documentconverter.cpp

sloedgy open sip stack source code

 /****************License************************************************
  *
  * Copyright 2000-2003.  ScanSoft, Inc.    
  *
  * Use of this software is subject to notices and obligations set forth 
  * in the SpeechWorks Public License - Software Version 1.2 which is 
  * included with this software. 
  *
  * ScanSoft is a registered trademark of ScanSoft, Inc., and OpenSpeech, 
  * SpeechWorks and the SpeechWorks logo are registered trademarks or 
  * trademarks of SpeechWorks International, Inc. in the United States 
  * and other countries.
  *
  ***********************************************************************/
 

 /***********************************************************************
  * This class performs several jobs.  Among them:
  *
  * 1) <grammar> contents are collapsed and blindly copied.
  *
  * 2) Explicit and implicit prompts are converted into <prompt> elements with
  *      full SSML headers and trailers.
  *      NOTE: <value>, <enumerate>, and <audio> don't contribute to the copy
  *      depth as they require special handling.
  *
  * 3) PCDATA consisting only of whitespace is discarded.
  *
  * 4) <error>, <help>, <noinput> and <nomatch> are converted into a <catch>
  *    with the appropriate event name.
  *
  * 5) <choice> elements may be assigned dtmf values.
  *
  * 6) Attribute conflicts are detected whenever possible.
  *
  * 7) <metadata> is ignored (except in <grammar> or <prompt>).
  *
  * 8) Assign xml:base and xml:lang to <grammar>, <prompt>, <field>, and <menu>.
  * 
  * NOTE: By performing these duties, the DocumentConverter greatly simplifies
  *       the rest of the interpreter.  Unfortunately, this means that this
  *       code is tricky to write and maintain.  PLEASE BE VERY CAREFUL THAT
  *       YOU UNDERSTAND WHAT IS HAPPENING HERE BEFORE MAKING ANY CHANGES.
  * 
  ***********************************************************************
  *
  * 1 Grammar processing
  *
  * The content of <grammar> elements is blindly copied.  This is controlled
  * by 'copyDepth' and 'inGrammar'.  This is possible because VoiceXML
  * grammars are entirely static.  Reducing them down to PCDATA at this stage
  * allows the interpreter to treat the entire grammar as a black box without
  * regard to supported types.
  *
  ***********************************************************************
  * 
  * 2 Prompt processing 
  *
  * A frequently noted defect of VoiceXML is that prompts may be both explicitly
  * declared (i.e. within <prompt>) or implicitly declared as raw CDATA.  In
  * this code, we convert all implicit prompts into explicit ones and collapse
  * most of the SSML into PCDATA.  At the end, prompts are composed only of 
  * those elements requiring special handling (i.e. <audio>, <enumerate>, and 
  * <value>) and PCDATA.
  *
  ***********************************************************************
  * 
  * 3 Whitespace stripping
  *
  * Because of implicit prompts, the XML parser is required to treat most
  * whitespace as significant information.  These 'empty prompts' are stripped
  * at this level to simplify later processing.
  *
  * Likewise, <choice> and <option> content is whitespace stripped (though this
  * task must be performed by the VXMLDocumentRep since Xerces may split PCDATA
  * across multiple calls to DocumentConverter::characters).
  *
  ***********************************************************************
  * 
  * 4 Event collapsing
  *
  * VoiceXML defines <error>, <help>, <noinput> and <nomatch> as aliases
  * for <catch event="...">.  These are converted at this layer (see
  * startElement) to simplify later processing.
  *
  ***********************************************************************
  * 
  * 5 Menu choice processing
  *
  * DTMF values may be assigned to <choice>s if necessary.
  *
  ***********************************************************************
  * 
  * 6 Attribute conflict detection
  *
  * There are many cases in the VoiceXML specification where conflicting
  * attributes result in error.badfetch.  These should (in future) be
  * detected by the XML schema and are caught at this level.
  *
  * One difficult case appears inside of <grammar> and <script> where either
  * the 'src' attribute or content is legal, but not both.  As grammars may
  * not appear inside scripts and vice versa, this is detected with a simple
  * scheme using the variables 'contentForbidden' and 'hasContent'.
  *
  ***********************************************************************
  *
  * 7 <metadata> stripped
  *
  * This content has no runtime impact.  It is stripped at this level
  * UNLESS it is inside a <grammar> or <prompt> tag.
  *
  ***********************************************************************
  *
  * 8 <prompt> & <grammar> always have xml:base and xml:lang attributes.
  *   <field>, <menu> always has xml:lang
  *
  * In an effort to keep these elements independent from the rest of VXML
  * this attribute is always set.  This allows the SRGS or SSML to
  * incorporate an xml:base in the root element which might be useful if
  * the contents are being sent to a seperate server.
  *
  * Likewise, a language is required for the type attribute on fields (used
  * for builtins) and for generating the CDATA grammars in menu choices.
  *
  ***********************************************************************/
  
#include <vxibuildopts.h>
#if P_VXI

 #include "DocumentConverter.hpp"
 #include "vxibuildopts.h"
 #include <xercesc/sax2/Attributes.hpp>
 #include <xercesc/sax/SAXParseException.hpp>
 #include <xercesc/sax/Locator.hpp>
 #include <vector>
 #include <algorithm>
 #include <sstream>                     // by ProcessNodeFinal
 #include "VXMLDocumentRep.hpp"
 #include "XMLChConverter.hpp"
 #include <iostream>
 
 //#############################################################################
 
#ifdef _MSC_VER
#pragma warning( disable:4062 )
#pragma warning( disable:4061 )
#endif

 enum {
   PRIV_ELEM_RangeStart    = 0x200,   // This should always be the first entry
 
   // Events
   PRIV_ELEM_ERROR,
   PRIV_ELEM_HELP,
   PRIV_ELEM_NOINPUT,
   PRIV_ELEM_NOMATCH,
   PRIV_ELEM_METADATA,
 
   // SSML
   PRIV_ELEM_BREAK,
   PRIV_ELEM_DESC,
   PRIV_ELEM_EMPHASIS,
   PRIV_ELEM_MARK,
   PRIV_ELEM_PARAGRAPH,
   PRIV_ELEM_PHONEME,
   PRIV_ELEM_PROSODY,
   PRIV_ELEM_SAYAS,
   PRIV_ELEM_SENTENCE,
   PRIV_ELEM_SUB,
   PRIV_ELEM_VOICE
 };
 
 enum {
   PRIV_ATTRIB_RangeStart  = 0x200,   // This should always be the first entry
   PRIV_ATTRIB_SCHEMALOC
 };
 
 
 struct VXMLElementInfo {
 public:
   const VXIchar * key;
   int value;
 
   VXMLElementInfo(const VXIchar * k, int v)
     : key(k), value(v)  { }
 
   VXMLElementInfo(const VXMLElementInfo & k)
     : key(k.key), value(k.value)   { }
 
   VXMLElementInfo& operator=(const VXMLElementInfo & k)
     { key = k.key; value = k.value; return *this; }
 };
 
 
 bool operator< (const VXMLElementInfo & x, const VXMLElementInfo & y)
 { if (x.key == NULL || y.key == NULL) return false;
   return wcscmp(x.key, y.key) < 0; }
 
 
 struct VXMLAttribute {
 public:
   const VXIchar * key;
   int value;
 
   VXMLAttribute(const VXIchar * k, int v)
     : key(k), value(v)  { }
 
   VXMLAttribute(const VXMLAttribute & k)
     : key(k.key), value(k.value)       { }
 
   VXMLAttribute& operator=(const VXMLAttribute & k)
     { key = k.key; value = k.value; return *this; }
 };
 
 
 bool operator< (const VXMLAttribute & x, const VXMLAttribute & y)
 { if (x.key == NULL || y.key == NULL) return false;
   return wcscmp(x.key, y.key) < 0; }
 
 
 typedef std::vector<VXMLAttribute>   TABLE_ATTRS;
 typedef std::vector<VXMLElementInfo> TABLE_ELEMS;
 
 TABLE_ATTRS  attrs;
 TABLE_ELEMS  elems;
 
 //#############################################################################
  
 static void InitializeTables()
 {
   // (1) Elements
 
   // (1.1) VXML
   elems.push_back(VXMLElementInfo(L"assign"      , NODE_ASSIGN));
   elems.push_back(VXMLElementInfo(L"audio"       , NODE_AUDIO));
   elems.push_back(VXMLElementInfo(L"block"       , NODE_BLOCK));
   elems.push_back(VXMLElementInfo(L"cancel"      , NODE_CANCEL));
   elems.push_back(VXMLElementInfo(L"catch"       , NODE_CATCH));
   elems.push_back(VXMLElementInfo(L"choice"      , NODE_CHOICE));
   elems.push_back(VXMLElementInfo(L"clear"       , NODE_CLEAR));
   elems.push_back(VXMLElementInfo(L"disconnect"  , NODE_DISCONNECT));
   elems.push_back(VXMLElementInfo(L"else"        , NODE_ELSE));
   elems.push_back(VXMLElementInfo(L"elseif"      , NODE_ELSEIF));
   elems.push_back(VXMLElementInfo(L"enumerate"   , NODE_ENUMERATE));
   elems.push_back(VXMLElementInfo(L"exit"        , NODE_EXIT));
   elems.push_back(VXMLElementInfo(L"field"       , NODE_FIELD));
   elems.push_back(VXMLElementInfo(L"filled"      , NODE_FILLED));
   elems.push_back(VXMLElementInfo(L"form"        , NODE_FORM));
   elems.push_back(VXMLElementInfo(L"goto"        , NODE_GOTO));
   elems.push_back(VXMLElementInfo(L"grammar"     , NODE_GRAMMAR));
   elems.push_back(VXMLElementInfo(L"if"          , NODE_IF));
   elems.push_back(VXMLElementInfo(L"initial"     , NODE_INITIAL));
   elems.push_back(VXMLElementInfo(L"link"        , NODE_LINK));
   elems.push_back(VXMLElementInfo(L"log"         , NODE_LOG));
   elems.push_back(VXMLElementInfo(L"menu"        , NODE_MENU));
   elems.push_back(VXMLElementInfo(L"meta"        , NODE_META));
   elems.push_back(VXMLElementInfo(L"object"      , NODE_OBJECT));
   elems.push_back(VXMLElementInfo(L"option"      , NODE_OPTION));
   elems.push_back(VXMLElementInfo(L"param"       , NODE_PARAM));
   elems.push_back(VXMLElementInfo(L"prompt"      , NODE_PROMPT));
   elems.push_back(VXMLElementInfo(L"property"    , NODE_PROPERTY));
   elems.push_back(VXMLElementInfo(L"record"      , NODE_RECORD));
   elems.push_back(VXMLElementInfo(L"return"      , NODE_RETURN));
   elems.push_back(VXMLElementInfo(L"reprompt"    , NODE_REPROMPT));
   elems.push_back(VXMLElementInfo(L"script"      , NODE_SCRIPT));
   elems.push_back(VXMLElementInfo(L"subdialog"   , NODE_SUBDIALOG));
   elems.push_back(VXMLElementInfo(L"submit"      , NODE_SUBMIT));
   elems.push_back(VXMLElementInfo(L"throw"       , NODE_THROW));
   elems.push_back(VXMLElementInfo(L"transfer"    , NODE_TRANSFER));
   elems.push_back(VXMLElementInfo(L"value"       , NODE_VALUE));
   elems.push_back(VXMLElementInfo(L"var"         , NODE_VAR));
   elems.push_back(VXMLElementInfo(L"vxml"        , NODE_VXML));
 
   // (1.2) from Defaults document
   elems.push_back(VXMLElementInfo(L"defaults"    , DEFAULTS_ROOT));
   elems.push_back(VXMLElementInfo(L"language"    , DEFAULTS_LANGUAGE));
 
   // (1.3) Internals elements (these are converted to others)
   elems.push_back(VXMLElementInfo(L"error"       , PRIV_ELEM_ERROR));
   elems.push_back(VXMLElementInfo(L"help"        , PRIV_ELEM_HELP));
   elems.push_back(VXMLElementInfo(L"noinput"     , PRIV_ELEM_NOINPUT));
   elems.push_back(VXMLElementInfo(L"nomatch"     , PRIV_ELEM_NOMATCH));
   elems.push_back(VXMLElementInfo(L"metadata"    , PRIV_ELEM_METADATA));
 
   // (1.4) SSML
   elems.push_back(VXMLElementInfo(L"break"       , PRIV_ELEM_BREAK));
   elems.push_back(VXMLElementInfo(L"desc"        , PRIV_ELEM_DESC));
   elems.push_back(VXMLElementInfo(L"emphasis"    , PRIV_ELEM_EMPHASIS));
   elems.push_back(VXMLElementInfo(L"mark"        , PRIV_ELEM_MARK));
   elems.push_back(VXMLElementInfo(L"p"           , PRIV_ELEM_PARAGRAPH));
   elems.push_back(VXMLElementInfo(L"paragraph"   , PRIV_ELEM_PARAGRAPH));
   elems.push_back(VXMLElementInfo(L"phoneme"     , PRIV_ELEM_PHONEME));
   elems.push_back(VXMLElementInfo(L"prosody"     , PRIV_ELEM_PROSODY));
   elems.push_back(VXMLElementInfo(L"s"           , PRIV_ELEM_SENTENCE));
   elems.push_back(VXMLElementInfo(L"say-as"      , PRIV_ELEM_SAYAS));
   elems.push_back(VXMLElementInfo(L"sentence"    , PRIV_ELEM_SENTENCE));
   elems.push_back(VXMLElementInfo(L"sub"         , PRIV_ELEM_SUB));
   elems.push_back(VXMLElementInfo(L"voice"       , PRIV_ELEM_VOICE));
 
   // (2) Element attributes
 
   // (2.1) VXML 1.0
   attrs.push_back(VXMLAttribute(L"_itemname"     , ATTRIBUTE__ITEMNAME));
   attrs.push_back(VXMLAttribute(L"aai"           , ATTRIBUTE_AAI));
   attrs.push_back(VXMLAttribute(L"aaiexpr"       , ATTRIBUTE_AAIEXPR));
   attrs.push_back(VXMLAttribute(L"accept"        , ATTRIBUTE_ACCEPT));
   attrs.push_back(VXMLAttribute(L"application"   , ATTRIBUTE_APPLICATION));
   attrs.push_back(VXMLAttribute(L"archive"       , ATTRIBUTE_ARCHIVE));
   attrs.push_back(VXMLAttribute(L"bargein"       , ATTRIBUTE_BARGEIN));
   attrs.push_back(VXMLAttribute(L"bargeintype"   , ATTRIBUTE_BARGEINTYPE));
   attrs.push_back(VXMLAttribute(L"base"          , ATTRIBUTE_BASE));
   attrs.push_back(VXMLAttribute(L"beep"          , ATTRIBUTE_BEEP));
   attrs.push_back(VXMLAttribute(L"bridge"        , ATTRIBUTE_BRIDGE));
   attrs.push_back(VXMLAttribute(L"charset"       , ATTRIBUTE_CHARSET));
   attrs.push_back(VXMLAttribute(L"classid"       , ATTRIBUTE_CLASSID));
   attrs.push_back(VXMLAttribute(L"codebase"      , ATTRIBUTE_CODEBASE));
   attrs.push_back(VXMLAttribute(L"codetype"      , ATTRIBUTE_CODETYPE));
   attrs.push_back(VXMLAttribute(L"cond"          , ATTRIBUTE_COND));
   attrs.push_back(VXMLAttribute(L"connecttimeout", ATTRIBUTE_CONNECTTIME));
   attrs.push_back(VXMLAttribute(L"content"       , ATTRIBUTE_CONTENT));
   attrs.push_back(VXMLAttribute(L"count"         , ATTRIBUTE_COUNT));
   attrs.push_back(VXMLAttribute(L"data"          , ATTRIBUTE_DATA));
   attrs.push_back(VXMLAttribute(L"dest"          , ATTRIBUTE_DEST));
   attrs.push_back(VXMLAttribute(L"destexpr"      , ATTRIBUTE_DESTEXPR));
   attrs.push_back(VXMLAttribute(L"dtmf"          , ATTRIBUTE_DTMF));
   attrs.push_back(VXMLAttribute(L"dtmfterm"      , ATTRIBUTE_DTMFTERM));
   attrs.push_back(VXMLAttribute(L"enctype"       , ATTRIBUTE_ENCTYPE));
   attrs.push_back(VXMLAttribute(L"event"         , ATTRIBUTE_EVENT));
   attrs.push_back(VXMLAttribute(L"eventexpr"     , ATTRIBUTE_EVENTEXPR));
   attrs.push_back(VXMLAttribute(L"expr"          , ATTRIBUTE_EXPR));
   attrs.push_back(VXMLAttribute(L"expritem"      , ATTRIBUTE_EXPRITEM));
   attrs.push_back(VXMLAttribute(L"fetchaudio"    , ATTRIBUTE_FETCHAUDIO));
   attrs.push_back(VXMLAttribute(L"fetchhint"     , ATTRIBUTE_FETCHHINT));
   attrs.push_back(VXMLAttribute(L"fetchtimeout"  , ATTRIBUTE_FETCHTIMEOUT));
   attrs.push_back(VXMLAttribute(L"finalsilence"  , ATTRIBUTE_FINALSILENCE));
   attrs.push_back(VXMLAttribute(L"http-equiv"    , ATTRIBUTE_HTTP_EQUIV));
   attrs.push_back(VXMLAttribute(L"id"            , ATTRIBUTE_ID));