magic-xml 0.2013.04.14

data/lib/magic_xml.rb

@@ -0,0 +1,1379 @@
+# Needed for parsing
+require 'rexml/parsers/baseparser'
+# Needed for fetching XMLs from the Internet
+require 'uri'
+require 'net/http'
+
+# FIXME: Make comment formatting RDoc-friendly. It's not always so now.
+
+# In Ruby 2 Symbol will be a subclass of String, and
+# this won't be needed any more. Before then...
+class Symbol
+    include Comparable
+    def <=>(other)
+        raise ArgumentError.new("comparison of #{self.class} with #{other.class} failed") unless other.is_a? Symbol
+        to_s <=> other.to_s
+    end
+    
+    alias_method :eqeqeq_before_magic_xml, :===
+    def ===(*args, &blk)
+        if args.size >= 1 and args[0].is_a? XML
+            self == args[0].name
+        else
+            eqeqeq_before_magic_xml(*args, &blk)
+        end
+    end
+end
+
+class Hash
+    alias_method :eqeqeq_before_magic_xml, :===
+    def ===(*args, &blk)
+        if args.size >= 1 and args[0].is_a? XML
+            all?{|k,v| v === args[0][k]}
+        else
+            eqeqeq_before_magic_xml(*args, &blk)
+        end
+    end
+end
+
+class String
+    # Escape string for output as XML text (< > &)
+    def xml_escape
+        replacements = {"<" => "&lt;", ">" => "&gt;", "&" => "&amp;" }
+        gsub(/([<>&])/) { replacements[$1] }
+    end
+    # Escape characters for output as XML attribute values (< > & ' ")
+    def xml_attr_escape
+        replacements = {"<" => "&lt;", ">" => "&gt;", "&" => "&amp;", "\"" => "&quot;", "'" => "&apos;"}
+        gsub(/([<>&\'\"])/) { replacements[$1] }
+    end
+    # Unescape entities
+    # Supports:
+    # * Full set of HTML-compatible named entities
+    # * Decimal entities &#1234;
+    # * Hex entities &#xA0b1;
+    def xml_unescape(extra_entities=nil)
+        @@xhtml_entity_replacements ||= {
+            'nbsp' => 160,
+            'iexcl' => 161,
+            'cent' => 162,
+            'pound' => 163,
+            'curren' => 164,
+            'yen' => 165,
+            'brvbar' => 166,
+            'sect' => 167,
+            'uml' => 168,
+            'copy' => 169,
+            'ordf' => 170,
+            'laquo' => 171,
+            'not' => 172,
+            'shy' => 173,
+            'reg' => 174,
+            'macr' => 175,
+            'deg' => 176,
+            'plusmn' => 177,
+            'sup2' => 178,
+            'sup3' => 179,
+            'acute' => 180,
+            'micro' => 181,
+            'para' => 182,
+            'middot' => 183,
+            'cedil' => 184,
+            'sup1' => 185,
+            'ordm' => 186,
+            'raquo' => 187,
+            'frac14' => 188,
+            'frac12' => 189,
+            'frac34' => 190,
+            'iquest' => 191,
+            'Agrave' => 192,
+            'Aacute' => 193,
+            'Acirc' => 194,
+            'Atilde' => 195,
+            'Auml' => 196,
+            'Aring' => 197,
+            'AElig' => 198,
+            'Ccedil' => 199,
+            'Egrave' => 200,
+            'Eacute' => 201,
+            'Ecirc' => 202,
+            'Euml' => 203,
+            'Igrave' => 204,
+            'Iacute' => 205,
+            'Icirc' => 206,
+            'Iuml' => 207,
+            'ETH' => 208,
+            'Ntilde' => 209,
+            'Ograve' => 210,
+            'Oacute' => 211,
+            'Ocirc' => 212,
+            'Otilde' => 213,
+            'Ouml' => 214,
+            'times' => 215,
+            'Oslash' => 216,
+            'Ugrave' => 217,
+            'Uacute' => 218,
+            'Ucirc' => 219,
+            'Uuml' => 220,
+            'Yacute' => 221,
+            'THORN' => 222,
+            'szlig' => 223,
+            'agrave' => 224,
+            'aacute' => 225,
+            'acirc' => 226,
+            'atilde' => 227,
+            'auml' => 228,
+            'aring' => 229,
+            'aelig' => 230,
+            'ccedil' => 231,
+            'egrave' => 232,
+            'eacute' => 233,
+            'ecirc' => 234,
+            'euml' => 235,
+            'igrave' => 236,
+            'iacute' => 237,
+            'icirc' => 238,
+            'iuml' => 239,
+            'eth' => 240,
+            'ntilde' => 241,
+            'ograve' => 242,
+            'oacute' => 243,
+            'ocirc' => 244,
+            'otilde' => 245,
+            'ouml' => 246,
+            'divide' => 247,
+            'oslash' => 248,
+            'ugrave' => 249,
+            'uacute' => 250,
+            'ucirc' => 251,
+            'uuml' => 252,
+            'yacute' => 253,
+            'thorn' => 254,
+            'yuml' => 255,
+            'quot' => 34,
+            'apos' => 39, # Wasn't present in the HTML entities set, but is defined in XML standard
+            'amp' => 38,
+            'lt' => 60,
+            'gt' => 62,
+            'OElig' => 338,
+            'oelig' => 339,
+            'Scaron' => 352,
+            'scaron' => 353,
+            'Yuml' => 376,
+            'circ' => 710,
+            'tilde' => 732,
+            'ensp' => 8194,
+            'emsp' => 8195,
+            'thinsp' => 8201,
+            'zwnj' => 8204,
+            'zwj' => 8205,
+            'lrm' => 8206,
+            'rlm' => 8207,
+            'ndash' => 8211,
+            'mdash' => 8212,
+            'lsquo' => 8216,
+            'rsquo' => 8217,
+            'sbquo' => 8218,
+            'ldquo' => 8220,
+            'rdquo' => 8221,
+            'bdquo' => 8222,
+            'dagger' => 8224,
+            'Dagger' => 8225,
+            'permil' => 8240,
+            'lsaquo' => 8249,
+            'rsaquo' => 8250,
+            'euro' => 8364,
+            'fnof' => 402,
+            'Alpha' => 913,
+            'Beta' => 914,
+            'Gamma' => 915,
+            'Delta' => 916,
+            'Epsilon' => 917,
+            'Zeta' => 918,
+            'Eta' => 919,
+            'Theta' => 920,
+            'Iota' => 921,
+            'Kappa' => 922,
+            'Lambda' => 923,
+            'Mu' => 924,
+            'Nu' => 925,
+            'Xi' => 926,
+            'Omicron' => 927,
+            'Pi' => 928,
+            'Rho' => 929,
+            'Sigma' => 931,
+            'Tau' => 932,
+            'Upsilon' => 933,
+            'Phi' => 934,
+            'Chi' => 935,
+            'Psi' => 936,
+            'Omega' => 937,
+            'alpha' => 945,
+            'beta' => 946,
+            'gamma' => 947,
+            'delta' => 948,
+            'epsilon' => 949,
+            'zeta' => 950,
+            'eta' => 951,
+            'theta' => 952,
+            'iota' => 953,
+            'kappa' => 954,
+            'lambda' => 955,
+            'mu' => 956,
+            'nu' => 957,
+            'xi' => 958,
+            'omicron' => 959,
+            'pi' => 960,
+            'rho' => 961,
+            'sigmaf' => 962,
+            'sigma' => 963,
+            'tau' => 964,
+            'upsilon' => 965,
+            'phi' => 966,
+            'chi' => 967,
+            'psi' => 968,
+            'omega' => 969,
+            'thetasym' => 977,
+            'upsih' => 978,
+            'piv' => 982,
+            'bull' => 8226,
+            'hellip' => 8230,
+            'prime' => 8242,
+            'Prime' => 8243,
+            'oline' => 8254,
+            'frasl' => 8260,
+            'weierp' => 8472,
+            'image' => 8465,
+            'real' => 8476,
+            'trade' => 8482,
+            'alefsym' => 8501,
+            'larr' => 8592,
+            'uarr' => 8593,
+            'rarr' => 8594,
+            'darr' => 8595,
+            'harr' => 8596,
+            'crarr' => 8629,
+            'lArr' => 8656,
+            'uArr' => 8657,
+            'rArr' => 8658,
+            'dArr' => 8659,
+            'hArr' => 8660,
+            'forall' => 8704,
+            'part' => 8706,
+            'exist' => 8707,
+            'empty' => 8709,
+            'nabla' => 8711,
+            'isin' => 8712,
+            'notin' => 8713,
+            'ni' => 8715,
+            'prod' => 8719,
+            'sum' => 8721,
+            'minus' => 8722,
+            'lowast' => 8727,
+            'radic' => 8730,
+            'prop' => 8733,
+            'infin' => 8734,
+            'ang' => 8736,
+            'and' => 8743,
+            'or' => 8744,
+            'cap' => 8745,
+            'cup' => 8746,
+            'int' => 8747,
+            'there4' => 8756,
+            'sim' => 8764,
+            'cong' => 8773,
+            'asymp' => 8776,
+            'ne' => 8800,
+            'equiv' => 8801,
+            'le' => 8804,
+            'ge' => 8805,
+            'sub' => 8834,
+            'sup' => 8835,
+            'nsub' => 8836,
+            'sube' => 8838,
+            'supe' => 8839,
+            'oplus' => 8853,
+            'otimes' => 8855,
+            'perp' => 8869,
+            'sdot' => 8901,
+            'lceil' => 8968,
+            'rceil' => 8969,
+            'lfloor' => 8970,
+            'rfloor' => 8971,
+            'lang' => 9001,
+            'rang' => 9002,
+            'loz' => 9674,
+            'spades' => 9824,
+            'clubs' => 9827,
+            'hearts' => 9829,
+            'diams' => 9830,
+        }
+        gsub(/&(?:([a-zA-Z]+)|#([0-9]+)|#x([a-fA-F0-9]+));/) {
+            if $1 then
+                v = @@xhtml_entity_replacements[$1]
+                # Nonstandard entity
+                unless v
+                    if extra_entities.is_a? Proc
+                        v = extra_entities.call($1)
+                    # Well, we expect a Hash here, but any container will do.
+                    # As long as it's not a nil.
+                    elsif extra_entities
+                        v = extra_entities[$1]
+                    end
+                end
+                raise "Unknown escape #{$1}" unless v
+            elsif $2
+                v = $2.to_i
+            else
+                v = $3.hex
+            end
+            # v can be a String or an Integer
+            if v.is_a? String then v else [v].pack('U') end
+        }
+    end
+    def xml_parse
+        XML.parse(self)
+    end
+end
+
+class File
+    def xml_parse
+        XML.parse(self)
+    end
+end
+
+class Array
+    # children of any element
+    def children(*args, &blk)
+        res = []
+        each{|c|
+            res += c.children(*args, &blk) if c.is_a? XML
+        }
+        res
+    end
+    # descendants of any element
+    def descendants(*args, &blk)
+        res = []
+        each{|c|
+            res += c.descendants(*args, &blk) if c.is_a? XML
+        }
+        res
+    end
+end
+
+# Methods of Enumerable.
+# It is not easy to design good methods, because XML
+# is not really "a container", it just acts as one sometimes.
+# Generally:
+# * Methods that return nil should work
+# * Methods that return an element should work
+# * Methods that return a container should return XML container, not Array
+# * Conversion methods should convert
+#
+# FIXME: Many methods use .dup, but do we want a shallow or a deep copy ?
+class XML
+    include Enumerable
+    # Default any? is ok
+    # Default all? is ok
+
+    # Iterate over children, possibly with a selector
+    def each(*selector, &blk)
+        children(*selector, &blk)
+        self
+    end
+
+    # Sort XML children of XML element.
+    def sort_by(*args, &blk)
+        self.dup{ @contents = @contents.select{|c| c.is_a? XML}.sort_by(*args, &blk) }
+    end
+
+    # Sort children of XML element.
+    def children_sort_by(*args, &blk)
+        self.dup{ @contents = @contents.sort_by(*args, &blk) }
+    end
+
+    # Sort children of XML element.
+    #
+    # Using sort is highly wrong, as XML (and XML-extras) is not even Comparable.
+    # Use sort_by instead.
+    #
+    # Unless you define your own XML#<=> operator, or do something equally weird.
+    def sort(*args, &blk)
+        self.dup{ @contents = @contents.sort(*args, &blk) }
+    end
+
+    #collect/map
+    #detect/find
+    #each_cons
+    #each_slice
+    #each_with_index
+    #to_a
+    #entries
+    #enum_cons
+    #enum_slice
+    #enum
+    # grep
+    # include?/member?
+    # inject
+    # max/min
+    # max_by/min_by - Ruby 1.9
+    # partition
+    # reject
+    # sort
+    # sort_by
+    # to_set
+    # zip
+    # And Enumerable::Enumerator-generating methods
+end
+
+# Class methods
+class XML
+    # XML.foo! == xml!(:foo)
+    # XML.foo  == xml(:foo)
+    def self.method_missing(meth, *args, &blk) 
+        if meth.to_s =~ /^(.*)!$/
+            xml!($1.to_sym, *args, &blk)
+        else
+            XML.new(meth, *args, &blk)
+        end
+    end
+
+    # Read file and parse
+    def self.from_file(file)
+        file = File.open(file) if file.is_a? String
+        parse(file)
+    end
+
+    # Fetch URL and parse
+    # Supported:
+    # http://.../
+    # https://.../
+    # file:foo.xml
+    # string:<foo/>
+    def self.from_url(url)
+        if url =~ /^string:(.*)$/m
+            parse($1)
+        elsif url =~ /^file:(.*)$/m
+            from_file($1)
+        elsif url =~ /^http(s?):/
+            ssl = ($1 == "s")
+            # No, seriously - Ruby needs something better than net/http
+            # Something that groks basic auth and queries and redirects automatically:
+            # HTTP_LIBRARY.get_content("http://username:passwd/u.r.l/?query")
+            # URI parsing must go inside the library, client programs
+            # should have nothing to do with it
+
+            # net/http is really inconvenient to use here
+            u = URI.parse(url)
+            # You're not seeing this:
+            if u.query then
+                path = u.path + "?" + u.query
+            else
+                path = u.path
+            end
+            req = Net::HTTP::Get.new(path)
+            if u.userinfo
+                username, passwd = u.userinfo.split(/:/,2)
+                req.basic_auth username, passwd
+            end
+            if ssl
+                # NOTE: You need libopenssl-ruby installed
+                # if you want to use HTTPS. Ubuntu is broken
+                # as it doesn't provide it in the default packages.
+                require 'net/https'
+                http = Net::HTTP.new(u.host, u.port)
+                http.use_ssl = true
+                http.verify_mode = OpenSSL::SSL::VERIFY_NONE
+            else
+                http = Net::HTTP.new(u.host, u.port)
+            end
+            
+            res = http.start {|http| http.request(req) }
+            # TODO: Throw a more meaningful exception
+            parse(res.body)
+        else
+            raise "URL protocol #{url} not supported (http, https, file, string are supported)"
+        end
+    end
+
+    # Like CDuce load_xml
+    # The path can be:
+    # * file handler
+    # * URL (a string with :)
+    # * file name (a string without :)
+    def self.load(obj)
+        if obj.is_a? String
+            if obj.include? ":"
+                from_url(obj)
+            else
+                from_file(obj)
+            end
+        else
+            parse(obj)
+        end
+    end
+
+    # Parse XML in mixed stream/tree mode
+    # Basically the idea is that every time we get start element,
+    # we ask the block what to do about it.
+    # If it wants a tree below it, it should call e.tree
+    # If a tree was requested, elements below the current one
+    # are *not* processed. If it wasn't, they are.
+    #
+    # For example:
+    #  <foo><bar/></foo><foo2/>
+    #  yield <foo> ... </foo>
+    #  .complete! called
+    #  process <foo2> next
+    #
+    # But:
+    #  <foo><bar/></foo><foo2/>
+    #  yield <foo> ... </foo>
+    #  .complete! not called
+    #  process <bar> next
+    #
+    # FIXME: yielded values are not reusable for now
+    # FIXME: make more object-oriented
+    def self.parse_as_twigs(stream)
+        parser = REXML::Parsers::BaseParser.new stream
+        # We don't really need to keep the stack ;-)
+        stack = []
+        while true
+            event = parser.pull
+            case event[0]
+            when :start_element
+                # Now the evil part evil
+                attrs = {}
+                event[2].each{|k,v| attrs[k.to_sym] = v.xml_unescape}
+                node = XML.new(event[1].to_sym, attrs, *event[3..-1])
+                
+                # I can't say it's superelegant
+                class <<node
+                    attr_accessor :do_complete
+                    def complete!
+                        if @do_complete
+                            @do_complete.call
+                            @do_complete = nil
+                        end
+                    end
+                end
+                node.do_complete = proc{
+                    parse_subtree(node, parser)
+                }
+
+                yield(node)
+                if node.do_complete
+                    stack.push node
+                    node.do_complete = nil # It's too late, complete! shouldn't do anything now
+                end
+            when :end_element
+                stack.pop
+            when :end_document
+                return
+            else
+                # FIXME: Do the right thing.
+                # For now, ignore *everything* else
+                # This is totally incorrect, user might want to 
+                # see text, comments and stuff like that anyway
+            end
+        end
+    end
+    
+    # Basically it's a copy of self.parse, ugly ...
+    def self.parse_subtree(start_node, parser)
+        stack = [start_node]
+        res = nil
+        while true
+            event = parser.pull
+            case event[0]
+            when :start_element
+                attrs = {}
+                event[2].each{|k,v| attrs[k.to_sym] = v.xml_unescape}
+                stack << XML.new(event[1].to_sym, attrs, *event[3..-1])
+                if stack.size == 1
+                    res = stack[0] 
+                else
+                    stack[-2] << stack[-1]
+                end
+            when :end_element
+                stack.pop
+                return if stack == []
+            # Needs unescaping
+            when :text
+                 # Ignore whitespace
+                 if stack.size == 0
+                     next if event[1] !~ /\S/
+                     raise "Non-whitespace text out of document root"
+                 end
+                 stack[-1] << event[1].xml_unescape
+            # CDATA is already unescaped
+            when :cdata
+                 if stack.size == 0
+                     raise "CDATA out of the document root"
+                 end
+                 stack[-1] << event[1]
+            when :end_document
+                raise "Parse error: end_document inside a subtree, tags are not balanced"
+            when :xmldecl,:start_doctype,:end_doctype,:elementdecl,:processing_instruction
+                # Positivery ignore
+            when :comment,:externalentity,:entity,:attlistdecl,:notationdecl
+                # Ignore ???
+                #print "Ignored XML event #{event[0]} when parsing\n"
+            else
+                # Huh ? What's that ?
+                #print "Unknown XML event #{event[0]} when parsing\n"
+            end
+        end
+        res
+
+    end
+
+    # Parse XML using REXML. Available options:
+    # * :extra_entities => Proc or Hash (default = nil)
+    # * :remove_pretty_printing => true/false (default = false)
+    # * :comments => true/false (default = false)
+    # * :pi => true/false (default = false)
+    # * :normalize => true/false (default = false) - normalize
+    # * :multiple_roots => true/false (default=false) - document
+    #      can have any number of roots (instread of one).
+    #      Return all in an array instead of root/nil.
+    #      Also include non-elements (String/PI/Comment) in the return set !!!
+    #
+    # FIXME: :comments/:pi will break everything
+    # if there are comments/PIs outside document root.
+    # Now PIs are outside the document root more often than not,
+    # so we're pretty much screwed here.
+    #
+    # FIXME: Integrate all kinds of parse, and make them support extra options
+    #
+    # FIXME: Benchmark normalize!
+    #
+    # FIXME: Benchmark dup-based Enumerable methods
+    #
+    # FIXME: Make it possible to include bogus XML_Document superparent,
+    #        and to make it support out-of-root PIs/Comments
+    def self.parse(stream, options={})
+        extra_entities = options[:extra_entities]
+
+        parser = REXML::Parsers::BaseParser.new stream
+        stack = [[]]
+        
+        while true
+            event = parser.pull
+            case event[0]
+            when :start_element
+                attrs = {}
+                event[2].each{|k,v| attrs[k.to_sym] = v.xml_unescape(extra_entities) }
+                stack << XML.new(event[1].to_sym, attrs, event[3..-1])
+                stack[-2] << stack[-1]
+            when :end_element
+                stack.pop
+            # Needs unescaping
+            when :text
+                 e = event[1].xml_unescape(extra_entities)
+                 # Either inside root or in multi-root mode
+                 if stack.size > 1 or options[:multiple_roots]
+                     stack[-1] << e
+                 elsif event[1] !~ /\S/
+                     # Ignore out-of-root whitespace in single-root mode
+                 else
+                     raise "Non-whitespace text out of document root (and not in multiroot mode): #{event[1]}"
+                 end
+            # CDATA is already unescaped
+            when :cdata
+                e = event[1]
+                if stack.size > 1 or options[:multiple_roots]
+                    stack[-1] << e
+                else
+                    raise "CDATA out of the document root"
+                end
+            when :comment
+                next unless options[:comments]
+                e = XML_Comment.new(event[1])
+                if stack.size > 1 or options[:multiple_roots]
+                    stack[-1] << e
+                else
+                    # FIXME: Ugly !
+                    raise "Comments out of the document root"
+                end
+            when :processing_instruction
+                # FIXME: Real PI node
+                next unless options[:pi]
+                e = XML_PI.new(event[1], event[2])
+                if stack.size > 1 or options[:multiple_roots]
+                    stack[-1] << e
+                else
+                    # FIXME: Ugly !
+                    raise "Processing instruction out of the document root"
+                end
+            when :end_document
+                break
+            when :xmldecl,:start_doctype,:end_doctype,:elementdecl
+                # Positivery ignore
+            when :externalentity,:entity,:attlistdecl,:notationdecl
+                # Ignore ???
+                #print "Ignored XML event #{event[0]} when parsing\n"
+            else
+                # Huh ? What's that ?
+                #print "Unknown XML event #{event[0]} when parsing\n"
+            end
+        end
+        roots = stack[0]
+        
+        roots.each{|root| root.remove_pretty_printing!} if options[:remove_pretty_printing]
+        # :remove_pretty_printing does :normalize anyway
+        roots.each{|root| root.normalize!} if options[:normalize]
+        if options[:multiple_roots]
+            roots
+        else
+            roots[0]
+        end
+    end
+
+    # Parse a sequence. Equivalent to XML.parse(stream, :multiple_roots => true).
+    def self.parse_sequence(stream, options={})
+        o = options.dup
+        o[:multiple_roots] = true
+        parse(stream, o)
+    end
+
+    # Renormalize a string containing XML document
+    def self.renormalize(stream)
+        parse(stream).to_s
+    end
+
+    # Renormalize a string containing a sequence of XML documents
+    # and strings
+    # XMLrenormalize_sequence("<hello   />, <world></world>!") =>
+    # "<hello/>, <world/>!"
+    def self.renormalize_sequence(stream)
+        parse_sequence(stream).join
+    end
+end
+
+# Instance methods (other than those of Enumerable)
+class XML
+    attr_accessor :name, :attrs, :contents
+
+    # initialize can be run in many ways
+    # * XML.new
+    # * XML.new(:tag_symbol)
+    # * XML.new(:tag_symbol, {attributes})
+    # * XML.new(:tag_symbol, "children", "more", XML.new(...))
+    # * XML.new(:tag_symbol, {attributes}, "and", "children")
+    # * XML.new(:tag_symbol) { monadic code }
+    # * XML.new(:tag_symbol, {attributes}) { monadic code }
+    #
+    # Or even:
+    # * XML.new(:tag_symbol, "children") { and some monadic code }
+    # * XML.new(:tag_symbol, {attributes}, "children") { and some monadic code }
+    # But typically you won't be mixing these two style
+    #
+    # Attribute values can will be converted to strings
+    def initialize(*args, &blk)
+        @name     = nil
+        @attrs    = {}
+        @contents = []
+        @name = args.shift if args.size != 0
+        if args.size != 0 and args[0].is_a? Hash
+            args.shift.each{|k,v|
+                # Do automatic conversion here
+                # This also assures that the hashes are *not* shared
+                self[k] = v
+            }
+        end
+        # Expand Arrays passed as arguments
+        self << args
+        # FIXME: We'd rather not have people say @name = :foo there :-)
+        if blk
+            instance_eval(&blk)
+        end
+    end
+
+    # Convert to a well-formatted XML
+    def to_s
+        "<#{@name}" + @attrs.sort.map{|k,v| " #{k}='#{v.xml_attr_escape}'"}.join +
+        if @contents.size == 0
+            "/>"
+        else
+            ">" + @contents.map{|x| if x.is_a? String then x.xml_escape else x.to_s end}.join + "</#{name}>"
+        end
+    end
+
+    # Convert to a well-formatted XML, but without children information.
+    # This is a reasonable format for irb and debugging.
+    # If you want to see a few levels of children, call inspect(2) and so on
+    def inspect(include_children=0)
+        "<#{@name}" + @attrs.sort.map{|k,v| " #{k}='#{v.xml_attr_escape}'"}.join +
+        if @contents.size == 0
+            "/>"
+        elsif include_children == 0
+            ">...</#{name}>"
+        else
+            ">" + @contents.map{|x| if x.is_a? String then x.xml_escape else x.inspect(include_children-1) end}.join + "</#{name}>"
+        end
+    end
+
+    # Read attributes.
+    # Also works with pseudoattributes:
+    #  img[:@x] == img.child(:x).text # or nil if there isn't any.
+    def [](key)
+        if key.to_s[0] == ?@
+            tag = key.to_s[1..-1].to_sym
+            c = child(tag)
+            if c
+                c.text
+            else
+                nil
+            end
+        else
+            @attrs[key]
+        end
+    end
+
+    # Set attributes.
+    # Value is automatically converted to String, so you can say:
+    #  img[:x] = 200
+    # Also works with pseudoattributes:
+    #  foo[:@bar] = "x"
+    def []=(key, value)
+        if key.to_s[0] == ?@
+            tag = key.to_s[1..-1].to_sym
+            c = child(tag)
+            if c
+                c.contents = [value.to_s]
+            else
+                self << XML.new(tag, value.to_s)
+            end
+        else
+            @attrs[key] = value.to_s
+        end
+    end
+
+    # Add children.
+    # Possible uses:
+    # * Add single element
+    #  self << xml(...)
+    #  self << "foo"
+    # Add nothing:
+    #  self << nil  
+    # Add multiple elements (also works recursively):
+    #  self << [a, b, c] 
+    #  self << [a, [b, c], d] 
+    def <<(cnt)
+        if cnt.nil?
+            # skip
+        elsif cnt.is_a? Array
+            cnt.each{|elem| self << elem}
+        else
+            @contents << cnt
+        end
+        self
+    end
+
+    # Equality test, works as if XMLs were normalized, so:
+    #  XML.new(:foo, "Hello, ", "world") == XML.new(:foo, "Hello, world")
+    def ==(x)
+        return false unless x.is_a? XML
+        return false unless name == x.name and attrs == x.attrs
+        # Now the hard part, strings can be split in different ways
+        # empty string children are possible etc.
+        self_i = 0
+        othr_i = 0
+        while self_i != contents.size or othr_i != x.contents.size
+            # Ignore ""s
+            if contents[self_i].is_a? String and contents[self_i] == ""
+                self_i += 1
+                next
+            end
+            if x.contents[othr_i].is_a? String and x.contents[othr_i] == ""
+                othr_i += 1
+                next
+            end
+
+            # If one is finished and the other contains non-empty elements,
+            # they are not equal
+            return false if self_i == contents.size or othr_i == x.contents.size
+
+            # Are they both Strings ?
+            # Strings can be divided in different ways, and calling normalize!
+            # here would be rather expensive, so let's use this complicated
+            # algorithm
+            if contents[self_i].is_a? String and x.contents[othr_i].is_a? String
+                a = contents[self_i]
+                b = x.contents[othr_i]
+                self_i += 1
+                othr_i += 1
+                while a != "" or b != ""
+                    if a == b
+                        a = ""
+                        b = ""
+                    elsif a.size > b.size and a[0, b.size] == b
+                        a = a[b.size..-1]
+                        if x.contents[othr_i].is_a? String
+                            b = x.contents[othr_i]
+                            othr_i += 1
+                            next
+                        end
+                    elsif b.size > a.size and b[0, a.size] == a
+                        b = b[a.size..-1]
+                        if contents[self_i].is_a? String
+                            a = contents[self_i]
+                            self_i += 1
+                            next
+                        end
+                    else
+                        return false
+                    end
+                end
+                next
+            end
+
+            # OK, so at least one of them is not a String.
+            # Hopefully they're either both XMLs or one is an XML and the
+            # other is a String. It is also possible that contents contains
+            # something illegal, but we aren't catching that,
+            # so xml(:foo, Garbage.new) is going to at least equal itself.
+            # And we aren't, because xml(:foo, Garbage.new) == xml(:bar, Garbage.new)
+            # is going to return an honest false, and incoherent sanity
+            # check is worse than no sanity check.
+            #
+            # Oh yeah, they can be XML_PI or XML_Comment. In such case, this
+            # is ok.
+            return false unless contents[self_i] == x.contents[othr_i]
+            self_i += 1
+            othr_i += 1
+        end
+        return true
+    end
+
+    alias_method :real_method_missing, :method_missing
+    # Define all foo!-methods for monadic interface, so you can write:
+    # 
+    def method_missing(meth, *args, &blk) 
+        if meth.to_s =~ /^(.*)!$/
+            self << XML.new($1.to_sym, *args, &blk)
+        else
+            real_method_missing(meth, *args, &blk)
+        end
+    end
+
+    # Make monadic interface more "official"
+    # * node.exec! { foo!; bar! }
+    # is equivalent to
+    # * node << xml(:foo) << xml(:bar)
+    def exec!(&blk)
+        instance_eval(&blk)
+    end
+
+    # Select a subtree
+    # NOTE: Uses object_id of the start/end tags !
+    # They have to be the same, not just identical !
+    # <foo>0<a>1</a><b/><c/><d>2</d><e/>3</foo>.range(<a>1</a>, <d>2</d>)
+    # returns
+    # <foo><b/><c/></foo>
+    # start and end and their descendants are not included in
+    # the result tree.
+    # Either start or end can be nil.
+    # * If both start and end are nil, return whole tree.
+    # * If start is nil, return subtree up to range_end.
+    # * If start is not inside the tree, return nil.
+    # * If end is nil, return subtree from start
+    # * If end is not inside the tree, return subtree from start.
+    # * If end is before or below start, or they're the same node, the result is unspecified.
+    # * if end comes directly after start, or as first node when start==nil, return path reaching there.
+    def range(range_start, range_end, end_reached_cb=nil)
+        if range_start == nil
+            result = XML.new(name, attrs)
+        else
+            result = nil
+        end
+        @contents.each {|c|
+            # end reached !
+            if range_end and c.object_id == range_end.object_id
+                end_reached_cb.call if end_reached_cb
+                break
+            end
+            # start reached !
+            if range_start and c.object_id == range_start.object_id
+                result = XML.new(name, attrs)
+                next
+            end
+            if result # We already started
+                if c.is_a? XML
+                    break_me = false
+                    result.add! c.range(nil, range_end, lambda{ break_me = true })
+                    if break_me
+                        end_reached_cb.call if end_reached_cb
+                        break
+                    end
+                else # String/XML_PI/XML_Comment
+                    result.add! c
+                end
+            else
+                # Strings/XML_PI/XML_Comment obviously cannot start a range
+                if c.is_a? XML
+                    break_me = false
+                    r = c.range(range_start, range_end, lambda{ break_me = true })
+                    if r
+                        # start reached !
+                        result = XML.new(name, attrs, r)
+                    end
+                    if break_me
+                        # end reached !
+                        end_reached_cb.call if end_reached_cb
+                        break
+                    end
+                end
+            end
+        }
+        return result
+    end
+
+    # XML#subsequence is similar to XML#range, but instead of
+    # trimmed subtree in returns a list of elements
+    # The same elements are included in both cases, but here
+    # we do not include any parents !
+    #
+    # <foo><a/><b/><c/></foo>.range(a,c) => <foo><b/></foo>
+    # <foo><a/><b/><c/></foo>.subsequence(a,c) => <b/>
+    #
+    # <foo><a><a1/></a><b/><c/></foo>.range(a1,c) => <foo><a/><b/></foo> # Does <a/> make sense ?
+    # <foo><a><a1/></a><b/><c/></foo>.subsequence(a1,c) => <b/>
+    #
+    # <foo><a><a1/><a2/></a><b/><c/></foo>.range(a1,c) => <foo><a><a2/></a><b/></foo>
+    # <foo><a><a1/><a2/></a><b/><c/></foo>.subsequence(a1,c) => <a2/><b/>
+    #
+    # And we return [], not nil if nothing matches
+    def subsequence(range_start, range_end, start_seen_cb=nil, end_seen_cb=nil)
+        result = []
+        start_seen = range_start.nil?
+        @contents.each{|c|
+            if range_end and range_end.object_id == c.object_id
+                end_seen_cb.call if end_seen_cb
+                break 
+            end
+            if range_start and range_start.object_id == c.object_id
+                start_seen = true
+                start_seen_cb.call if start_seen_cb
+                next
+            end
+            if start_seen
+                if c.is_a? XML
+                    break_me = false
+                    result += c.subsequence(nil, range_end, nil, lambda{break_me=true})
+                    break if break_me
+                else # String/XML_PI/XML_Comment
+                    result << c
+                end
+            else
+                # String/XML_PI/XML_Comment cannot start a subsequence
+                if c.is_a? XML
+                    break_me = false
+                    result += c.subsequence(range_start, range_end, lambda{start_seen=true}, lambda{break_me=true})
+                    break if break_me
+                end
+            end
+        }
+        # Include starting tag if it was right from the range_start
+        # Otherwise, return just the raw sequence
+        result = [XML.new(@name, @attrs, result)] if range_start == nil
+        return result
+    end
+
+    # =~ for a few reasonable patterns
+    def =~(pattern)
+        if pattern.is_a? Symbol
+            @name == pattern
+        elsif pattern.is_a? Regexp
+            rv = text =~ pattern
+        else # Hash, Pattern_any, Pattern_all
+            pattern === self
+        end
+    end
+    
+    # Get rid of pretty-printing whitespace. Also normalizes the XML.
+    def remove_pretty_printing!(exceptions=nil)
+        normalize!
+        real_remove_pretty_printing!(exceptions)
+        normalize!
+    end
+
+    # normalize! is already recursive, so only one call at top level is needed.
+    # This helper method lets us avoid extra calls to normalize!.
+    def real_remove_pretty_printing!(exceptions=nil)
+        return if exceptions and exceptions.include? @name
+        each{|c|
+            if c.is_a? String
+                c.sub!(/^\s+/, "")
+                c.sub!(/\s+$/, "")
+                c.gsub!(/\s+/, " ")
+            elsif c.is_a? XML_PI or c.is_a? XML_Comment
+            else
+                c.real_remove_pretty_printing!(exceptions)
+            end
+        }
+    end
+
+    protected :real_remove_pretty_printing!
+
+    # Add pretty-printing whitespace. Also normalizes the XML.
+    def add_pretty_printing!
+        normalize!
+        real_add_pretty_printing!
+        normalize!
+    end
+    
+    def real_add_pretty_printing!(indent = "")
+        return if @contents.empty?
+        each{|c|
+            if c.is_a? XML
+                c.real_add_pretty_printing!(indent+"  ")
+            elsif c.is_a? String
+                c.gsub!(/\n\s*/, "\n#{indent}  ")
+            end
+        }
+        @contents = @contents.inject([]){|children, c| children + ["\n#{indent}  ", c]}+["\n#{indent}"]
+    end
+
+    protected :real_add_pretty_printing!
+
+    alias_method :raw_dup, :dup
+    # This is not a trivial method - first it does a *deep* copy,
+    # second it takes a block which is instance_eval'ed,
+    # so you can do things like:
+    # * node.dup{ @name = :foo }
+    # * node.dup{ self[:color] = "blue" }
+    def dup(&blk)
+        new_obj = self.raw_dup
+        # Attr values stay shared - ugly
+        new_obj.attrs = new_obj.attrs.dup
+        new_obj.contents = new_obj.contents.map{|c| c.dup}
+        
+        new_obj.instance_eval(&blk) if blk
+        return new_obj
+    end
+
+
+    # Add some String children (all attributes get to_s'ed)
+    def text!(*args)
+        args.each{|s| self << s.to_s}
+    end
+    # Add XML child
+    def xml!(*args, &blk)
+        @contents << XML.new(*args, &blk)
+    end
+
+    alias_method :add!, :<<
+    
+    # Normalization means joining strings
+    # and getting rid of ""s, recursively
+    def normalize!
+        new_contents = []
+        @contents.each{|c|
+            if c.is_a? String
+                next if c == ""
+                if new_contents[-1].is_a? String
+                    new_contents[-1] += c
+                    next
+                end
+            else
+                c.normalize!
+            end
+            new_contents.push c
+        }
+        @contents = new_contents
+    end
+
+    # Return text below the node, stripping all XML tags,
+    # "<foo>Hello, <bar>world</bar>!</foo>".xml_parse.text
+    # returns "Hello, world!"
+    def text
+        res = ""
+        @contents.each{|c|
+            if c.is_a? XML
+                res << c.text
+            elsif c.is_a? String
+                res << c
+            end # Ignore XML_PI/XML_Comment
+        }
+        res
+    end
+
+    # Equivalent to node.children(pat, *rest)[0]
+    # Returns nil if there aren't any matching children
+    def child(pat=nil, *rest)
+        children(pat, *rest) {|c|
+            return c
+        }
+        return nil
+    end
+
+    # Equivalent to node.descendants(pat, *rest)[0]
+    # Returns nil if there aren't any matching descendants
+    def descendant(pat=nil, *rest)
+        descendants(pat, *rest) {|c|
+            return c
+        }
+        return nil
+    end
+
+    # XML#children(pattern, more_patterns)
+    # Return all children of a node with tags matching tag.
+    # Also:
+    # * children(:a, :b) == children(:a).children(:b)
+    # * children(:a, :*, :c) == children(:a).descendants(:c)
+    def children(pat=nil, *rest, &blk)
+        return descendants(*rest, &blk) if pat == :*
+        res = []
+        @contents.each{|c|
+            if pat.nil? or pat === c
+                if rest == []
+                    res << c
+                    yield c if block_given?
+                else
+                    res += c.children(*rest, &blk)
+                end
+            end
+        }
+        res
+    end
+    
+    # * XML#descendants
+    # * XML#descendants(pattern)
+    # * XML#descendants(pattern, more_patterns)
+    #
+    # Return all descendants of a node matching the pattern.
+    # If pattern==nil, simply return all descendants.
+    # Optionally run a block on each of them if a block was given.
+    # If pattern==nil, also match Strings !
+    def descendants(pat=nil, *rest, &blk)
+        res = []
+        @contents.each{|c|
+            if pat.nil? or pat === c
+                if rest == []
+                    res << c
+                    yield c if block_given?
+                else
+                    res += c.children(*rest, &blk)
+                end
+            end
+            if c.is_a? XML
+                res += c.descendants(pat, *rest, &blk)
+            end
+        }
+        res
+    end
+    
+    # Change elements based on pattern
+    def deep_map(pat, &blk)
+        if self =~ pat
+            yield self
+        else
+            r = XML.new(self.name, self.attrs)
+            each{|c|
+                if c.is_a? XML
+                    r << c.deep_map(pat, &blk)
+                else
+                    r << c
+                end
+            }
+            r
+        end
+    end
+
+    # FIXME: do we want a shallow or a deep copy here ?
+    # Map children, but leave the name/attributes
+    def map(pat=nil)
+        r = XML.new(self.name, self.attrs)
+        each{|c|
+            if !pat || (c.is_a?(XML) && c =~ pat)
+                r << yield(c)
+            else
+                r << c
+            end
+        }
+        r
+    end
+end
+
+# FIXME: Is this even sane ?
+# * What about escaping and all that stuff ?
+# * Rest of the code assumes that everything is either XML or String
+class XML_PI
+    def initialize(c, t)
+        @c = c
+        @t = t
+    end
+    def to_s
+        "<?#{@c}#{@t}?>"
+    end
+end
+
+# FIXME: Is this even sane ?
+# * What about escaping and all that stuff ?
+# * Rest of the code assumes that everything is either XML or String
+# * There are some limitations on where one can put -s in the comment. Do not overdo.
+class XML_Comment
+    def initialize(c)
+        @c = c
+    end
+    def to_s
+        "<!--#{@c}-->"
+    end
+end
+
+# Syntactic sugar for XML.new
+def xml(*args, &blk)
+    XML.new(*args, &blk)
+end
+
+# xml! in XML { ... } - context adds node to parent
+# xml! in main context prints the argument (and returns it anyway)
+def xml!(*args, &blk)
+    node = xml(*args, &blk)
+    print node
+    node
+end
+
+# Perl 6 is supposed to have native support for something like that.
+# Constructor takes multiple patterns. The object matches if they all match.
+#
+# Usage:
+#  case foo
+#  when all(:foo, {:color => 'blue'}, /Hello/)
+#       print foo
+#  end
+class Patterns_all
+    def initialize(*patterns)
+        @patterns = patterns
+    end
+    def ===(obj)
+        @patterns.all?{|p| p === obj}
+    end
+end
+
+def all(*patterns)
+    Patterns_all.new(*patterns)
+end
+
+# Perl 6 is supposed to have native support for something like that.
+# Constructor takes multiple patterns. The object matches if they all match.
+#
+# Usage:
+#  case foo
+#  when all(:foo, any({:color => 'blue'}, {:color => 'red'}), /Hello/)
+#       print foo
+#  end
+class Patterns_any
+    def initialize(*patterns)
+        @patterns = patterns
+    end
+    def ===(obj)
+        @patterns.any?{|p| p === obj}
+    end
+end
+
+def any(*patterns)
+    Patterns_any.new(*patterns)
+end