For instance, a <tr> tag should
+      implicitly close the previous <tr> tag within the same <table>,
+      but not close a <tr> tag in another table.
+
+       <table><tr>Blah<tr>Blah
+        should be transformed into:
+       <table><tr>Blah</tr><tr>Blah
+        but,
+       <tr>Blah<table><tr>Blah
+        should NOT be transformed into
+       <tr>Blah<table></tr><tr>Blah
+
+    Differing assumptions about tag nesting rules are a major source
+    of problems with the BeautifulSoup class. If BeautifulSoup is not
+    treating as nestable a tag your page author treats as nestable,
+    try subclassing this tree builder or using another parser's tree
+    builder."""
+
+    assume_html = True
+    preserve_whitespace_tags = buildSet(['pre', 'textarea'])
+    quote_tags = buildSet(['script', 'textarea'])
+    self_closing_tags = buildSet(['br' , 'hr', 'input', 'img', 'meta',
+                                  'spacer', 'link', 'frame', 'base'])
+
+    #According to the HTML standard, each of these inline tags can
+    #contain another tag of the same type. Furthermore, it's common
+    #to actually use these tags this way.
+    nestable_inline_tags = ['span', 'font', 'q', 'object', 'bdo', 'sub', 'sup',
+                            'center']
+
+    #According to the HTML standard, these block tags can contain
+    #another tag of the same type. Furthermore, it's common
+    #to actually use these tags this way.
+    nestable_block_tags = ['blockquote', 'div', 'fieldset', 'ins', 'del']
+
+    #Lists can contain other lists, but there are restrictions.
+    nestable_list_tags = { 'ol' : [],
+                           'ul' : [],
+                           'li' : ['ul', 'ol'],
+                           'dl' : [],
+                           'dd' : ['dl'],
+                           'dt' : ['dl'] }
+
+    #Tables can contain other tables, but there are restrictions.
+    nestable_table_tags = {'table' : [],
+                           'tr' : ['table', 'tbody', 'tfoot', 'thead'],
+                           'td' : ['tr'],
+                           'th' : ['tr'],
+                           'thead' : ['table'],
+                           'tbody' : ['table'],
+                           'tfoot' : ['table'],
+                           }
+
+    non_nestable_block_tags = ['address', 'form', 'p', 'pre']
+
+    #If one of these tags is encountered, all tags up to the next tag of
+    #this type are popped.
+    reset_nesting_tags = buildTagMap(None, nestable_block_tags, 'noscript',
+                                     non_nestable_block_tags,
+                                     nestable_list_tags,
+                                     nestable_table_tags)
+
+    nestable_tags = buildTagMap([], nestable_inline_tags, nestable_block_tags,
+                                nestable_list_tags, nestable_table_tags)
+
+
+    def __init__(self, *args, **kwargs):
+        if not kwargs.has_key('smartQuotesTo'):
+            kwargs['smartQuotesTo'] = self.HTML_ENTITIES
+        HTMLParserXMLTreeBuilder.__init__(self, *args, **kwargs)
+
+# Some aliases to use if you don't care about the underlying
+# implementation.
+XMLTreeBuilder = HTMLParserXMLTreeBuilder
+HTMLTreeBuilder = HTMLParserTreeBuilder
+
+class ICantBelieveItsValidHTMLTreeBuilder(HTMLParserTreeBuilder):
+    """The  is oriented towards skipping over
+    common HTML errors like unclosed tags. However, sometimes it makes
+    errors of its own. For instance, consider this fragment:
+
+     <b>Foo<b>Bar</b></b>
+
+    This is perfectly valid (if bizarre) HTML. However, the
+    BeautifulSoup class will implicitly close the first b tag when it
+    encounters the second 'b'. It will think the author wrote
+    "<b>Foo<b>Bar", and didn't close the first 'b' tag, because
+    there's no real-world reason to bold something that's already
+    bold. When it encounters '</b></b>' it will close two more 'b'
+    tags, for a grand total of three tags closed instead of two. This
+    can throw off the rest of your document structure. The same is
+    true of a number of other tags, listed below.
+
+    It's much more common for someone to forget to close a 'b' tag
+    than to actually use nested 'b' tags, and the BeautifulSoup class
+    handles the common case. This class handles the not-co-common
+    case: where you can't believe someone wrote what they did, but
+    it's valid HTML and BeautifulSoup screwed up by assuming it
+    wouldn't be."""
+    i_cant_believe_theyre_nestable_inline_tags = \
+     ['em', 'big', 'i', 'small', 'tt', 'abbr', 'acronym', 'strong',
+      'cite', 'code', 'dfn', 'kbd', 'samp', 'strong', 'var', 'b',
+      'big']
+
+    i_cant_believe_theyre_nestable_block_tags = ['noscript']
+
+    nestable_tags = buildTagMap([], HTMLParserTreeBuilder.nestable_tags,
+                                i_cant_believe_theyre_nestable_block_tags,
+                                i_cant_believe_theyre_nestable_inline_tags)
diff --git a/builder.py.3.diff b/builder.py.3.diff
new file mode 100644
index 0000000..91b510d
--- /dev/null
+++ b/builder.py.3.diff
@@ -0,0 +1,4 @@
+90c90
+< from HTMLParser import HTMLParser, HTMLParseError
+---
+> from html.parser import HTMLParser, HTMLParseError
diff --git a/element.py b/element.py
new file mode 100644
index 0000000..3bdb01f
--- /dev/null
+++ b/element.py
@@ -0,0 +1,870 @@
+import re
+import types
+try:
+    from htmlentitydefs import name2codepoint
+except ImportError:
+    name2codepoint = {}
+
+from util import isString, isList
+
+DEFAULT_OUTPUT_ENCODING = "utf-8"
+
+class Entities:
+    """A mixin class that knows about XML entities."""
+
+    HTML_ENTITIES = "html"
+    XML_ENTITIES = "xml"
+    XHTML_ENTITIES = "xhtml"
+
+    def _invert(h):
+        "Cheap function to invert a hash."
+        i = {}
+        for k,v in h.items():
+            i[v] = k
+        return i
+
+    XML_ENTITIES_TO_SPECIAL_CHARS = { "apos" : "'",
+                                      "quot" : '"',
+                                      "amp" : "&",
+                                      "lt" : "<",
+                                      "gt" : ">" }
+
+    XML_SPECIAL_CHARS_TO_ENTITIES = _invert(XML_ENTITIES_TO_SPECIAL_CHARS)
+
+class PageElement:
+    """Contains the navigational information for some part of the page
+    (either a tag or a piece of text)"""
+
+    def setup(self, parent=None, previous=None):
+        """Sets up the initial relations between this element and
+        other elements."""
+        self.parent = parent
+        self.previous = previous
+        self.next = None
+        self.previousSibling = None
+        self.nextSibling = None
+        if self.parent and self.parent.contents:
+            self.previousSibling = self.parent.contents[-1]
+            self.previousSibling.nextSibling = self
+
+    def replaceWith(self, replaceWith):
+        oldParent = self.parent
+        myIndex = self.parent.contents.index(self)
+        if hasattr(replaceWith, 'parent') and replaceWith.parent == self.parent:
+            # We're replacing this element with one of its siblings.
+            index = self.parent.contents.index(replaceWith)
+            if index and index < myIndex:
+                # Furthermore, it comes before this element. That
+                # means that when we extract it, the index of this
+                # element will change.
+                myIndex = myIndex - 1
+        self.extract()
+        oldParent.insert(myIndex, replaceWith)
+
+    def extract(self):
+        """Destructively rips this element out of the tree."""
+        if self.parent:
+            try:
+                self.parent.contents.remove(self)
+            except ValueError:
+                pass
+
+        #Find the two elements that would be next to each other if
+        #this element (and any children) hadn't been parsed. Connect
+        #the two.
+        lastChild = self._lastRecursiveChild()
+        nextElement = lastChild.next
+
+        if self.previous:
+            self.previous.next = nextElement
+        if nextElement:
+            nextElement.previous = self.previous
+        self.previous = None
+        lastChild.next = None
+
+        self.parent = None
+        if self.previousSibling:
+            self.previousSibling.nextSibling = self.nextSibling
+        if self.nextSibling:
+            self.nextSibling.previousSibling = self.previousSibling
+        self.previousSibling = self.nextSibling = None
+        return self
+
+    def _lastRecursiveChild(self):
+        "Finds the last element beneath this object to be parsed."
+        lastChild = self
+        while hasattr(lastChild, 'contents') and lastChild.contents:
+            lastChild = lastChild.contents[-1]
+        return lastChild
+
+    def insert(self, position, newChild):
+        if (isinstance(newChild, basestring)
+            or isinstance(newChild, unicode)) \
+            and not isinstance(newChild, NavigableString):
+            newChild = NavigableString(newChild)
+
+        position =  min(position, len(self.contents))
+        if hasattr(newChild, 'parent') and newChild.parent != None:
+            # We're 'inserting' an element that's already one
+            # of this object's children.
+            if newChild.parent == self:
+                index = self.find(newChild)
+                if index and index < position:
+                    # Furthermore we're moving it further down the
+                    # list of this object's children. That means that
+                    # when we extract this element, our target index
+                    # will jump down one.
+                    position = position - 1
+            newChild.extract()
+
+        newChild.parent = self
+        previousChild = None
+        if position == 0:
+            newChild.previousSibling = None
+            newChild.previous = self
+        else:
+            previousChild = self.contents[position-1]
+            newChild.previousSibling = previousChild
+            newChild.previousSibling.nextSibling = newChild
+            newChild.previous = previousChild._lastRecursiveChild()
+        if newChild.previous:
+            newChild.previous.next = newChild
+
+        newChildsLastElement = newChild._lastRecursiveChild()
+
+        if position >= len(self.contents):
+            newChild.nextSibling = None
+
+            parent = self
+            parentsNextSibling = None
+            while not parentsNextSibling:
+                parentsNextSibling = parent.nextSibling
+                parent = parent.parent
+                if not parent: # This is the last element in the document.
+                    break
+            if parentsNextSibling:
+                newChildsLastElement.next = parentsNextSibling
+            else:
+                newChildsLastElement.next = None
+        else:
+            nextChild = self.contents[position]
+            newChild.nextSibling = nextChild
+            if newChild.nextSibling:
+                newChild.nextSibling.previousSibling = newChild
+            newChildsLastElement.next = nextChild
+
+        if newChildsLastElement.next:
+            newChildsLastElement.next.previous = newChildsLastElement
+        self.contents.insert(position, newChild)
+
+    def append(self, tag):
+        """Appends the given tag to the contents of this tag."""
+        self.insert(len(self.contents), tag)
+
+    def findNext(self, name=None, attrs={}, text=None, **kwargs):
+        """Returns the first item that matches the given criteria and
+        appears after this Tag in the document."""
+        return self._findOne(self.findAllNext, name, attrs, text, **kwargs)
+
+    def findAllNext(self, name=None, attrs={}, text=None, limit=None,
+                    **kwargs):
+        """Returns all items that match the given criteria and appear
+        after this Tag in the document."""
+        return self._findAll(name, attrs, text, limit, self.nextGenerator,
+                             **kwargs)
+
+    def findNextSibling(self, name=None, attrs={}, text=None, **kwargs):
+        """Returns the closest sibling to this Tag that matches the
+        given criteria and appears after this Tag in the document."""
+        return self._findOne(self.findNextSiblings, name, attrs, text,
+                             **kwargs)
+
+    def findNextSiblings(self, name=None, attrs={}, text=None, limit=None,
+                         **kwargs):
+        """Returns the siblings of this Tag that match the given
+        criteria and appear after this Tag in the document."""
+        return self._findAll(name, attrs, text, limit,
+                             self.nextSiblingGenerator, **kwargs)
+    fetchNextSiblings = findNextSiblings # Compatibility with pre-3.x
+
+    def findPrevious(self, name=None, attrs={}, text=None, **kwargs):
+        """Returns the first item that matches the given criteria and
+        appears before this Tag in the document."""
+        return self._findOne(self.findAllPrevious, name, attrs, text, **kwargs)
+
+    def findAllPrevious(self, name=None, attrs={}, text=None, limit=None,
+                        **kwargs):
+        """Returns all items that match the given criteria and appear
+        before this Tag in the document."""
+        return self._findAll(name, attrs, text, limit, self.previousGenerator,
+                           **kwargs)
+    fetchPrevious = findAllPrevious # Compatibility with pre-3.x
+
+    def findPreviousSibling(self, name=None, attrs={}, text=None, **kwargs):
+        """Returns the closest sibling to this Tag that matches the
+        given criteria and appears before this Tag in the document."""
+        return self._findOne(self.findPreviousSiblings, name, attrs, text,
+                             **kwargs)
+
+    def findPreviousSiblings(self, name=None, attrs={}, text=None,
+                             limit=None, **kwargs):
+        """Returns the siblings of this Tag that match the given
+        criteria and appear before this Tag in the document."""
+        return self._findAll(name, attrs, text, limit,
+                             self.previousSiblingGenerator, **kwargs)
+    fetchPreviousSiblings = findPreviousSiblings # Compatibility with pre-3.x
+
+    def findParent(self, name=None, attrs={}, **kwargs):
+        """Returns the closest parent of this Tag that matches the given
+        criteria."""
+        # NOTE: We can't use _findOne because findParents takes a different
+        # set of arguments.
+        r = None
+        l = self.findParents(name, attrs, 1)
+        if l:
+            r = l[0]
+        return r
+
+    def findParents(self, name=None, attrs={}, limit=None, **kwargs):
+        """Returns the parents of this Tag that match the given
+        criteria."""
+
+        return self._findAll(name, attrs, None, limit, self.parentGenerator,
+                             **kwargs)
+    fetchParents = findParents # Compatibility with pre-3.x
+
+    #These methods do the real heavy lifting.
+
+    def _findOne(self, method, name, attrs, text, **kwargs):
+        r = None
+        l = method(name, attrs, text, 1, **kwargs)
+        if l:
+            r = l[0]
+        return r
+
+    def _findAll(self, name, attrs, text, limit, generator, **kwargs):
+        "Iterates over a generator looking for things that match."
+
+        if isinstance(name, SoupStrainer):
+            strainer = name
+        else:
+            # Build a SoupStrainer
+            strainer = SoupStrainer(name, attrs, text, **kwargs)
+        results = ResultSet(strainer)
+        g = generator()
+        while True:
+            try:
+                i = g.next()
+            except StopIteration:
+                break
+            if i:
+                found = strainer.search(i)
+                if found:
+                    results.append(found)
+                    if limit and len(results) >= limit:
+                        break
+        return results
+
+    #These Generators can be used to navigate starting from both
+    #NavigableStrings and Tags.
+    def nextGenerator(self):
+        i = self
+        while i:
+            i = i.next
+            yield i
+
+    def nextSiblingGenerator(self):
+        i = self
+        while i:
+            i = i.nextSibling
+            yield i
+
+    def previousGenerator(self):
+        i = self
+        while i:
+            i = i.previous
+            yield i
+
+    def previousSiblingGenerator(self):
+        i = self
+        while i:
+            i = i.previousSibling
+            yield i
+
+    def parentGenerator(self):
+        i = self
+        while i:
+            i = i.parent
+            yield i
+
+    # Utility methods
+    def substituteEncoding(self, str, encoding=None):
+        encoding = encoding or "utf-8"
+        return str.replace("%SOUP-ENCODING%", encoding)
+
+    def toEncoding(self, s, encoding=None):
+        """Encodes an object to a string in some encoding, or to Unicode.
+        ."""
+        if isinstance(s, unicode):
+            if encoding:
+                s = s.encode(encoding)
+        elif isinstance(s, str):
+            if encoding:
+                s = s.encode(encoding)
+            else:
+                s = unicode(s)
+        else:
+            if encoding:
+                s  = self.toEncoding(str(s), encoding)
+            else:
+                s = unicode(s)
+        return s
+
+class NavigableString(unicode, PageElement):
+
+    def __new__(cls, value):
+        """Create a new NavigableString.
+
+        When unpickling a NavigableString, this method is called with
+        the string in DEFAULT_OUTPUT_ENCODING. That encoding needs to be
+        passed in to the superclass's __new__ or the superclass won't know
+        how to handle non-ASCII characters.
+        """
+        if isinstance(value, unicode):
+            return unicode.__new__(cls, value)
+        return unicode.__new__(cls, value, DEFAULT_OUTPUT_ENCODING)
+
+    def __getnewargs__(self):
+        return (unicode(self),)
+
+    def __getattr__(self, attr):
+        """text.string gives you text. This is for backwards
+        compatibility for Navigable*String, but for CData* it lets you
+        get the string without the CData wrapper."""
+        if attr == 'string':
+            return self
+        else:
+            raise AttributeError, "'%s' object has no attribute '%s'" % (self.__class__.__name__, attr)
+
+    def encode(self, encoding=DEFAULT_OUTPUT_ENCODING):
+        return self.decode().encode(encoding)
+
+    def decodeGivenEventualEncoding(self, eventualEncoding):
+        return self
+
+class CData(NavigableString):
+
+    def decodeGivenEventualEncoding(self, eventualEncoding):
+        return u'<![CDATA[' + self + u']]>'
+
+class ProcessingInstruction(NavigableString):
+
+    def decodeGivenEventualEncoding(self, eventualEncoding):
+        output = self
+        if u'%SOUP-ENCODING%' in output:
+            output = self.substituteEncoding(output, eventualEncoding)
+        return u'<?' + output + u'?>'
+
+class Comment(NavigableString):
+    def decodeGivenEventualEncoding(self, eventualEncoding):
+        return u'<!--' + self + u'-->'
+
+class Declaration(NavigableString):
+    def decodeGivenEventualEncoding(self, eventualEncoding):
+        return u'<!' + self + u'>'
+
+class Tag(PageElement, Entities):
+
+    """Represents a found HTML tag with its attributes and contents."""
+
+    def _convertEntities(self, builder, match):
+        """Used in a call to re.sub to replace HTML, XML, and numeric
+        entities with the appropriate Unicode characters. If HTML
+        entities are being converted, any unrecognized entities are
+        escaped."""
+        x = match.group(1)
+        if builder.convertHTMLEntities and x in name2codepoint:
+            return unichr(name2codepoint[x])
+        elif x in self.XML_ENTITIES_TO_SPECIAL_CHARS:
+            if builder.convertXMLEntities:
+                return self.XML_ENTITIES_TO_SPECIAL_CHARS[x]
+            else:
+                return u'&%s;' % x
+        elif len(x) > 0 and x[0] == '#':
+            # Handle numeric entities
+            if len(x) > 1 and x[1] == 'x':
+                return unichr(int(x[2:], 16))
+            else:
+                return unichr(int(x[1:]))
+
+        elif self.escapeUnrecognizedEntities:
+            return u'&amp;%s;' % x
+        else:
+            return u'&%s;' % x
+
+    def __init__(self, parser, builder, name, attrs=None, parent=None,
+                 previous=None):
+        "Basic constructor."
+
+        # We don't actually store the parser object: that lets extracted
+        # chunks be garbage-collected
+        self.parserClass = parser.__class__
+        self.name = name
+        self.isSelfClosing = builder.isSelfClosingTag(name)
+        if attrs == None:
+            attrs = []
+        if isinstance(attrs, types.DictType):
+            self.attrMap = attrs
+        self.attrs = attrs
+        self.contents = []
+        self.setup(parent, previous)
+        self.hidden = False
+        self.containsSubstitutions = False
+        self.escapeUnrecognizedEntities = parser.escapeUnrecognizedEntities
+
+        # Convert any HTML, XML, or numeric entities in the attribute values.
+        convert_one = lambda x: self._convertEntities(parser.builder, x)
+        def convert(kval):
+            k, val = kval
+            if val is None:
+                return kval
+            return (k, re.sub("&(#\d+|#x[0-9a-fA-F]+|\w+);", convert_one, val))
+        if isinstance(attrs, types.DictType):
+            self.attrs = [convert(kv) for kv in attrs.items()]
+        else:
+            self.attrs = map(convert, attrs)
+
+    def get(self, key, default=None):
+        """Returns the value of the 'key' attribute for the tag, or
+        the value given for 'default' if it doesn't have that
+        attribute."""
+        return self._getAttrMap().get(key, default)
+
+    def has_key(self, key):
+        return self._getAttrMap().has_key(key)
+
+    def __getitem__(self, key):
+        """tag[key] returns the value of the 'key' attribute for the tag,
+        and throws an exception if it's not there."""
+        return self._getAttrMap()[key]
+
+    def __iter__(self):
+        "Iterating over a tag iterates over its contents."
+        return iter(self.contents)
+
+    def __len__(self):
+        "The length of a tag is the length of its list of contents."
+        return len(self.contents)
+
+    def __contains__(self, x):
+        return x in self.contents
+
+    def __nonzero__(self):
+        "A tag is non-None even if it has no contents."
+        return True
+
+    def __setitem__(self, key, value):
+        """Setting tag[key] sets the value of the 'key' attribute for the
+        tag."""
+        self._getAttrMap()
+        self.attrMap[key] = value
+        found = False
+        for i in range(0, len(self.attrs)):
+            if self.attrs[i][0] == key:
+                self.attrs[i] = (key, value)
+                found = True
+        if not found:
+            self.attrs.append((key, value))
+        self._getAttrMap()[key] = value
+
+    def __delitem__(self, key):
+        "Deleting tag[key] deletes all 'key' attributes for the tag."
+        for item in self.attrs:
+            if item[0] == key:
+                self.attrs.remove(item)
+                #We don't break because bad HTML can define the same
+                #attribute multiple times.
+            self._getAttrMap()
+            if self.attrMap.has_key(key):
+                del self.attrMap[key]
+
+    def __call__(self, *args, **kwargs):
+        """Calling a tag like a function is the same as calling its
+        findAll() method. Eg. tag('a') returns a list of all the A tags
+        found within this tag."""
+        return apply(self.findAll, args, kwargs)
+
+    def __getattr__(self, tag):
+        #print "Getattr %s.%s" % (self.__class__, tag)
+        if len(tag) > 3 and tag.rfind('Tag') == len(tag)-3:
+            return self.find(tag[:-3])
+        elif tag.find('__') != 0:
+            return self.find(tag)
+        raise AttributeError, "'%s' object has no attribute '%s'" % (self.__class__, tag)
+
+    def __eq__(self, other):
+        """Returns true iff this tag has the same name, the same attributes,
+        and the same contents (recursively) as the given tag.
+
+        NOTE: right now this will return false if two tags have the
+        same attributes in a different order. Should this be fixed?"""
+        if not hasattr(other, 'name') or not hasattr(other, 'attrs') or not hasattr(other, 'contents') or self.name != other.name or self.attrs != other.attrs or len(self) != len(other):
+            return False
+        for i in range(0, len(self.contents)):
+            if self.contents[i] != other.contents[i]:
+                return False
+        return True
+
+    def __ne__(self, other):
+        """Returns true iff this tag is not identical to the other tag,
+        as defined in __eq__."""
+        return not self == other
+
+    def __repr__(self, encoding=DEFAULT_OUTPUT_ENCODING):
+        """Renders this tag as a string."""
+        return self.decode(eventualEncoding=encoding)
+
+    BARE_AMPERSAND_OR_BRACKET = re.compile("([<>]|"
+                                           + "&(?!#\d+;|#x[0-9a-fA-F]+;|\w+;)"
+                                           + ")")
+
+    def _sub_entity(self, x):
+        """Used with a regular expression to substitute the
+        appropriate XML entity for an XML special character."""
+        return "&" + self.XML_SPECIAL_CHARS_TO_ENTITIES[x.group(0)[0]] + ";"
+
+    def __unicode__(self):
+        return self.decode()
+
+    def __str__(self):
+        return self.encode()
+
+    def encode(self, encoding=DEFAULT_OUTPUT_ENCODING,
+               prettyPrint=False, indentLevel=0):
+        return self.decode(prettyPrint, indentLevel, encoding).encode(encoding)
+
+    def decode(self, prettyPrint=False, indentLevel=0,
+               eventualEncoding=DEFAULT_OUTPUT_ENCODING):
+        """Returns a string or Unicode representation of this tag and
+        its contents. To get Unicode, pass None for encoding."""
+
+        attrs = []
+        if self.attrs:
+            for key, val in self.attrs:
+                fmt = '%s="%s"'
+                if isString(val):
+                    if (self.containsSubstitutions
+                        and eventualEncoding is not None
+                        and '%SOUP-ENCODING%' in val):
+                        val = self.substituteEncoding(val, eventualEncoding)
+
+                    # The attribute value either:
+                    #
+                    # * Contains no embedded double quotes or single quotes.
+                    #   No problem: we enclose it in double quotes.
+                    # * Contains embedded single quotes. No problem:
+                    #   double quotes work here too.
+                    # * Contains embedded double quotes. No problem:
+                    #   we enclose it in single quotes.
+                    # * Embeds both single _and_ double quotes. This
+                    #   can't happen naturally, but it can happen if
+                    #   you modify an attribute value after parsing
+                    #   the document. Now we have a bit of a
+                    #   problem. We solve it by enclosing the
+                    #   attribute in single quotes, and escaping any
+                    #   embedded single quotes to XML entities.
+                    if '"' in val:
+                        fmt = "%s='%s'"
+                        if "'" in val:
+                            # TODO: replace with apos when
+                            # appropriate.
+                            val = val.replace("'", "&squot;")
+
+                    # Now we're okay w/r/t quotes. But the attribute
+                    # value might also contain angle brackets, or
+                    # ampersands that aren't part of entities. We need
+                    # to escape those to XML entities too.
+                    val = self.BARE_AMPERSAND_OR_BRACKET.sub(self._sub_entity, val)
+                if val is None:
+                    # Handle boolean attributes.
+                    decoded = key
+                else:
+                    decoded = fmt % (key, val)
+                attrs.append(decoded)
+        close = ''
+        closeTag = ''
+        if self.isSelfClosing:
+            close = ' /'
+        else:
+            closeTag = '</%s>' % self.name
+
+        indentTag, indentContents = 0, 0
+        if prettyPrint:
+            indentTag = indentLevel
+            space = (' ' * (indentTag-1))
+            indentContents = indentTag + 1
+        contents = self.decodeContents(prettyPrint, indentContents,
+                                       eventualEncoding)
+        if self.hidden:
+            s = contents
+        else:
+            s = []
+            attributeString = ''
+            if attrs:
+                attributeString = ' ' + ' '.join(attrs)
+            if prettyPrint:
+                s.append(space)
+            s.append('<%s%s%s>' % (self.name, attributeString, close))
+            if prettyPrint:
+                s.append("\n")
+            s.append(contents)
+            if prettyPrint and contents and contents[-1] != "\n":
+                s.append("\n")
+            if prettyPrint and closeTag:
+                s.append(space)
+            s.append(closeTag)
+            if prettyPrint and closeTag and self.nextSibling:
+                s.append("\n")
+            s = ''.join(s)
+        return s
+
+    def decompose(self):
+        """Recursively destroys the contents of this tree."""
+        contents = [i for i in self.contents]
+        for i in contents:
+            if isinstance(i, Tag):
+                i.decompose()
+            else:
+                i.extract()
+        self.extract()
+
+    def prettify(self, encoding=DEFAULT_OUTPUT_ENCODING):
+        return self.encode(encoding, True)
+
+    def encodeContents(self, encoding=DEFAULT_OUTPUT_ENCODING,
+                       prettyPrint=False, indentLevel=0):
+        return self.decodeContents(prettyPrint, indentLevel).encode(encoding)
+
+    def decodeContents(self, prettyPrint=False, indentLevel=0,
+                       eventualEncoding=DEFAULT_OUTPUT_ENCODING):
+        """Renders the contents of this tag as a string in the given
+        encoding. If encoding is None, returns a Unicode string.."""
+        s=[]
+        for c in self:
+            text = None
+            if isinstance(c, NavigableString):
+                text = c.decodeGivenEventualEncoding(eventualEncoding)
+            elif isinstance(c, Tag):
+                s.append(c.decode(prettyPrint, indentLevel, eventualEncoding))
+            if text and prettyPrint:
+                text = text.strip()
+            if text:
+                if prettyPrint:
+                    s.append(" " * (indentLevel-1))
+                s.append(text)
+                if prettyPrint:
+                    s.append("\n")
+        return ''.join(s)
+
+    #Soup methods
+
+    def find(self, name=None, attrs={}, recursive=True, text=None,
+             **kwargs):
+        """Return only the first child of this Tag matching the given
+        criteria."""
+        r = None
+        l = self.findAll(name, attrs, recursive, text, 1, **kwargs)
+        if l:
+            r = l[0]
+        return r
+    findChild = find
+
+    def findAll(self, name=None, attrs={}, recursive=True, text=None,
+                limit=None, **kwargs):
+        """Extracts a list of Tag objects that match the given
+        criteria.  You can specify the name of the Tag and any
+        attributes you want the Tag to have.
+
+        The value of a key-value pair in the 'attrs' map can be a
+        string, a list of strings, a regular expression object, or a
+        callable that takes a string and returns whether or not the
+        string matches for some custom definition of 'matches'. The
+        same is true of the tag name."""
+        generator = self.recursiveChildGenerator
+        if not recursive:
+            generator = self.childGenerator
+        return self._findAll(name, attrs, text, limit, generator, **kwargs)
+    findChildren = findAll
+
+    # Pre-3.x compatibility methods. Will go away in 4.0.
+    first = find
+    fetch = findAll
+
+    def fetchText(self, text=None, recursive=True, limit=None):
+        return self.findAll(text=text, recursive=recursive, limit=limit)
+
+    def firstText(self, text=None, recursive=True):
+        return self.find(text=text, recursive=recursive)
+
+    # 3.x compatibility methods. Will go away in 4.0.
+    def renderContents(self, encoding=DEFAULT_OUTPUT_ENCODING,
+                       prettyPrint=False, indentLevel=0):
+        if encoding is None:
+            return self.decodeContents(prettyPrint, indentLevel, encoding)
+        else:
+            return self.encodeContents(encoding, prettyPrint, indentLevel)
+
+
+    #Private methods
+
+    def _getAttrMap(self):
+        """Initializes a map representation of this tag's attributes,
+        if not already initialized."""
+        if not getattr(self, 'attrMap'):
+            self.attrMap = {}
+            for (key, value) in self.attrs:
+                self.attrMap[key] = value
+        return self.attrMap
+
+    #Generator methods
+    def childGenerator(self):
+        for i in range(0, len(self.contents)):
+            yield self.contents[i]
+        raise StopIteration
+
+    def recursiveChildGenerator(self):
+        if not len(self.contents):
+            raise StopIteration
+        stopNode = self._lastRecursiveChild().next
+        current = self.contents[0]
+        while current is not stopNode:
+            yield current
+            current = current.next
+
+# Next, a couple classes to represent queries and their results.
+class SoupStrainer:
+    """Encapsulates a number of ways of matching a markup element (tag or
+    text)."""
+
+    def __init__(self, name=None, attrs={}, text=None, **kwargs):
+        self.name = name
+        if isString(attrs):
+            kwargs['class'] = attrs
+            attrs = None
+        if kwargs:
+            if attrs:
+                attrs = attrs.copy()
+                attrs.update(kwargs)
+            else:
+                attrs = kwargs
+        self.attrs = attrs
+        self.text = text
+
+    def __str__(self):
+        if self.text:
+            return self.text
+        else:
+            return "%s|%s" % (self.name, self.attrs)
+
+    def searchTag(self, markupName=None, markupAttrs={}):
+        found = None
+        markup = None
+        if isinstance(markupName, Tag):
+            markup = markupName
+            markupAttrs = markup
+        callFunctionWithTagData = callable(self.name) \
+                                and not isinstance(markupName, Tag)
+
+        if (not self.name) \
+               or callFunctionWithTagData \
+               or (markup and self._matches(markup, self.name)) \
+               or (not markup and self._matches(markupName, self.name)):
+            if callFunctionWithTagData:
+                match = self.name(markupName, markupAttrs)
+            else:
+                match = True
+                markupAttrMap = None
+                for attr, matchAgainst in self.attrs.items():
+                    if not markupAttrMap:
+                         if hasattr(markupAttrs, 'get'):
+                            markupAttrMap = markupAttrs
+                         else:
+                            markupAttrMap = {}
+                            for k,v in markupAttrs:
+                                markupAttrMap[k] = v
+                    attrValue = markupAttrMap.get(attr)
+                    if not self._matches(attrValue, matchAgainst):
+                        match = False
+                        break
+            if match:
+                if markup:
+                    found = markup
+                else:
+                    found = markupName
+        return found
+
+    def search(self, markup):
+        #print 'looking for %s in %s' % (self, markup)
+        found = None
+        # If given a list of items, scan it for a text element that
+        # matches.
+        if isList(markup) and not isinstance(markup, Tag):
+            for element in markup:
+                if isinstance(element, NavigableString) \
+                       and self.search(element):
+                    found = element
+                    break
+        # If it's a Tag, make sure its name or attributes match.
+        # Don't bother with Tags if we're searching for text.
+        elif isinstance(markup, Tag):
+            if not self.text:
+                found = self.searchTag(markup)
+        # If it's text, make sure the text matches.
+        elif isinstance(markup, NavigableString) or \
+                 isString(markup):
+            if self._matches(markup, self.text):
+                found = markup
+        else:
+            raise Exception, "I don't know how to match against a %s" \
+                  % markup.__class__
+        return found
+
+    def _matches(self, markup, matchAgainst):
+        #print "Matching %s against %s" % (markup, matchAgainst)
+        result = False
+        if matchAgainst == True and type(matchAgainst) == types.BooleanType:
+            result = markup != None
+        elif callable(matchAgainst):
+            result = matchAgainst(markup)
+        else:
+            #Custom match methods take the tag as an argument, but all
+            #other ways of matching match the tag name as a string.
+            if isinstance(markup, Tag):
+                markup = markup.name
+            if markup is not None and not isString(markup):
+                markup = unicode(markup)
+            #Now we know that chunk is either a string, or None.
+            if hasattr(matchAgainst, 'match'):
+                # It's a regexp object.
+                result = markup and matchAgainst.search(markup)
+            elif (isList(matchAgainst)
+                  and (markup is not None or not isString(matchAgainst))):
+                result = markup in matchAgainst
+            elif hasattr(matchAgainst, 'items'):
+                result = markup.has_key(matchAgainst)
+            elif matchAgainst and isString(markup):
+                if isinstance(markup, unicode):
+                    matchAgainst = unicode(matchAgainst)
+                else:
+                    matchAgainst = str(matchAgainst)
+
+            if not result:
+                result = matchAgainst == markup
+        return result
+
+class ResultSet(list):
+    """A ResultSet is just a list that keeps track of the SoupStrainer
+    that created it."""
+    def __init__(self, source):
+        list.__init__([])
+        self.source = source
diff --git a/BeautifulSoup.py.3.diff b/element.py.3.diff
index ad3d6e1..4549edd 100644
--- a/BeautifulSoup.py.3.diff
+++ b/element.py.3.diff
@@ -1,7 +1,3 @@
-90c90
-< from HTMLParser import HTMLParser, HTMLParseError
----
-> from html.parser import HTMLParser, HTMLParseError
 92c92
 <     from htmlentitydefs import name2codepoint
 ---
diff --git a/to3.sh b/to3.sh
index f24a121..26b3246 100755
--- a/to3.sh
+++ b/to3.sh
@@ -1,6 +1,6 @@
 #!/bin/sh
 mkdir python3
-for i in BeautifulSoup*.py dammit.py
+for i in BeautifulSoupTests.py builder.py element.py dammit.py
 do
     cp $i python3/
     2to3-3.0 -x next $i | patch -p0 python3/$i
diff --git a/util.py b/util.py
new file mode 100644
index 0000000..693a7e2
--- /dev/null
+++ b/util.py
@@ -0,0 +1,29 @@
+# Helper functions and mixin classes for Beautiful Soup
+
+import types
+try:
+    set
+except NameError:
+    from sets import Set as set
+
+def isList(l):
+    """Convenience method that works with all 2.x versions of Python
+    to determine whether or not something is listlike."""
+    return ((hasattr(l, '__iter__') and not isString(l))
+            or (type(l) in (types.ListType, types.TupleType)))
+
+def isString(s):
+    """Convenience method that works with all 2.x versions of Python
+    to determine whether or not something is stringlike."""
+    try:
+        return isinstance(s, unicode) or isinstance(s, basestring)
+    except NameError:
+        return isinstance(s, str)
+
+def buildSet(args=None):
+    """Turns a list or a string into a set."""
+    if isinstance(args, str):
+        return set([args])
+    if args is None:
+        return set()
+    return set(args)