void GramError::Report() const
{
#ifdef PROBCHECK
Prob::report(GetSentence(), ruleTerm->GetRule(),
NMarkedSections(), start, stop);
#endif // PROBCHECK
}
/* main - main program
*
* Parameters:
* argc - number of command line arguments
* argv - pointer to command line argument array
*
* Returns:
* none
*/
void
main(int argc, char **argv)
{
DATABASE_S_T data; /* main database */
/* open dictionary then read it in
*/
if ( OpenDictionary( DEFAULT_DICTIONARY, &data.dict ) )
{
fprintf(stderr, "Error: opening %s as dictionary\n", DEFAULT_DICTIONARY);
exit(1);
}
else
{
ReadDictionary( &data.dict );
}
/* get sentence input
*/
if (argc EQ 1)
{
#ifdef DEBUG
DisplayDictionary( &data.dict );
#endif
if ( GetSentence( &data.current_sent ) )
{
fprintf(stderr, "Error: getting sentence\n" );
exit(1);
}
}
else
{
fprintf(stderr, "Error: command line input not supported yet\n" );
exit(1);
}
/* parse the sentence
*/
ParseSentence( &data.dict, &data.current_sent );
#ifdef DEBUG
DisplaySentence( &data.current_sent );
#endif
/* display the sentence diagram
*/
DiagramSentence( &data.current_sent );
exit(0);
}
void CoreferenceDocumentNumeric::ComputeGlobalWordPositions(
CoreferenceDocument* instance) {
sentence_cumulative_lengths_.clear();
sentence_cumulative_lengths_.resize(instance->GetNumSentences());
int offset = 0;
for (int i = 0; i < GetNumSentences(); ++i) {
CoreferenceSentenceNumeric *sentence = GetSentence(i);
// Subtract 1 since there is an extra start symbol.
sentence_cumulative_lengths_[i] = offset + sentence->size() - 1;
offset += sentence->size() - 1;
//LOG(INFO) << "sentence_cumulative_lengths_[" << i << "] = "
// << sentence_cumulative_lengths_[i];
}
}
// Given a partial rule application ending at startPos-1 and given the sets of
// source and target non-terminals covering the span [startPos, endPos],
// determines the full or partial rule applications that can be produced through
// extending the current rule application by a single non-terminal.
void ChartRuleLookupManagerMemory::ExtendPartialRuleApplication(
const DottedRuleInMemory &prevDottedRule,
size_t startPos,
size_t endPos,
size_t stackInd,
DottedRuleColl & dottedRuleColl)
{
// source non-terminal labels for the remainder
const NonTerminalSet &sourceNonTerms =
GetSentence().GetLabelSet(startPos, endPos);
// target non-terminal labels for the remainder
const ChartCellLabelSet &targetNonTerms = GetTargetLabelSet(startPos, endPos);
// note where it was found in the prefix tree of the rule dictionary
const PhraseDictionaryNodeSCFG &node = prevDottedRule.GetLastNode();
const PhraseDictionaryNodeSCFG::NonTerminalMap & nonTermMap =
node.GetNonTerminalMap();
const size_t numChildren = nonTermMap.size();
if (numChildren == 0) {
return;
}
const size_t numSourceNonTerms = sourceNonTerms.size();
const size_t numTargetNonTerms = targetNonTerms.GetSize();
const size_t numCombinations = numSourceNonTerms * numTargetNonTerms;
// We can search by either:
// 1. Enumerating all possible source-target NT pairs that are valid for
// the span and then searching for matching children in the node,
// or
// 2. Iterating over all the NT children in the node, searching
// for each source and target NT in the span's sets.
// We'll do whichever minimises the number of lookups:
if (numCombinations <= numChildren*2) {
// loop over possible source non-terminal labels (as found in input tree)
NonTerminalSet::const_iterator p = sourceNonTerms.begin();
NonTerminalSet::const_iterator sEnd = sourceNonTerms.end();
for (; p != sEnd; ++p) {
const Word & sourceNonTerm = *p;
// loop over possible target non-terminal labels (as found in chart)
ChartCellLabelSet::const_iterator q = targetNonTerms.begin();
ChartCellLabelSet::const_iterator tEnd = targetNonTerms.end();
for (; q != tEnd; ++q) {
const ChartCellLabel &cellLabel = q->second;
// try to match both source and target non-terminal
const PhraseDictionaryNodeSCFG * child =
node.GetChild(sourceNonTerm, cellLabel.GetLabel());
// nothing found? then we are done
if (child == NULL) {
continue;
}
// create new rule
#ifdef USE_BOOST_POOL
DottedRuleInMemory *rule = m_dottedRulePool.malloc();
new (rule) DottedRuleInMemory(*child, cellLabel, prevDottedRule);
#else
DottedRuleInMemory *rule = new DottedRuleInMemory(*child, cellLabel,
prevDottedRule);
#endif
dottedRuleColl.Add(stackInd, rule);
}
}
}
else
{
// loop over possible expansions of the rule
PhraseDictionaryNodeSCFG::NonTerminalMap::const_iterator p;
PhraseDictionaryNodeSCFG::NonTerminalMap::const_iterator end =
nonTermMap.end();
for (p = nonTermMap.begin(); p != end; ++p) {
// does it match possible source and target non-terminals?
const PhraseDictionaryNodeSCFG::NonTerminalMapKey &key = p->first;
const Word &sourceNonTerm = key.first;
if (sourceNonTerms.find(sourceNonTerm) == sourceNonTerms.end()) {
continue;
}
const Word &targetNonTerm = key.second;
const ChartCellLabel *cellLabel = targetNonTerms.Find(targetNonTerm);
if (!cellLabel) {
continue;
}
// create new rule
const PhraseDictionaryNodeSCFG &child = p->second;
#ifdef USE_BOOST_POOL
DottedRuleInMemory *rule = m_dottedRulePool.malloc();
new (rule) DottedRuleInMemory(child, *cellLabel, prevDottedRule);
#else
DottedRuleInMemory *rule = new DottedRuleInMemory(child, *cellLabel,
prevDottedRule);
#endif
dottedRuleColl.Add(stackInd, rule);
}
}
}
bool Satori::Save(bool isOnUnload) {
GetSender().next_event();
// メンバ変数を里々変数化
for (std::map<int, string>::iterator it=reserved_talk.begin(); it!=reserved_talk.end() ; ++it)
variables[string("次から")+itos(it->first)+"回目のトーク"] = it->second;
// 起動時間累計を設定
variables["ゴースト起動時間累計秒"] =
uitos(posix_get_current_sec() - sec_count_at_load + sec_count_total,"%lu");
// (互換用)
variables["ゴースト起動時間累計ミリ秒"] =
uitos((posix_get_current_sec() - sec_count_at_load + sec_count_total)*1000,"%lu");
variables["ゴースト起動時間累計(ms)"] =
uitos((posix_get_current_sec() - sec_count_at_load + sec_count_total)*1000,"%lu");
if ( isOnUnload ) {
secure_flag = true;
(void)GetSentence("OnSatoriUnload");
}
string theFullPath = mBaseFolder + "satori_savedata.tmp";
std::ofstream out(theFullPath.c_str());
bool temp = GetSender().is_validated();
GetSender().validate();
GetSender().sender() << "saving " << theFullPath << "... " ;
GetSender().validate(temp);
if ( !out.is_open() )
{
GetSender().sender() << "failed." << std::endl;
return false;
}
string line = "＊セーブデータ";
string data;
out << ENCODE(line) << std::endl;
for (strmap::const_iterator it=variables.begin() ; it!=variables.end() ; ++it) {
string str = zen2han(it->first);
if ( str[0]=='S' && aredigits(str.c_str()+1) ) {
continue;
}
if ( str == "今回は喋らない" || str == "今回は会話時サーフェス戻し" || str == "今回は会話時サーフィス戻し" || str == "今回は自動アンカー" ) {
continue;
}
data = it->second;
replace(data,"φ","φφ");
replace(data,"（","φ（");
replace(data,"）","φ）");
m_escaper.unescape_for_dic(data);
string line = string("＄")+it->first+"\t"+data; // 変数を保存
out << ENCODE(line) << std::endl;
}
for (std::map<string, std::vector<Word> >::const_iterator i=mAppendedWords.begin() ; i!=mAppendedWords.end() ; ++i )
{
if ( ! i->second.empty() ) {
out << std::endl << ENCODE( string("＠") + i->first ) << std::endl;
for (std::vector<Word>::const_iterator j=i->second.begin() ; j!=i->second.end() ; ++j )
{
out << ENCODE( *j ) << std::endl;
}
}
}
out.flush();
out.close();
GetSender().sender() << "ok." << std::endl;
//バックアップ
string realFullPath = mBaseFolder + "satori_savedata." + (fEncodeSavedata?"sat":"txt");
string realFullPathBackup = mBaseFolder + "satori_savebackup." + (fEncodeSavedata?"sat":"txt");
#ifdef POSIX
unlink(realFullPathBackup.c_str());
rename(realFullPath.c_str(),realFullPathBackup.c_str());
rename(theFullPath.c_str(),realFullPath.c_str());
#else
::DeleteFile(realFullPathBackup.c_str());
::MoveFile(realFullPath.c_str(),realFullPathBackup.c_str());
::MoveFile(theFullPath.c_str(),realFullPath.c_str());
#endif
//いらないほうを消す
string delFullPath = mBaseFolder + "satori_savedata." + (fEncodeSavedata?"txt":"sat");
string delFullPathBackup = mBaseFolder + "satori_savebackup." + (fEncodeSavedata?"txt":"sat");
#ifdef POSIX
unlink(delFullPath.c_str());
unlink(delFullPathBackup.c_str());
#else
::DeleteFile(delFullPath.c_str());
::DeleteFile(delFullPathBackup.c_str());
#endif
return true;
}
bool Satori::load(const string& iBaseFolder)
{
GetSender().next_event();
setlocale(LC_ALL, "Japanese");
#ifdef _WINDOWS
_setmbcp(_MB_CP_LOCALE);
#endif
mBaseFolder = iBaseFolder;
GetSender().sender() << "■SATORI::Load on " << mBaseFolder << "" << std::endl;
#if POSIX
// 「/」で終わっていなければ付ける。
if (mBaseFolder[mBaseFolder.size() - 1] != '/') {
mBaseFolder += '/';
}
#else
// 「\」で終わっていなければ付ける。
if (mBaseFolder[mBaseFolder.size() - 1] != '\\') {
mBaseFolder += '\\';
}
#endif
#ifdef _MSC_VER
// 本体のあるフォルダをサーチ
{
TCHAR buf[MAX_PATH+1];
::GetModuleFileName(NULL, buf, MAX_PATH);
char* p = FindFinalChar(buf, DIR_CHAR);
if ( p==NULL )
mExeFolder = "";
else {
*(++p) = '\0';
mExeFolder = buf;
}
}
GetSender().sender() << "本体の所在: " << mExeFolder << "" << std::endl;
#endif // _MSC_VER
// メンバ初期化
InitMembers();
#ifdef _MSC_VER
// システムの設定を読んでおく
OSVERSIONINFO ovi;
ovi.dwOSVersionInfoSize = sizeof(OSVERSIONINFO);
::GetVersionEx(&ovi);
string os;
if ( ovi.dwPlatformId == VER_PLATFORM_WIN32_WINDOWS ) {
if ( ovi.dwMinorVersion == 0 ) { mOSType=SATORI_OS_WIN95; os="Windows 95"; }
else if ( ovi.dwMinorVersion == 10 ) { mOSType=SATORI_OS_WIN98; os="Windows 98"; }
else if ( ovi.dwMinorVersion == 90 ) { mOSType=SATORI_OS_WINME; os="Windows Me"; }
else { mOSType = SATORI_OS_UNDEFINED; os="undefined"; }
} else {
if ( ovi.dwMinorVersion == 0 ) {
if ( ovi.dwMajorVersion == 4 ) { mOSType=SATORI_OS_WINNT; os="Windows NT"; }
else if ( ovi.dwMajorVersion == 5 ) { mOSType=SATORI_OS_WIN2K; os="Windows 2000"; }
}
else { mOSType = SATORI_OS_WINXP; os="Windows XP or later"; }
}
GetSender().sender() << "ＯＳ種別: " << os << std::endl;
if ( mOSType==SATORI_OS_WIN95 ) {
is_single_monitor = true;
} else {
BOOL (WINAPI* pEnumDisplayMonitors)(HDC,LPRECT,MONITORENUMPROC,LPARAM);
(FARPROC&)pEnumDisplayMonitors = ::GetProcAddress(::GetModuleHandle("user32.dll"), "EnumDisplayMonitors");
if ( pEnumDisplayMonitors==NULL ) {
is_single_monitor = true;
}
else {
RECT rectData[2];
memset(rectData,0,sizeof(rectData));
(*pEnumDisplayMonitors)(NULL,NULL,(MONITORENUMPROC)MonitorEnumFunc,(LPARAM)(rectData));
max_screen_rect = rectData[0];
desktop_rect = rectData[1];
RECT* rect;
rect = &desktop_rect;
GetSender().sender() << "プライマリデスクトップ: (" <<
rect->left << "," << rect->top << "," << rect->right << "," << rect->bottom << ")" << std::endl;
rect = &max_screen_rect;
GetSender().sender() << "仮想デスクトップ: (" <<
rect->left << "," << rect->top << "," << rect->right << "," << rect->bottom << ")" << std::endl;
is_single_monitor = ( ::EqualRect(&max_screen_rect, &desktop_rect)!=FALSE );
GetSender().sender() << (is_single_monitor ?
"モニタは一つだけと判断、見切れ判定を呼び出し元に任せます。" :
"複数のモニタが接続されていると判断、見切れ判定は里々が行います。") << std::endl;
}
}
#endif // _MSC_VER
// 置換辞書読み取り
strmap_from_file(replace_before_dic, mBaseFolder+"replace.txt", "\t");
strmap_from_file(replace_after_dic, mBaseFolder+"replace_after.txt", "\t");
// キャラデータ読み込み
mCharacters.load(mBaseFolder + "characters.ini");
for ( inimap::const_iterator i=mCharacters.begin() ; i!=mCharacters.end() ; ++i ) {
const strmap& m = i->second;
strmap::const_iterator j;
// 置換辞書に追加
j = m.find("popular-name");
if ( j != m.end() && j->second.size()>0 )
replace_before_dic[j->second + "："] = string("\xff\x01") + zen2han(i->first); //0xff0x01はあとで変換
j = m.find("initial-letter");
if ( j != m.end() && j->second.size()>0 )
replace_before_dic[j->second + "："] = string("\xff\x01") + zen2han(i->first); //0xff0x01はあとで変換
j = m.find("base-surface");
if ( j != m.end() && j->second.size()>0 )
system_variable_operation( string("サーフェス加算値") + i->first, j->second);
}
//for ( strmap::const_iterator j=replace_before_dic.begin() ; j!=replace_before_dic.end() ; ++j )
// cout << j->first << ": " << j->second << endl;
// ランダマイズ
randomize();
//------------------------------------------
// コンフィグ読み込み
LoadDictionary(mBaseFolder + "satori_conf.txt", false);
// 変数初期化実行
GetSentence("初期化");
// SAORI読み込み
Family<Word>* f = words.get_family("SAORI");
mShioriPlugins->load(mBaseFolder);
if ( f != NULL )
{
std::vector<const Word*> els;
f->get_elements_pointers(els);
for (std::vector<const Word*>::const_iterator i=els.begin(); i!=els.end() ; ++i)
{
if ( (*i)->size()>0 && !mShioriPlugins->load_a_plugin(**i) )
{
GetSender().sender() << "SAORI読み込み中にエラーが発生: " << **i << std::endl;
}
}
}
mShioriPlugins->load_default_entry();
talks.clear();
words.clear();
//------------------------------------------
// セーブデータ読み込み
//bool oldConf = fEncodeSavedata;
bool loadResult = LoadDictionary(mBaseFolder + "satori_savedata." + (fEncodeSavedata?"sat":"txt"), false);
GetSentence("セーブデータ");
bool execResult = talks.get_family("セーブデータ") != NULL;
if ( ! loadResult || ! execResult ) {
loadResult = LoadDictionary(mBaseFolder + "satori_savebackup." + (fEncodeSavedata?"sat":"txt"), false);
GetSentence("セーブデータ");
execResult = talks.get_family("セーブデータ") != NULL;
}
talks.clear();
reload_flag = false;
if ( variables.find("ゴースト起動時間累計秒") != variables.end() ) {
sec_count_total = zen2ul(variables["ゴースト起動時間累計秒"]);
}
else if ( variables.find("ゴースト起動時間累計ミリ秒") != variables.end() ) {
sec_count_total = zen2ul(variables["ゴースト起動時間累計ミリ秒"]) / 1000;
}
else {
sec_count_total = zen2ul(variables["ゴースト起動時間累計(ms)"]) / 1000;
}
variables["起動回数"] = itos( zen2int(variables["起動回数"])+1 );
// 「単語の追加」で登録された単語を覚えておく
const std::map< string, Family<Word> >& m = words.compatible();
for (std::map< string, Family<Word> >::const_iterator it = m.begin() ; it != m.end() ; ++it )
{
std::vector<const Word*> v;
it->second.get_elements_pointers(v);
for (std::vector<const Word*>::const_iterator itx = v.begin() ; itx < v.end() ; ++itx ) {
mAppendedWords[it->first].push_back(**itx);
}
}
//------------------------------------------
// 指定フォルダの辞書を読み込み
int loadcount = 0;
strvec::iterator i = dic_folder.begin();
if ( i==dic_folder.end() ) {
loadcount += LoadDicFolder(mBaseFolder); // ルートフォルダの辞書
} else {
for ( ; i!=dic_folder.end() ; ++i )
loadcount += LoadDicFolder(mBaseFolder + *i + DIR_CHAR); // サブフォルダの辞書
}
is_dic_loaded = loadcount != 0;
//------------------------------------------
secure_flag = true;
system_variable_operation("単語群「＊」の重複回避", "有効、トーク中");
system_variable_operation("文「＊」の重複回避", "有効");
//system_variable_operation("単語群「季節の食べ物」の重複回避", "有効、トーク中");
GetSentence("OnSatoriLoad");
on_loaded_script = GetSentence("OnSatoriBoot");
diet_script(on_loaded_script);
GetSender().sender() << "loaded." << std::endl;
GetSender().flush();
return true;
}
void ChartRuleLookupManagerMemory::GetChartRuleCollection(
const WordsRange &range,
bool adhereTableLimit,
ChartTranslationOptionList &outColl)
{
size_t relEndPos = range.GetEndPos() - range.GetStartPos();
size_t absEndPos = range.GetEndPos();
// MAIN LOOP. create list of nodes of target phrases
ProcessedRuleColl &processedRuleCol = *m_processedRuleColls[range.GetStartPos()];
const ProcessedRuleList &runningNodes = processedRuleCol.GetRunningNodes();
// Note that runningNodes can be expanded as the loop runs (through calls to
// ExtendPartialRuleApplication()).
for (size_t ind = 0; ind < runningNodes.size(); ++ind)
{
const ProcessedRule &prevProcessedRule = *runningNodes[ind];
const PhraseDictionaryNodeSCFG &prevNode = prevProcessedRule.GetLastNode();
const WordConsumed *prevWordConsumed = prevProcessedRule.GetLastWordConsumed();
size_t startPos = (prevWordConsumed == NULL) ? range.GetStartPos() : prevWordConsumed->GetWordsRange().GetEndPos() + 1;
// search for terminal symbol
if (startPos == absEndPos)
{
const Word &sourceWord = GetSentence().GetWord(absEndPos);
const PhraseDictionaryNodeSCFG *node = prevNode.GetChild(sourceWord);
if (node != NULL)
{
WordConsumed *newWordConsumed = new WordConsumed(absEndPos, absEndPos
, sourceWord
, prevWordConsumed);
ProcessedRule *processedRule = new ProcessedRule(*node, newWordConsumed);
processedRuleCol.Add(relEndPos+1, processedRule);
}
}
// search for non-terminals
size_t endPos, stackInd;
if (startPos > absEndPos)
continue;
else if (startPos == range.GetStartPos() && range.GetEndPos() > range.GetStartPos())
{ // start.
endPos = absEndPos - 1;
stackInd = relEndPos;
}
else
{
endPos = absEndPos;
stackInd = relEndPos + 1;
}
const NonTerminalSet &sourceNonTerms =
GetSentence().GetLabelSet(startPos, endPos);
const NonTerminalSet &targetNonTerms =
GetCellCollection().GetHeadwords(WordsRange(startPos, endPos));
ExtendPartialRuleApplication(prevNode, prevWordConsumed, startPos,
endPos, stackInd, sourceNonTerms,
targetNonTerms, processedRuleCol);
}
// return list of target phrases
ProcessedRuleList &nodes = processedRuleCol.Get(relEndPos + 1);
size_t rulesLimit = StaticData::Instance().GetRuleLimit();
ProcessedRuleList::const_iterator iterNode;
for (iterNode = nodes.begin(); iterNode != nodes.end(); ++iterNode)
{
const ProcessedRule &processedRule = **iterNode;
const PhraseDictionaryNodeSCFG &node = processedRule.GetLastNode();
const WordConsumed *wordConsumed = processedRule.GetLastWordConsumed();
assert(wordConsumed);
const TargetPhraseCollection *targetPhraseCollection = node.GetTargetPhraseCollection();
if (targetPhraseCollection != NULL)
{
outColl.Add(*targetPhraseCollection, *wordConsumed, adhereTableLimit, rulesLimit);
}
}
outColl.CreateChartRules(rulesLimit);
}
void ChartRuleLookupManagerOnDisk::GetChartRuleCollection(
const WordsRange &range,
ChartParserCallback &outColl)
{
const StaticData &staticData = StaticData::Instance();
size_t relEndPos = range.GetEndPos() - range.GetStartPos();
size_t absEndPos = range.GetEndPos();
// MAIN LOOP. create list of nodes of target phrases
DottedRuleStackOnDisk &expandableDottedRuleList = *m_expandableDottedRuleListVec[range.GetStartPos()];
// sort save nodes so only do nodes with most counts
expandableDottedRuleList.SortSavedNodes();
const DottedRuleStackOnDisk::SavedNodeColl &savedNodeColl = expandableDottedRuleList.GetSavedNodeColl();
//cerr << "savedNodeColl=" << savedNodeColl.size() << " ";
const ChartCellLabel &sourceWordLabel = GetSourceAt(absEndPos);
for (size_t ind = 0; ind < (savedNodeColl.size()) ; ++ind) {
const SavedNodeOnDisk &savedNode = *savedNodeColl[ind];
const DottedRuleOnDisk &prevDottedRule = savedNode.GetDottedRule();
const OnDiskPt::PhraseNode &prevNode = prevDottedRule.GetLastNode();
size_t startPos = prevDottedRule.IsRoot() ? range.GetStartPos() : prevDottedRule.GetWordsRange().GetEndPos() + 1;
// search for terminal symbol
if (startPos == absEndPos) {
OnDiskPt::Word *sourceWordBerkeleyDb = m_dbWrapper.ConvertFromMoses(Input, m_inputFactorsVec, sourceWordLabel.GetLabel());
if (sourceWordBerkeleyDb != NULL) {
const OnDiskPt::PhraseNode *node = prevNode.GetChild(*sourceWordBerkeleyDb, m_dbWrapper);
if (node != NULL) {
// TODO figure out why source word is needed from node, not from sentence
// prob to do with factors or non-term
//const Word &sourceWord = node->GetSourceWord();
DottedRuleOnDisk *dottedRule = new DottedRuleOnDisk(*node, sourceWordLabel, prevDottedRule);
expandableDottedRuleList.Add(relEndPos+1, dottedRule);
// cache for cleanup
m_sourcePhraseNode.push_back(node);
}
delete sourceWordBerkeleyDb;
}
}
// search for non-terminals
size_t endPos, stackInd;
if (startPos > absEndPos)
continue;
else if (startPos == range.GetStartPos() && range.GetEndPos() > range.GetStartPos()) {
// start.
endPos = absEndPos - 1;
stackInd = relEndPos;
} else {
endPos = absEndPos;
stackInd = relEndPos + 1;
}
// size_t nonTermNumWordsCovered = endPos - startPos + 1;
// get target nonterminals in this span from chart
const ChartCellLabelSet &chartNonTermSet =
GetTargetLabelSet(startPos, endPos);
//const Word &defaultSourceNonTerm = staticData.GetInputDefaultNonTerminal()
// ,&defaultTargetNonTerm = staticData.GetOutputDefaultNonTerminal();
// go through each SOURCE lhs
const NonTerminalSet &sourceLHSSet = GetSentence().GetLabelSet(startPos, endPos);
NonTerminalSet::const_iterator iterSourceLHS;
for (iterSourceLHS = sourceLHSSet.begin(); iterSourceLHS != sourceLHSSet.end(); ++iterSourceLHS) {
const Word &sourceLHS = *iterSourceLHS;
OnDiskPt::Word *sourceLHSBerkeleyDb = m_dbWrapper.ConvertFromMoses(Input, m_inputFactorsVec, sourceLHS);
if (sourceLHSBerkeleyDb == NULL) {
delete sourceLHSBerkeleyDb;
continue; // vocab not in pt. node definately won't be in there
}
const OnDiskPt::PhraseNode *sourceNode = prevNode.GetChild(*sourceLHSBerkeleyDb, m_dbWrapper);
delete sourceLHSBerkeleyDb;
if (sourceNode == NULL)
continue; // didn't find source node
// go through each TARGET lhs
ChartCellLabelSet::const_iterator iterChartNonTerm;
for (iterChartNonTerm = chartNonTermSet.begin(); iterChartNonTerm != chartNonTermSet.end(); ++iterChartNonTerm) {
const ChartCellLabel &cellLabel = iterChartNonTerm->second;
//cerr << sourceLHS << " " << defaultSourceNonTerm << " " << chartNonTerm << " " << defaultTargetNonTerm << endl;
//bool isSyntaxNonTerm = (sourceLHS != defaultSourceNonTerm) || (chartNonTerm != defaultTargetNonTerm);
bool doSearch = true; //isSyntaxNonTerm ? nonTermNumWordsCovered <= maxSyntaxSpan :
// nonTermNumWordsCovered <= maxDefaultSpan;
if (doSearch) {
OnDiskPt::Word *chartNonTermBerkeleyDb = m_dbWrapper.ConvertFromMoses(Output, m_outputFactorsVec, cellLabel.GetLabel());
if (chartNonTermBerkeleyDb == NULL)
continue;
const OnDiskPt::PhraseNode *node = sourceNode->GetChild(*chartNonTermBerkeleyDb, m_dbWrapper);
delete chartNonTermBerkeleyDb;
if (node == NULL)
continue;
// found matching entry
//const Word &sourceWord = node->GetSourceWord();
DottedRuleOnDisk *dottedRule = new DottedRuleOnDisk(*node, cellLabel, prevDottedRule);
expandableDottedRuleList.Add(stackInd, dottedRule);
m_sourcePhraseNode.push_back(node);
}
} // for (iterChartNonTerm
delete sourceNode;
} // for (iterLabelListf
// return list of target phrases
DottedRuleCollOnDisk &nodes = expandableDottedRuleList.Get(relEndPos + 1);
// source LHS
DottedRuleCollOnDisk::const_iterator iterDottedRuleColl;
for (iterDottedRuleColl = nodes.begin(); iterDottedRuleColl != nodes.end(); ++iterDottedRuleColl) {
// node of last source word
const DottedRuleOnDisk &prevDottedRule = **iterDottedRuleColl;
if (prevDottedRule.Done())
continue;
prevDottedRule.Done(true);
const OnDiskPt::PhraseNode &prevNode = prevDottedRule.GetLastNode();
//get node for each source LHS
const NonTerminalSet &lhsSet = GetSentence().GetLabelSet(range.GetStartPos(), range.GetEndPos());
NonTerminalSet::const_iterator iterLabelSet;
for (iterLabelSet = lhsSet.begin(); iterLabelSet != lhsSet.end(); ++iterLabelSet) {
const Word &sourceLHS = *iterLabelSet;
OnDiskPt::Word *sourceLHSBerkeleyDb = m_dbWrapper.ConvertFromMoses(Input, m_inputFactorsVec, sourceLHS);
if (sourceLHSBerkeleyDb == NULL)
continue;
const TargetPhraseCollection *targetPhraseCollection = NULL;
const OnDiskPt::PhraseNode *node = prevNode.GetChild(*sourceLHSBerkeleyDb, m_dbWrapper);
if (node) {
UINT64 tpCollFilePos = node->GetValue();
std::map<UINT64, const TargetPhraseCollection*>::const_iterator iterCache = m_cache.find(tpCollFilePos);
if (iterCache == m_cache.end()) {
const OnDiskPt::TargetPhraseCollection *tpcollBerkeleyDb = node->GetTargetPhraseCollection(m_dictionary.GetTableLimit(), m_dbWrapper);
std::vector<float> weightT = staticData.GetWeights(&m_dictionary);
targetPhraseCollection
= tpcollBerkeleyDb->ConvertToMoses(m_inputFactorsVec
,m_outputFactorsVec
,m_dictionary
,weightT
,m_filePath
, m_dbWrapper.GetVocab());
delete tpcollBerkeleyDb;
m_cache[tpCollFilePos] = targetPhraseCollection;
} else {
// just get out of cache
targetPhraseCollection = iterCache->second;
}
CHECK(targetPhraseCollection);
if (!targetPhraseCollection->IsEmpty()) {
AddCompletedRule(prevDottedRule, *targetPhraseCollection,
range, outColl);
}
} // if (node)
delete node;
delete sourceLHSBerkeleyDb;
}
}
} // for (size_t ind = 0; ind < savedNodeColl.size(); ++ind)
//cerr << numDerivations << " ";
}
