moses-smt/mosesdecoder
1
1
Fork 0
mirror of https://github.com/moses-smt/mosesdecoder.git synced 2024-12-25 12:52:29 +03:00
Code Issues Packages Projects Releases Wiki Activity

remove locking. Make wordIndex variable local

Browse Source
This commit is contained in:
Hieu Hoang 2012-11-30 13:50:59 +00:00
parent 5fd9cbb529
commit 7abb3c878a
2 changed files with 11 additions and 49 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

42
moses/TranslationModel/fuzzy-match/FuzzyMatchWrapper.cpp
View File

@ -49,35 +49,13 @@ namespace tmmt
cerr << "loading completed" << endl;
}
FuzzyMatchWrapper::WordIndex &FuzzyMatchWrapper::GetWordIndex(long translationId)
{
boost::shared_lock<boost::shared_mutex> read_lock(m_accessLock);
std::map<long, WordIndex>::iterator iter = m_wordIndex.find(translationId);
assert(iter != m_wordIndex.end());
return iter->second;
}
void FuzzyMatchWrapper::AddWordIndex(long translationId)
{
boost::unique_lock<boost::shared_mutex> lock(m_accessLock);
WordIndex &ret = m_wordIndex[translationId];
}
void FuzzyMatchWrapper::DeleteWordIndex(long translationId)
{
boost::unique_lock<boost::shared_mutex> lock(m_accessLock);
size_t ret = m_wordIndex.erase(translationId);
CHECK(ret == 1);
}
string FuzzyMatchWrapper::Extract(long translationId, const string &dirNameStr)
{
const Moses::StaticData &staticData = Moses::StaticData::Instance();
AddWordIndex(translationId);
string fuzzyMatchFile = ExtractTM(translationId, dirNameStr);
WordIndex wordIndex;
string fuzzyMatchFile = ExtractTM(wordIndex, translationId, dirNameStr);
// create extrac files
create_xml(fuzzyMatchFile);
@ -104,12 +82,11 @@ namespace tmmt
+ " -phrase-translation-table " + fuzzyMatchFile + ".pt";
system(cmd.c_str());
DeleteWordIndex(translationId);
return fuzzyMatchFile + ".pt.gz";
}
string FuzzyMatchWrapper::ExtractTM(long translationId, const string &dirNameStr)
string FuzzyMatchWrapper::ExtractTM(WordIndex &wordIndex, long translationId, const string &dirNameStr)
{
const std::vector< std::vector< WORD_ID > > &source = suffixArray->GetCorpus();
@ -277,7 +254,7 @@ namespace tmmt
int pruned_match_count = 0;
if (short_match_max_length( input_length ))
{
init_short_matches(translationId, input[sentenceInd] );
init_short_matches(wordIndex, translationId, input[sentenceInd] );
}
vector< int > best_tm;
typedef map< int, vector< Match > >::iterator I;
@ -289,7 +266,7 @@ namespace tmmt
int tmID = tm->first;
int tm_length = suffixArray->GetSentenceLength(tmID);
vector< Match > &match = tm->second;
add_short_matches( translationId, match, source[tmID], input_length, best_cost );
add_short_matches(wordIndex, translationId, match, source[tmID], input_length, best_cost );
//cerr << "match in sentence " << tmID << ": " << match.size() << " [" << tm_length << "]" << endl;
@ -838,13 +815,12 @@ int FuzzyMatchWrapper::short_match_max_length( int input_length )
(to be used by the next function)
(done here, because this has be done only once for an input sentence) */
void FuzzyMatchWrapper::init_short_matches(long translationId, const vector< WORD_ID > &input )
void FuzzyMatchWrapper::init_short_matches(WordIndex &wordIndex, long translationId, const vector< WORD_ID > &input )
{
int max_length = short_match_max_length( input.size() );
if (max_length == 0)
return;
WordIndex &wordIndex = GetWordIndex(translationId);
wordIndex.clear();
// store input words and their positions in hash map
@ -861,14 +837,12 @@ void FuzzyMatchWrapper::init_short_matches(long translationId, const vector< WOR
/* add all short matches to list of matches for a sentence */
void FuzzyMatchWrapper::add_short_matches(long translationId, vector< Match > &match, const vector< WORD_ID > &tm, int input_length, int best_cost )
void FuzzyMatchWrapper::add_short_matches(WordIndex &wordIndex, long translationId, vector< Match > &match, const vector< WORD_ID > &tm, int input_length, int best_cost )
{
int max_length = short_match_max_length( input_length );
if (max_length == 0)
return;
WordIndex &wordIndex = GetWordIndex(translationId);
int tm_length = tm.size();
map< WORD_ID,vector< int > >::iterator input_word_hit;
for(int t_pos=0; t_pos<tm.size(); t_pos++)

18
moses/TranslationModel/fuzzy-match/FuzzyMatchWrapper.h
View File

@ -9,9 +9,6 @@
#ifndef moses_FuzzyMatchWrapper_h
#define moses_FuzzyMatchWrapper_h
#ifdef WITH_THREADS
#include <boost/thread/shared_mutex.hpp>
#endif
#include <fstream>
#include <string>
#include "SuffixArray.h"
@ -46,12 +43,6 @@ protected:
int multiple_max;
typedef std::map< WORD_ID,std::vector< int > > WordIndex;
std::map<long, WordIndex> m_wordIndex;
//WordIndex m_wordIndex;
#ifdef WITH_THREADS
//reader-writer lock
mutable boost::shared_mutex m_accessLock;
#endif
// global cache for word pairs
std::map< std::pair< WORD_ID, WORD_ID >, unsigned int > lsed;
@ -69,21 +60,18 @@ protected:
unsigned int compute_length( const std::vector< tmmt::WORD_ID > &sentence );
unsigned int letter_sed( WORD_ID aIdx, WORD_ID bIdx );
unsigned int sed( const std::vector< WORD_ID > &a, const std::vector< WORD_ID > &b, std::string &best_path, bool use_letter_sed );
void init_short_matches(long translationId, const std::vector< WORD_ID > &input );
void init_short_matches(WordIndex &wordIndex, long translationId, const std::vector< WORD_ID > &input );
int short_match_max_length( int input_length );
void add_short_matches(long translationId, std::vector< Match > &match, const std::vector< WORD_ID > &tm, int input_length, int best_cost );
void add_short_matches(WordIndex &wordIndex, long translationId, std::vector< Match > &match, const std::vector< WORD_ID > &tm, int input_length, int best_cost );
std::vector< Match > prune_matches( const std::vector< Match > &match, int best_cost );
int parse_matches( std::vector< Match > &match, int input_length, int tm_length, int &best_cost );
void create_extract(int sentenceInd, int cost, const std::vector< WORD_ID > &sourceSentence, const std::vector<SentenceAlignment> &targets, const std::string &inputStr, const std::string &path, std::ofstream &outputFile);
std::string ExtractTM(long translationId, const std::string &inputPath);
std::string ExtractTM(WordIndex &wordIndex, long translationId, const std::string &inputPath);
Vocabulary &GetVocabulary()
{ return suffixArray->GetVocabulary(); }
WordIndex &GetWordIndex(long translationId);
void AddWordIndex(long translationId);
void DeleteWordIndex(long translationId);
};
}