//////////////////////////////////////////////////////////////////////////////// // // Microsoft Research Singularity // // Copyright (c) Microsoft Corporation. All rights reserved. // // Note: This is a very limited Xml Reader. // // BUG: The parser cannot handle attributes in single quotes // Actually, that's a feature -- XML requires the quotes. // namespace Microsoft.Singularity.Xml { using System; using System.IO; using System.Collections; using System.Text; /// /// This is a simple XML Parser that does no validation. All it does it parse the syntax of XML /// public class XmlReader { private States state; private int lineNumber; XmlNode doc; public XmlReader() { state = States.START; lineNumber = 1; } private enum TokenType { NAME, STRING_LITERAL, OPERATOR, NONE } private enum States { START, NEED_ELEMENT_NAME, NEED_ATTRIBUTE_NAME, NEED_EQUALS_SIGN, NEED_ATTRIBUTE_VALUE, NEED_CLOSURE_BRACKET, NEED_CLOSURE_NAME, NEED_FINAL_BRACKET, QUESTION_MARK_ELEMENT, GET_STRINGS } public XmlNode Parse(byte[] xmlBytes) { MemoryStream ms = new MemoryStream(xmlBytes); StreamReader sr = new StreamReader(ms); doc = new XmlNode("::xml"); return ParseHelper(doc, sr); } public XmlNode Parse(string filePath) { StreamReader sr = new StreamReader(filePath); doc = new XmlNode("::xml"); return ParseHelper(doc, sr); } private XmlNode ParseHelper(XmlNode doc, TextReader sr) { TokenType type; string token = null; ArrayList xmlNodes = new ArrayList(); XmlNode curNode = null; Stack st = new Stack(); string curAttributeName = null; while (ReadToken(sr, out token, out type)) { switch (state) { case States.START: //Console.WriteLine("START"); if (type == TokenType.OPERATOR && token.Equals("<")) { state = States.NEED_ELEMENT_NAME; } else { // All other text is interpreted as freestanding text. // It had better occur within a tag! if (st.Count == 0) { throw new XmlException("Line " + lineNumber + ": Text must be inside a tag"); } XmlNode lastNode = (XmlNode)st.Peek(); lastNode.AddText(token); // Next state depends on whether we're at the end of the text-within-a-tag // already or not. if ((char)sr.Peek() == '<') { // Looks like we're already at the end state = States.START; } else { state = States.GET_STRINGS; } } break; case States.NEED_ELEMENT_NAME: //Console.WriteLine("NEED_ELEMENT_NAME"); if (type == TokenType.NAME) { state = States.NEED_ATTRIBUTE_NAME; curNode = new XmlNode(token, st.Count); //Console.WriteLine("Saw beginning of element: " + token); } else if (type == TokenType.OPERATOR && token.Equals("/")) { state = States.NEED_CLOSURE_NAME; } else if (type == TokenType.OPERATOR && (token.Equals("?") || token.Equals("!") || token.Equals('-'))) { // then go to the question mark state and ignore this element state = States.QUESTION_MARK_ELEMENT; } else { throw new XmlException("Line " + lineNumber + ": Can only begin a tag with / or a name"); } break; case States.NEED_ATTRIBUTE_NAME: //Console.WriteLine("NEED_ATTRIBUTE_NAME"); if (type == TokenType.NAME) { state = States.NEED_EQUALS_SIGN; curAttributeName = token; } else if (type == TokenType.OPERATOR && token.Equals(">")) { // hold onto this node so we can check it later state = States.START; bool stackEmpty = st.Count == 0; if (!stackEmpty) { XmlNode parent = (XmlNode)st.Peek(); parent.AddChild(curNode); //DebugStub.WriteLine("add {1} to {0}", // __arglist(parent.Name, curNode.Name)); } else doc.AddChild(curNode); st.Push(curNode); curNode = null; } else if (type == TokenType.OPERATOR && token.Equals("/")) { // this node is almost complete state = States.NEED_CLOSURE_BRACKET; } else { throw new XmlException("Line " + lineNumber + ": Must have either attributes, '/>', or '>' after the name of an element"); } break; case States.NEED_EQUALS_SIGN: //Console.WriteLine("NEED_EQUALS_SIGN"); if (type == TokenType.OPERATOR && token.Equals("=")) { state = States.NEED_ATTRIBUTE_VALUE; } else { throw new XmlException("Line " + lineNumber + ": Need an '=' after an attribute name"); } break; case States.NEED_ATTRIBUTE_VALUE: //Console.WriteLine("NEED_ATTRIBUTE_VALUE"); if (type == TokenType.STRING_LITERAL) { state = States.NEED_ATTRIBUTE_NAME; string unescaped_attribute_value = ExpandEntityReferences(token); curNode[curAttributeName] = unescaped_attribute_value; //DebugStub.WriteLine(" {2}.attr[{0}]={1}", // __arglist(curAttributeName ,unescaped_attribute_value, curNode.Name)); curAttributeName = null; } else { throw new XmlException("Line " + lineNumber + ": Must have an attribute value after the '=' in an XML node"); } break; case States.NEED_CLOSURE_BRACKET: //Console.WriteLine("NEED_CLOSURE_BRACKET"); if (type == TokenType.OPERATOR && token.Equals(">")) { // this node is done, and we don't have to check it state = States.START; bool stackEmpty = st.Count == 0; if (!stackEmpty) { XmlNode parent = (XmlNode)st.Peek(); parent.AddChild(curNode); } else doc.AddChild(curNode); } else { throw new XmlException("Line " + lineNumber + ": Must have a '>' after a closing '/' in an XML node"); } break; case States.NEED_CLOSURE_NAME: //Console.WriteLine("NEED_CLOSURE_NAME"); if (type == TokenType.NAME) { // pop the last XmlNode and make sure that this name matches. // Otherwise we don't have balanced open and close tags state = States.NEED_FINAL_BRACKET; XmlNode xmln = (XmlNode)st.Pop(); if (!token.Equals(xmln.Name)) { throw new XmlException("Line " + lineNumber + ": " + token + " does not match " + xmln.Name); } //Console.WriteLine("Saw end of element: " + token); } else { throw new XmlException("Line " + lineNumber + ": Must have a name after an opening />"); } break; case States.NEED_FINAL_BRACKET: //Console.WriteLine("NEED_FINAL_BRACKET"); if (type == TokenType.OPERATOR && token.Equals(">")) { state = States.START; } else { throw new XmlException("Line " + lineNumber + ": Must have a > after a closure tag's name"); } break; case States.QUESTION_MARK_ELEMENT: // just stay in this state until you see a '>' while ('>' != ReadCharacter(sr)) { ; } state = States.START; break; case States.GET_STRINGS: { // stay in this state until you see a '<' StringBuilder sb = new StringBuilder(); XmlNode prevNode = (XmlNode)st.Peek(); if (type == TokenType.OPERATOR && token.Equals("<")) { throw new XmlException("Unexpected tag beginning while in text state"); } sb.Append(token); while ((char)sr.Peek() != '<') { sb.Append((char)sr.Read()); } prevNode.AddText(sb.ToString()); state = States.START; //Console.WriteLine("Grabbed string data"); } break; } } sr.Close(); return doc; } /// /// This method expands XML entity references found in an input string. /// If an invalid entity reference is encountered, this method will throw /// XmlException. /// /// The string to search for entity references. /// The expanded string. private static string ExpandEntityReferences(string input) { // In most cases, there are no entity references. Check for that case now. // If we do find an entity reference, then the work isn't wasted. int start = input.IndexOf('&'); if (start == -1) return input; StringBuilder buffer = new StringBuilder(); buffer.Append(input, 0, start); start++; for (;;) { // At this point, 'start' points to a named XML entity. // locate the entity name. int end = input.IndexOf(';', start); if (end == -1) throw new XmlException("An invalid entity reference was found. '&' is present, but there is no matching ';'."); int name_length = end - start; if (name_length == 0) throw new XmlException("An invalid entity reference was found. '&;' is not a valid entity reference."); string entity_name = input.Substring(start, name_length); string value; switch (entity_name) { case "amp": value = "&"; break; case "lt": value = "<"; break; case "gt": value = ">"; break; case "quot": value = "\""; break; default: throw new XmlException(String.Format("An invalid entity reference was found. The entity '&{0};' is not recognized.", entity_name)); } buffer.Append(value); // Are there any more entity references in this string? int next = input.IndexOf('&', end + 1); if (next == -1) { // There are no more entity references in the string. // Append the rest of the string. buffer.Append(input, end + 1, input.Length - end - 1); return buffer.ToString(); } // Yes, there are more references. Keep looping, preserving // the same meaning of 'start' as when we entered this loop. buffer.Append(input, end + 1, next - (end + 1)); start = next + 1; // skip over next '&' } } private char ReadCharacter(TextReader sr) { char tempChar = (char)sr.Read(); if (tempChar == '\n') { lineNumber++; } return tempChar; } private bool ReadToken(TextReader sr, out string token, out TokenType type) { // fortunately for us, the first character tells us exactly which of the three // types of token this is. [a-zA-Z] means it's a NAME, ["] means it's a STRING_LITERAL // and [<>/=] means it's an operator int nextVal = sr.Peek(); if (nextVal == -1) { // this is the end of the file token = ""; type = TokenType.NONE; return false; } char firstChar = unchecked((char)nextVal); // if it's in [a-zA-Z] if (('a' <= firstChar && firstChar <= 'z') || ('A' <= firstChar && firstChar <= 'Z')) { token = ReadName(sr); type = TokenType.NAME; return true; } else if (firstChar == '"') { token = ReadString(sr); type = TokenType.STRING_LITERAL; return true; } else if (firstChar == '<' || firstChar == '>' || firstChar == '/' || firstChar == '=' || firstChar == '?' || firstChar == '!' || firstChar == '-') { sr.Read(); token = firstChar.ToString(); type = TokenType.OPERATOR; return true; } else if (firstChar == ' ' || firstChar == '\n' || (int)firstChar == 13 || firstChar == '\t') { // throw away all whitespace while (firstChar == ' ' || firstChar == '\n' || (int)firstChar == 13 || firstChar == '\t') { char tempChar = (char)sr.Read(); if (tempChar == '\n') { // increment the line count lineNumber++; } firstChar = (char)sr.Peek(); } return ReadToken(sr, out token, out type); } else { sr.Read(); type = TokenType.NONE; token = firstChar.ToString(); return true; // throw new XmlException("Line " + lineNumber + ": '" + firstChar + "' is not a valid first character for a token"); } } private string ReadName(TextReader sr) { StringBuilder sb = new StringBuilder(); // add characters to our string until we see something that's not in [a-zA-Z0-9_-.:] char read = (char)sr.Peek(); while (('a' <= read && read <= 'z') || ('A' <= read && read <= 'Z') || ('0' <= read && read <= '9') || read == '-' || read == '_' || read == '.' || read == ':') { sb.Append((char)sr.Read()); read = (char)sr.Peek(); } return sb.ToString(); } private string ReadString(TextReader sr) { StringBuilder sb = new StringBuilder(); // we need to read this string and translate it into a string literal. // this means that we have (for the moment) two magic characters: \ and " // "\\" means \ and "\"" means " bool translateNext = false; char read = (char)sr.Peek(); if (read != '"') { throw new XmlException("Line " + lineNumber + ": Cannot start a string literal with " + (char)read + ". You must use a '\"'"); } // drop the '"' on the floor sr.Read(); read = (char)sr.Read(); while (read != '"' || translateNext) { if (!translateNext) { if (read == '\\') { translateNext = true; } else { sb.Append(read); read = (char)sr.Read(); } } else { translateNext = false; if (read == '\\') { sb.Append("\\"); read = (char)sr.Read(); } else if (read == '"') { sb.Append("\""); read = (char)sr.Read(); } else { throw new XmlException("Line " + lineNumber + ": Invalid escape sequence: \\" + read); } } } return sb.ToString(); } } }