X-Git-Url: http://git.nguyen.vg/gitweb/?a=blobdiff_plain;f=TCImplementation.cpp;h=1a29f2524eccf74a394c02aed5b62bf4c54bb252;hb=102e33b134075765e6d4e0c38bc1307568ce5602;hp=290ca9c15864bfadf1ad304ebf83733308d79983;hpb=40ddf9aca842bdc081b6350a4ebfe36b066c94c9;p=SXSI%2FTextCollection.git diff --git a/TCImplementation.cpp b/TCImplementation.cpp index 290ca9c..1a29f25 100644 --- a/TCImplementation.cpp +++ b/TCImplementation.cpp @@ -19,7 +19,10 @@ *****************************************************************************/ #include "TCImplementation.h" -//#include "HeapProfiler.h" // FIXME remove +//#define DEBUG_MEMUSAGE +#ifdef DEBUG_MEMUSAGE +#include "HeapProfiler.h" // FIXME remove +#endif #include #include @@ -33,63 +36,42 @@ using std::vector; using std::pair; using std::make_pair; using std::map; - +using std::string; namespace SXSI { // Save file version info -const uchar TCImplementation::versionFlag = 3; +const uchar TCImplementation::versionFlag = 8; /** * Constructor inits an empty dynamic FM-index. * Samplerate defaults to TEXTCOLLECTION_DEFAULT_SAMPLERATE. */ -TCImplementation::TCImplementation(uchar * bwt, ulong length, unsigned samplerate_, unsigned numberOfTexts_, ulong maxTextLength_) +TCImplementation::TCImplementation(uchar * bwt, ulong length, unsigned samplerate_, + unsigned numberOfTexts_, ulong maxTextLength_, ulong numberOfSamples_, + CSA::DeltaVector & notIndexed, const string & niText, char tsType) : n(length), samplerate(samplerate_), alphabetrank(0), sampled(0), suffixes(0), - suffixDocId(0), numberOfTexts(numberOfTexts_), numberOfAllTexts(numberOfTexts_), maxTextLength(maxTextLength_), endmarkerDocId(0), - emptyTextRank(0) + suffixDocId(0), numberOfTexts(numberOfTexts_), maxTextLength(maxTextLength_), Doc(0) { - - // Bitvector of empty texts, FIXME Remove? - { - //std::cout << std::endl << "texts: " << numberOfTexts << ", all texts " << numberOfAllTexts << ", size : " << emptyTextId.size() <::const_iterator it = emptyTextId.begin(); it != emptyTextId.end(); ++it) -// Tools::SetField(tempB, 1, (*it), 1); - emptyTextRank = new BSGAP(tempB, numberOfTexts, true); - } makewavelet(bwt); // Deletes bwt! bwt = 0; - + // Make sampling tables - maketables(); + maketables(numberOfSamples_, tsType, notIndexed, niText); } bool TCImplementation::EmptyText(DocId k) const { assert(k < (DocId)numberOfTexts); - if (emptyTextRank->IsBitSet(k)) - return true; - return false; + return false; // Empty texts are not indexed } -uchar* TCImplementation::GetText(DocId k) const +uchar * TCImplementation::GetText(DocId k) const { assert(k < (DocId)numberOfTexts); - ulong textRank = 0; - if (emptyTextRank->IsBitSet(k, &textRank)) - { - uchar* result = new uchar[1]; - result[0] = 0; - return result; - } - // Map to non-empty text - k -= textRank; //emptyTextRank->rank(k); - - TextPosition i = k; + return textStorage->GetText(k); +/* TextPosition i = k; string result; // Reserve average string length to avoid reallocs @@ -110,10 +92,11 @@ uchar* TCImplementation::GetText(DocId k) const res[i] = '\0'; for (ulong j = 0; j < i; ++j) res[i-j-1] = result[j]; - return res; + return res;*/ } + /* - * Not supported + * Substring queries are supported via the pointer returned by TextStorage::GetText uchar* TCImplementation::GetText(DocId k, TextPosition i, TextPosition j) const { assert(k < (DocId)numberOfTexts); @@ -121,15 +104,6 @@ uchar* TCImplementation::GetText(DocId k, TextPosition i, TextPosition j) const assert(i <= j); ulong textRank = 0; - if (emptyTextRank->IsBitSet(k, &textRank)) - { - uchar* result = new uchar[1]; - result[0] = 0; - return result; - } - - // Map to non-empty text - k -= textRank; // emptyTextRank->rank(k); // Start position of k'th text ulong start = (*textStartPos)[k]; @@ -137,6 +111,8 @@ uchar* TCImplementation::GetText(DocId k, TextPosition i, TextPosition j) const return Substring(i + start, j-i+1); }*/ + + /****************************************************************** * Existential queries */ @@ -191,11 +167,7 @@ bool TCImplementation::IsEqual(uchar const *pattern) const { TextPosition m = std::strlen((char *)pattern); if (m == 0) - { - if (numberOfAllTexts - numberOfTexts > 0u) - return true; // Empty texts exists return false; // No empty texts exists - } TextPosition sp = 0, ep = 0; // Match including end-marker @@ -211,11 +183,7 @@ bool TCImplementation::IsEqual(uchar const *pattern, DocId begin, DocId end) con { TextPosition m = std::strlen((char *)pattern); if (m == 0) - { - if (numberOfAllTexts - numberOfTexts > 0u) - return true; // Empty texts exists return false; // No empty texts exists - } TextPosition sp = 0, ep = 0; // Match including end-marker @@ -282,7 +250,7 @@ unsigned TCImplementation::CountPrefix(uchar const * pattern) const { TextPosition m = strlen((char *)pattern); if (m == 0) - return numberOfAllTexts; + return numberOfTexts; TextPosition sp = 0, ep = 0; Search(pattern, m, &sp, &ep); @@ -295,7 +263,7 @@ unsigned TCImplementation::CountPrefix(uchar const * pattern, DocId begin, DocId { TextPosition m = strlen((char *)pattern); if (m == 0) - return numberOfAllTexts; + return numberOfTexts; TextPosition sp = 0, ep = 0; Search(pattern, m, &sp, &ep); @@ -308,7 +276,7 @@ unsigned TCImplementation::CountSuffix(uchar const * pattern) const { TextPosition m = strlen((char *)pattern); if (m == 0) - return numberOfAllTexts; + return numberOfTexts; TextPosition sp = 0, ep = 0; // Search with end-marker @@ -321,7 +289,7 @@ unsigned TCImplementation::CountSuffix(uchar const * pattern, DocId begin, DocId { TextPosition m = strlen((char *)pattern); if (m == 0) - return numberOfAllTexts; + return numberOfTexts; TextPosition sp = 0, ep = 0; // Search with end-marker @@ -334,7 +302,7 @@ unsigned TCImplementation::CountEqual(uchar const *pattern) const { TextPosition m = strlen((char const *)pattern); if (m == 0) - return numberOfAllTexts - numberOfTexts; // Empty texts. + return 0; // No empty texts. TextPosition sp = 0, ep = 0; // Match including end-marker @@ -348,7 +316,7 @@ unsigned TCImplementation::CountEqual(uchar const *pattern, DocId begin, DocId e { TextPosition m = strlen((char const *)pattern); if (m == 0) - return numberOfAllTexts - numberOfTexts; // Empty texts. + return 0; // No empty texts. TextPosition sp = 0, ep = 0; // Match including end-marker @@ -362,7 +330,7 @@ unsigned TCImplementation::CountContains(uchar const * pattern) const { TextPosition m = strlen((char const *)pattern); if (m == 0) - return numberOfAllTexts; // Total number of texts. + return numberOfTexts; // Total number of texts. // Here counting is as slow as fetching the result set // because we have to filter out occ's that fall within same document. @@ -374,7 +342,7 @@ unsigned TCImplementation::CountContains(uchar const * pattern, DocId begin, Doc { TextPosition m = strlen((char const *)pattern); if (m == 0) - return numberOfAllTexts; // Total number of texts. + return numberOfTexts; // Total number of texts. // Here counting is as slow as fetching the result set // because we have to filter out occ's that fall within same document. @@ -387,7 +355,7 @@ unsigned TCImplementation::CountLessThan(uchar const * pattern) const { TextPosition m = strlen((char const *)pattern); if (m == 0) - return numberOfAllTexts - numberOfTexts; // Empty texts. + return 0; // No empty texts. TextPosition sp = 0, ep = 0; SearchLessThan(pattern, m, &sp, &ep); @@ -400,7 +368,7 @@ unsigned TCImplementation::CountLessThan(uchar const * pattern, DocId begin, Doc { TextPosition m = strlen((char const *)pattern); if (m == 0) - return numberOfAllTexts - numberOfTexts; // Empty texts. + return 0; // No empty texts. TextPosition sp = 0, ep = 0; SearchLessThan(pattern, m, &sp, &ep); @@ -420,33 +388,9 @@ TextCollection::document_result TCImplementation::Prefix(uchar const * pattern) TextPosition sp = 0, ep = 0; Search(pattern, m, &sp, &ep); - - TextCollection::document_result result; - // Report end-markers in result interval - unsigned resultSize = CountEndmarkers(sp, ep); - if (resultSize == 0) - return result; - - result.reserve(resultSize); // Try to avoid reallocation. - // Iterate through end-markers in [sp,ep]: - unsigned i = 0; - if (sp > 0) - i = alphabetrank->rank(0, sp - 1); - while (resultSize) - { - // End-marker we found belongs to the "preceeding" doc in the collection - DocId docId = ((*endmarkerDocId)[i] + 1) % numberOfTexts; - // Map to doc ID: - docId = emptyTextRank->select0(docId+1); - result.push_back(docId); - - -- resultSize; - ++ i; - } - - return result; + return EnumerateEndmarkers(sp, ep); } TextCollection::document_result TCImplementation::Prefix(uchar const * pattern, DocId begin, DocId end) const @@ -458,33 +402,8 @@ TextCollection::document_result TCImplementation::Prefix(uchar const * pattern, TextPosition sp = 0, ep = 0; Search(pattern, m, &sp, &ep); - TextCollection::document_result result; - - // Report end-markers in result interval - unsigned resultSize = CountEndmarkers(sp, ep); - if (resultSize == 0) - return result; - - result.reserve(resultSize); // Try to avoid reallocation. - - // Iterate through end-markers in [sp,ep] and [begin, end]: - unsigned i = 0; - if (sp > 0) - i = alphabetrank->rank(0, sp - 1); - while (resultSize) - { - // End-marker we found belongs to the "preceeding" doc in the collection - DocId docId = ((*endmarkerDocId)[i] + 1) % numberOfTexts; - // Map to doc ID: - docId = emptyTextRank->select0(docId+1); - if (docId >= begin && docId <= end) - result.push_back(docId); - - -- resultSize; - ++ i; - } - - return result; + // Return end-markers in [sp,ep] and [begin, end]: + return EnumerateEndmarkers(sp, ep, begin, end); } TextCollection::document_result TCImplementation::Suffix(uchar const * pattern) const @@ -500,14 +419,13 @@ TextCollection::document_result TCImplementation::Suffix(uchar const * pattern) TextCollection::document_result result; result.reserve(ep-sp+1); // Try to avoid reallocation. - ulong sampled_rank_i = 0; // Check each occurrence for (; sp <= ep; ++sp) { TextPosition i = sp; uchar c = alphabetrank->access(i); - while (c != '\0' && !sampled->IsBitSet(i, &sampled_rank_i)) + while (c != '\0' && !sampled->access(i)) { i = C[c]+alphabetrank->rank(c,i)-1; c = alphabetrank->access(i); @@ -518,18 +436,11 @@ TextCollection::document_result TCImplementation::Suffix(uchar const * pattern) { // Rank among the end-markers in BWT unsigned endmarkerRank = alphabetrank->rank(0, i) - 1; - - // End-marker that we found belongs to the "preceeding" doc in collection: - DocId docId = ((*endmarkerDocId)[endmarkerRank] + 1) % numberOfTexts; - // Map to doc ID: - docId = emptyTextRank->select0(docId+1); - result.push_back(docId); + result.push_back(Doc->access(endmarkerRank)); } else // Sampled position { - DocId docId = (*suffixDocId)[sampled_rank_i-1]; //sampled->rank(i)-1]; - // Map to doc ID: - docId = emptyTextRank->select0(docId+1); + DocId docId = (*suffixDocId)[sampled->rank1(i)-1]; result.push_back(docId); } } @@ -550,14 +461,13 @@ TextCollection::document_result TCImplementation::Suffix(uchar const * pattern, TextCollection::document_result result; result.reserve(ep-sp+1); // Try to avoid reallocation. - ulong sampled_rank_i = 0; // Check each occurrence, already within [begin, end] for (; sp <= ep; ++sp) { TextPosition i = sp; uchar c = alphabetrank->access(i); - while (c != '\0' && !sampled->IsBitSet(i, &sampled_rank_i)) + while (c != '\0' && !sampled->access(i)) { i = C[c]+alphabetrank->rank(c,i)-1; c = alphabetrank->access(i); @@ -568,18 +478,11 @@ TextCollection::document_result TCImplementation::Suffix(uchar const * pattern, { // Rank among the end-markers in BWT unsigned endmarkerRank = alphabetrank->rank(0, i) - 1; - - // End-marker that we found belongs to the "preceeding" doc in collection: - DocId docId = ((*endmarkerDocId)[endmarkerRank] + 1) % numberOfTexts; - // Map to doc ID: - docId = emptyTextRank->select0(docId+1); - result.push_back(docId); + result.push_back(Doc->access(endmarkerRank)); } else // Sampled position { - DocId docId = (*suffixDocId)[sampled_rank_i-1]; //sampled->rank(i)-1]; - // Map to doc ID: - docId = emptyTextRank->select0(docId+1); + DocId docId = (*suffixDocId)[sampled->rank1(i)-1]; result.push_back(docId); } } @@ -598,31 +501,8 @@ TextCollection::document_result TCImplementation::Equal(uchar const *pattern) co // Match including end-marker Search(pattern, m+1, &sp, &ep); - TextCollection::document_result result; - // Report end-markers in result interval - unsigned resultSize = CountEndmarkers(sp, ep); - if (resultSize == 0) - return result; - - result.reserve(resultSize); // Try to avoid reallocation. - - unsigned i = 0; - if (sp > 0) - i = alphabetrank->rank(0, sp - 1); - while (resultSize) - { - // End-marker we found belongs to the "previous" doc in the collection - DocId docId = ((*endmarkerDocId)[i] + 1) % numberOfTexts; - // Map to doc ID: - docId = emptyTextRank->select0(docId+1); - result.push_back(docId); - - -- resultSize; - ++ i; - } - - return result; + return EnumerateEndmarkers(sp, ep); } TextCollection::document_result TCImplementation::Equal(uchar const *pattern, DocId begin, DocId end) const @@ -635,31 +515,8 @@ TextCollection::document_result TCImplementation::Equal(uchar const *pattern, Do // Match including end-marker Search(pattern, m+1, &sp, &ep, begin, end); - TextCollection::document_result result; - // Report end-markers in result interval - unsigned resultSize = CountEndmarkers(sp, ep); - if (resultSize == 0) - return result; - - result.reserve(resultSize); // Try to avoid reallocation. - - unsigned i = 0; - if (sp > 0) - i = alphabetrank->rank(0, sp - 1); - while (resultSize) - { - // End-marker we found belongs to the "previous" doc in the collection - DocId docId = ((*endmarkerDocId)[i] + 1) % numberOfTexts; - // Map to doc ID: - docId = emptyTextRank->select0(docId+1); - result.push_back(docId); // already within [begin, end] - - -- resultSize; - ++ i; - } - - return result; + return EnumerateEndmarkers(sp, ep, begin, end); } @@ -675,40 +532,10 @@ TextCollection::document_result TCImplementation::Contains(uchar const * pattern // We want unique document indentifiers, using std::set to collect them std::set resultSet; - - ulong sampled_rank_i = 0; - // Check each occurrence - for (; sp <= ep; ++sp) - { - TextPosition i = sp; - uchar c = alphabetrank->access(i); - while (c != '\0' && !sampled->IsBitSet(i, &sampled_rank_i)) - { - i = C[c]+alphabetrank->rank(c,i)-1; // LF-mapping - c = alphabetrank->access(i); - } - if (c == '\0') - { - // Rank among the end-markers in BWT - unsigned endmarkerRank = alphabetrank->rank(0, i) - 1; - - // End-marker that we found belongs to the "preceeding" doc in collection: - DocId docId = ((*endmarkerDocId)[endmarkerRank] + 1) % numberOfTexts; - resultSet.insert(docId); - } - else - { - DocId di = (*suffixDocId)[sampled_rank_i-1]; //sampled->rank(i)-1]; - assert((unsigned)di < numberOfTexts); - resultSet.insert(di); - } - } + EnumerateDocuments(resultSet, sp, ep); // Convert std::set to std::vector TextCollection::document_result result(resultSet.begin(), resultSet.end()); - // Map to doc ID's - for (document_result::iterator it = result.begin(); it != result.end(); ++it) - *it = emptyTextRank->select0(*it+1); return result; } @@ -724,45 +551,18 @@ TextCollection::document_result TCImplementation::Contains(uchar const * pattern // We want unique document indentifiers, using std::set to collect them std::set resultSet; - - ulong sampled_rank_i = 0; - // Check each occurrence - for (; sp <= ep; ++sp) - { - TextPosition i = sp; - uchar c = alphabetrank->access(i); - while (c != '\0' && !sampled->IsBitSet(i, &sampled_rank_i)) - { - i = C[c]+alphabetrank->rank(c,i)-1; // LF-mapping - c = alphabetrank->access(i); - } - if (c == '\0') - { - // Rank among the end-markers in BWT - unsigned endmarkerRank = alphabetrank->rank(0, i) - 1; - - // End-marker that we found belongs to the "preceeding" doc in collection: - DocId docId = ((*endmarkerDocId)[endmarkerRank] + 1) % numberOfTexts; - if (docId >= begin && docId <= end) - resultSet.insert(docId); - } - else - { - DocId docId = (*suffixDocId)[sampled_rank_i-1]; //sampled->rank(i)-1]; - assert((unsigned)docId < numberOfTexts); - if (docId >= begin && docId <= end) - resultSet.insert(docId); - } - } + EnumerateDocuments(resultSet, sp, ep, begin, end); // Convert std::set to std::vector TextCollection::document_result result(resultSet.begin(), resultSet.end()); - // Map to doc ID's - for (document_result::iterator it = result.begin(); it != result.end(); ++it) - *it = emptyTextRank->select0(*it+1); return result; } + +/** + *** +* * FIXME Lessthan or equal + */ TextCollection::document_result TCImplementation::LessThan(uchar const * pattern) const { TextPosition m = strlen((char *)pattern); @@ -772,32 +572,8 @@ TextCollection::document_result TCImplementation::LessThan(uchar const * pattern TextPosition sp = 0, ep = 0; SearchLessThan(pattern, m, &sp, &ep); - TextCollection::document_result result; - // Report end-markers in result interval - unsigned resultSize = CountEndmarkers(sp, ep); - if (resultSize == 0) - return result; - - result.reserve(resultSize); // Try to avoid reallocation. - - // Iterate through end-markers in [sp,ep]: - unsigned i = 0; - if (sp > 0) - i = alphabetrank->rank(0, sp - 1); - while (resultSize) - { - // End-marker we found belongs to the "preceeding" doc in the collection - DocId docId = ((*endmarkerDocId)[i] + 1) % numberOfTexts; - // Map to doc ID: - docId = emptyTextRank->select0(docId+1); - result.push_back(docId); - - -- resultSize; - ++ i; - } - - return result; + return EnumerateEndmarkers(sp, ep); } TextCollection::document_result TCImplementation::LessThan(uchar const * pattern, DocId begin, DocId end) const @@ -809,35 +585,54 @@ TextCollection::document_result TCImplementation::LessThan(uchar const * pattern TextPosition sp = 0, ep = 0; SearchLessThan(pattern, m, &sp, &ep); - TextCollection::document_result result; - - // Report end-markers in result interval - unsigned resultSize = CountEndmarkers(sp, ep); - if (resultSize == 0) - return result; + // Iterate through end-markers in [sp,ep] and [begin, end]: + return EnumerateEndmarkers(sp, ep, begin, end); +} - result.reserve(resultSize); // Try to avoid reallocation. - // Iterate through end-markers in [sp,ep] and [begin, end]: - unsigned i = 0; - if (sp > 0) - i = alphabetrank->rank(0, sp - 1); - while (resultSize) - { - // End-marker we found belongs to the "preceeding" doc in the collection - DocId docId = ((*endmarkerDocId)[i] + 1) % numberOfTexts; - // Map to doc ID: - docId = emptyTextRank->select0(docId+1); - if (docId >= begin && docId <= end) - result.push_back(docId); +TextCollection::document_result TCImplementation::KMismaches(uchar const * pattern, unsigned k) const +{ + TextPosition m = strlen((char *)pattern); + if (m == 0) + return TextCollection::document_result(); // empty result set - -- resultSize; - ++ i; - } + suffix_range_vector ranges; + kmismatches(ranges, pattern, 0, n-1, m, k); + std::set resultSet; + + for (suffix_range_vector::iterator it = ranges.begin(); it != ranges.end(); ++it) + // Iterate through docs in [sp,ep]: + EnumerateDocuments(resultSet, (*it).first, (*it).second); + + // Convert std::set to std::vector + TextCollection::document_result result(resultSet.begin(), resultSet.end()); + return result; +} + +TextCollection::document_result TCImplementation::KErrors(uchar const * pattern, unsigned k) const +{ + TextPosition m = strlen((char *)pattern); + if (m == 0) + return TextCollection::document_result(); // empty result set + + suffix_range_vector ranges; + ulong *dd = new ulong[m+1]; + for (ulong i=0;i resultSet; + for (suffix_range_vector::iterator it = ranges.begin(); it != ranges.end(); ++it) + // Iterate through docs in [sp,ep]: + EnumerateDocuments(resultSet, (*it).first, (*it).second); + + // Convert std::set to std::vector + TextCollection::document_result result(resultSet.begin(), resultSet.end()); return result; } + /** * Full result set queries */ @@ -853,42 +648,7 @@ TextCollection::full_result TCImplementation::FullContains(uchar const * pattern full_result result; result.reserve(ep-sp+1); // Try to avoid reallocation. - - ulong sampled_rank_i = 0; - // Report each occurrence - for (; sp <= ep; ++sp) - { - TextPosition i = sp; - TextPosition dist = 0; - uchar c = alphabetrank->access(i); - while (c != '\0' && !sampled->IsBitSet(i, &sampled_rank_i)) - { - i = C[c]+alphabetrank->rank(c,i)-1; - c = alphabetrank->access(i); - ++ dist; - } - if (c == '\0') - { - // Rank among the end-markers in BWT - unsigned endmarkerRank = alphabetrank->rank(0, i) - 1; - - // End-marker that we found belongs to the "preceeding" doc in collection: - DocId docId = ((*endmarkerDocId)[endmarkerRank] + 1) % numberOfTexts; - // Map to doc ID: - docId = emptyTextRank->select0(docId+1); - result.push_back(make_pair(docId, dist)); - } - else - { - TextPosition textPos = (*suffixes)[sampled_rank_i-1]+dist; //sampled->rank(i)-1] + dist; - DocId docId = (*suffixDocId)[sampled_rank_i-1]; //sampled->rank(i)-1]; -// textPos = textPos - (*textStartPos)[docId]; // Offset inside the text - - // Map to doc ID: - docId = emptyTextRank->select0(docId+1); - result.push_back(make_pair(docId, textPos)); - } - } + EnumeratePositions(result, sp, ep); return result; } @@ -905,48 +665,49 @@ TextCollection::full_result TCImplementation::FullContains(uchar const * pattern full_result result; result.reserve(ep-sp+1); // Try to avoid reallocation. - - ulong sampled_rank_i = 0; - // Report each occurrence - for (; sp <= ep; ++sp) - { - TextPosition i = sp; - TextPosition dist = 0; - uchar c = alphabetrank->access(i); - while (c != '\0' && !sampled->IsBitSet(i, &sampled_rank_i)) - { - i = C[c]+alphabetrank->rank(c,i)-1; - c = alphabetrank->access(i); - ++ dist; - } - if (c == '\0') - { - // Rank among the end-markers in BWT - unsigned endmarkerRank = alphabetrank->rank(0, i) - 1; - - // End-marker that we found belongs to the "preceeding" doc in collection: - DocId docId = ((*endmarkerDocId)[endmarkerRank] + 1) % numberOfTexts; - // Map to doc ID: - docId = emptyTextRank->select0(docId+1); - if (docId >= begin && docId <= end) - result.push_back(make_pair(docId, dist)); - } - else - { - TextPosition textPos = (*suffixes)[sampled_rank_i-1]+dist; //sampled->rank(i)-1] + dist; - DocId docId = (*suffixDocId)[sampled_rank_i-1]; //sampled->rank(i)-1]; -// textPos = textPos - (*textStartPos)[docId]; // Offset inside the text - - // Map to doc ID: - docId = emptyTextRank->select0(docId+1); - if (docId >= begin && docId <= end) - result.push_back(make_pair(docId, textPos)); - } - } + EnumeratePositions(result, sp, ep, begin, end); return result; } +TextCollection::full_result TCImplementation::FullKMismatches(uchar const * pattern, unsigned k) const +{ + TextPosition m = strlen((char *)pattern); + if (m == 0) + return TextCollection::full_result(); // empty result set + + suffix_range_vector ranges; + ulong count = kmismatches(ranges, pattern, 0, n-1, m, k); + + TextCollection::full_result result; + result.reserve(count); // avoid reallocation. + for (suffix_range_vector::iterator it = ranges.begin(); it != ranges.end(); ++it) + // Iterate through docs in [sp,ep]: + EnumeratePositions(result, (*it).first, (*it).second); + return result; +} + +TextCollection::full_result TCImplementation::FullKErrors(uchar const * pattern, unsigned k) const +{ + TextPosition m = strlen((char *)pattern); + if (m == 0) + return TextCollection::full_result(); // empty result set + + suffix_range_vector ranges; + ulong *dd = new ulong[m+1]; + for (unsigned i=0;isave(file); - sampled->Save(file); + sampled->save(file); suffixes->Save(file); suffixDocId->Save(file); if (std::fwrite(&(this->numberOfTexts), sizeof(unsigned), 1, file) != 1) throw std::runtime_error("TCImplementation::Save(): file write error (numberOfTexts)."); - if (std::fwrite(&(this->numberOfAllTexts), sizeof(unsigned), 1, file) != 1) - throw std::runtime_error("TCImplementation::Save(): file write error (numberOfAllTexts)."); if (std::fwrite(&(this->maxTextLength), sizeof(ulong), 1, file) != 1) throw std::runtime_error("TCImplementation::Save(): file write error (maxTextLength)."); - endmarkerDocId->Save(file); - emptyTextRank->Save(file); + Doc->save(file); + textStorage->Save(file); fflush(file); } @@ -1010,11 +767,15 @@ void TCImplementation::Save(FILE *file) const * * Throws a std::runtime_error exception on i/o error. * For more info, see TCImplementation::Save(). + * + * index_mode_t is defined in TextCollection.h and + * defaults to both the index and "naive" text. + * + * Note: Samplerate can not be changed during load. */ -TCImplementation::TCImplementation(FILE *file, unsigned samplerate_) +TCImplementation::TCImplementation(FILE *file, index_mode_t im, unsigned samplerate_) : n(0), samplerate(samplerate_), alphabetrank(0), sampled(0), suffixes(0), - suffixDocId(0), numberOfTexts(0), numberOfAllTexts(0), maxTextLength(0), endmarkerDocId(0), - emptyTextRank(0) + suffixDocId(0), numberOfTexts(0), maxTextLength(0), Doc(0) { uchar verFlag = 0; if (std::fread(&verFlag, 1, 1, file) != 1) @@ -1038,19 +799,24 @@ TCImplementation::TCImplementation(FILE *file, unsigned samplerate_) throw std::runtime_error("TCImplementation::Load(): file read error (bwt end position)."); alphabetrank = static_sequence::load(file); - sampled = new BSGAP(file); + if (im == index_mode_text_only) { delete alphabetrank; alphabetrank = 0; } + + sampled = static_bitsequence::load(file); + if (im == index_mode_text_only) { delete sampled; sampled = 0; } suffixes = new BlockArray(file); + if (im == index_mode_text_only) { delete suffixes; suffixes = 0; } suffixDocId = new BlockArray(file); + if (im == index_mode_text_only) { delete suffixDocId; suffixDocId = 0; } if (std::fread(&(this->numberOfTexts), sizeof(unsigned), 1, file) != 1) throw std::runtime_error("TCImplementation::Load(): file read error (numberOfTexts)."); - if (std::fread(&(this->numberOfAllTexts), sizeof(unsigned), 1, file) != 1) - throw std::runtime_error("TCImplementation::Load(): file read error (numberOfAllTexts)."); if (std::fread(&(this->maxTextLength), sizeof(ulong), 1, file) != 1) throw std::runtime_error("TCImplementation::Load(): file read error (maxTextLength)."); - endmarkerDocId = new BlockArray(file); - emptyTextRank = new BSGAP(file); + Doc = new ArrayDoc(file); //static_sequence::load(file); + if (im == index_mode_text_only) { delete Doc; Doc = 0; } + + textStorage = TextStorage::Load(file); // FIXME Construct data structures with new samplerate //maketables(); @@ -1061,40 +827,97 @@ TCImplementation::TCImplementation(FILE *file, unsigned samplerate_) /** * Rest of the functions follow... */ +ulong TCImplementation::searchPrefix(uchar const *pattern, ulong i, ulong *sp, ulong *ep) const +{ + int c; + while (*sp<=*ep && i>=1) + { + c = (int)pattern[--i]; + *sp = C[c]+alphabetrank->rank(c,*sp-1); + *ep = C[c]+alphabetrank->rank(c,*ep)-1; + } + if (*sp<=*ep) + return *ep - *sp + 1; + else + return 0; +} -// FIXME Use 2D-search structure -unsigned TCImplementation::CountEndmarkers(TextPosition sp, TextPosition ep, DocId begin, DocId end) const +ulong TCImplementation::kmismatches(suffix_range_vector &result, uchar const *pattern, ulong sp, ulong ep, ulong j, unsigned k) const { - if (sp > ep) - return 0; - - ulong ranksp = 0; - if (sp != 0) - ranksp = alphabetrank->rank(0, sp - 1); - - unsigned resultSize = alphabetrank->rank(0, ep) - ranksp; - if (resultSize == 0) - return 0; + if (sp>ep) return 0; + if (j == 0) + { + result.push_back(std::make_pair(sp,ep)); + return ep-sp+1; + } + int c; + ulong spnew; + ulong epnew; + int knew; + ulong sum=0; + if (k==0) + { + sum = searchPrefix(pattern, j, &sp, &ep); + if (sp<=ep) + result.push_back(std::make_pair(sp, ep)); + return sum; + } + vector chars = alphabetrank->accessAll(sp, ep); + for (vector::iterator it = chars.begin(); it != chars.end(); ++it) + { + if (*it == 0) + continue; // skip '\0' + c = *it; + spnew = C[c]+alphabetrank->rank(c,sp-1); + epnew = C[c]+alphabetrank->rank(c,ep)-1; + if (c!=pattern[j-1]) knew = (int)k-1; else knew = k; + if (knew>=0) sum += kmismatches(result, pattern, spnew, epnew, j-1, knew); + } + return sum; +} - // Count end-markers in result interval and within [begin, end] - unsigned count = 0; - unsigned i = ranksp; - while (resultSize) +//first call kerrors(pattern,1,n,m+k,k,d,m), where d[i]=i +ulong TCImplementation::kerrors(suffix_range_vector &result, uchar const *pattern, ulong sp, ulong ep, ulong j, unsigned k, ulong const *d, ulong m) const +{ + ulong sum=0; + if (d[m]<=k) // range of suffixes with at most k-errors found { - // End-marker we found belongs to the "previous" doc in the collection - DocId docId = ((*endmarkerDocId)[i] + 1) % numberOfTexts; - // Map to doc ID: - docId = emptyTextRank->select0(docId+1); - if (docId >= begin && docId <= end) - ++ count; - - -- resultSize; - ++ i; + if (sp<=ep) + result.push_back(std::make_pair(sp, ep)); + sum += (sp<=ep)?ep-sp+1:0ul; } - return count; + if (sp>ep || j==0) + return sum; + ulong *dnew = new ulong[m+1]; + int c; + ulong spnew; + ulong p,lowerbound; + ulong epnew; + vector chars = alphabetrank->accessAll(sp, ep); + for (vector::iterator it = chars.begin(); it != chars.end(); ++it) + { + if (*it == 0) + continue; // skip '\0' + c = *it; + spnew = C[c]+alphabetrank->rank(c,sp-1); + epnew = C[c]+alphabetrank->rank(c,ep)-1; + if (spnew>epnew) continue; + dnew[0]=m+k-j+1; + lowerbound=k+1; + for (p=1; p<=m; p++) { + dnew[p]=myminofthree(d[p]+1,dnew[p-1]+1,(c==pattern[m-p])?d[p-1]:(d[p-1]+1)); + if (dnew[p]rank(c,i) @@ -1152,7 +975,7 @@ ulong TCImplementation::SearchLessThan(uchar const * pattern, TextPosition m, Te { // printf("i = %lu, c = %c, sp = %lu, ep = %lu\n", i, pattern[i], sp, ep); c = (int)pattern[--i]; - uint result = alphabetrank->rank(c,ep); + uint result = alphabetrank->rankLessThan(c,ep); if (result == ~0u) ep = 0; else @@ -1168,16 +991,12 @@ ulong TCImplementation::SearchLessThan(uchar const * pattern, TextPosition m, Te TCImplementation::~TCImplementation() { -#ifdef CSA_TEST_BWT - delete dynFMI; -#endif delete alphabetrank; delete sampled; delete suffixes; delete suffixDocId; - - delete endmarkerDocId; - delete emptyTextRank; + delete Doc; + delete textStorage; } void TCImplementation::makewavelet(uchar *bwt) @@ -1205,23 +1024,26 @@ void TCImplementation::makewavelet(uchar *bwt) // delete [] bwt; //alphabetrank = new RLWaveletTree(bwt, n); // Deletes bwt! // std::cerr << "heap usage: " << HeapProfiler::GetHeapConsumption()/(1024*1024) << " Mbytes" << std::endl; -// std::cerr << "max heap usage before WT: " << HeapProfiler::GetMaxHeapConsumption()/(1024*1024) << " Mbytes" << std::endl; -// HeapProfiler::ResetMaxHeapConsumption(); // FIXME remove +#ifdef DEBUG_MEMUSAGE + std::cerr << "max heap usage before WT: " << HeapProfiler::GetMaxHeapConsumption()/(1024*1024) << " Mbytes" << std::endl; + HeapProfiler::ResetMaxHeapConsumption(); +#endif alphabet_mapper * am = new alphabet_mapper_none(); - static_bitsequence_builder * bmb = new static_bitsequence_builder_rrr02(8); // FIXME samplerate? -// static_bitsequence_builder * bmb = new static_bitsequence_builder_brw32(16); // FIXME samplerate? - wt_coder * wtc = new wt_coder_binary(bwt,n,am); + static_bitsequence_builder * bmb = new static_bitsequence_builder_brw32(8); //rrr02(8); // FIXME samplerate? + wt_coder * wtc = new wt_coder_huff(bwt,n,am);//binary(bwt,n,am); // FIXME Huffman shape alphabetrank = new static_sequence_wvtree(bwt,n,wtc,bmb,am); delete bmb; bwt = 0; // already deleted -// std::cerr << "heap usage: " << HeapProfiler::GetHeapConsumption()/(1024*1024) << " Mbytes" << std::endl; -// std::cerr << "max heap usage after WT: " << HeapProfiler::GetMaxHeapConsumption()/(1024*1024) << " Mbytes" << std::endl; +#ifdef DEBUG_MEMUSAGE + std::cerr << "heap usage after WT: " << HeapProfiler::GetHeapConsumption()/(1024*1024) << " Mbytes" << std::endl; + std::cerr << "max heap usage after WT: " << HeapProfiler::GetMaxHeapConsumption()/(1024*1024) << " Mbytes" << std::endl; +#endif } -void TCImplementation::maketables() +void TCImplementation::maketables(ulong sampleLength, char tsType, CSA::DeltaVector & notIndexed, const string & niText) { // Calculate BWT end-marker position (of last inserted text) { @@ -1237,97 +1059,207 @@ void TCImplementation::maketables() this->bwtEndPos = i; } - // Build up arrays for text length and starting positions - // FIXME Temp, remove - //BlockArray* textLength = new BlockArray(numberOfTexts, Tools::CeilLog2(maxTextLength)); - BlockArray* textStartPos = new BlockArray(numberOfTexts, Tools::CeilLog2(this->n)); - //(*textLength)[0] = l; - (*textStartPos)[0] = 0; +#ifdef DEBUG_MEMUSAGE + std::cerr << "heap usage before BWT traverse: " << HeapProfiler::GetHeapConsumption()/(1024*1024) << " / " << HeapProfiler::GetMaxHeapConsumption()/(1024*1024) << " Mbytes, " << HeapProfiler::GetHeapConsumption() << " / " << HeapProfiler::GetMaxHeapConsumption() << std::endl; + HeapProfiler::ResetMaxHeapConsumption(); +#endif - // Construct samples - ulong sampleLength = (n%samplerate==0) ? n/samplerate : n/samplerate+1; - unsigned ceilLog2n = Tools::CeilLog2(n); - BlockArray* positions = new BlockArray(sampleLength, ceilLog2n); - BlockArray* tmpSuffix = new BlockArray(sampleLength, ceilLog2n); + // Build up array for text starting positions +// BlockArray* textStartPos = new BlockArray(numberOfTexts, Tools::CeilLog2(this->n)); +// (*textStartPos)[0] = 0; // Mapping from end-markers to doc ID's: - endmarkerDocId = new BlockArray(numberOfTexts, Tools::CeilLog2(numberOfTexts)); - - ulong *sampledpositions = new ulong[n/W+1]; - for (ulong i=0;in)); + uint *sampledpositions = new uint[n/(sizeof(uint)*8)+1]; + for (ulong i = 0; i < n / (sizeof(uint)*8) + 1; i++) + sampledpositions[i] = 0; ulong x,p=bwtEndPos; ulong sampleCount = 0; // Keeping track of text position of prev. end-marker seen - ulong posOfSuccEndmarker = n; + ulong posOfSuccEndmarker = n-1; DocId textId = numberOfTexts; ulong ulongmax = 0; ulongmax--; uint alphabetrank_i_tmp =0; - //positions: - for (ulong i=n-1;iGetPos(i) + // Text length = n + number of bytes not indexed. + TextStorageBuilder tsbuilder(n + niText.length()); + ulong tsb_i = n + niText.length(); // Iterator from text length to 0. + string::const_reverse_iterator nit_i = niText.rbegin(); // Iterator through non-indexed texts + + for (ulong i=n-1;iGetPos(i) x=(i==n-1)?0:i+1; - if (x % samplerate == 0 && posOfSuccEndmarker - x > samplerate) { - Tools::SetField(sampledpositions,1,p,1); - (*positions)[sampleCount] = p; - (*tmpSuffix)[sampleCount] = x; // FIXME remove + uchar c = alphabetrank->access(p, alphabetrank_i_tmp); + + tsbuilder[--tsb_i] = c; // Build TextStorage + + if ((posOfSuccEndmarker - i) % samplerate == 0 && c != '\0') + { + set_field(sampledpositions,1,p,1); + (*positions)[sampleCount] = p; sampleCount ++; } - uchar c = alphabetrank->access(p, alphabetrank_i_tmp); if (c == '\0') { - --textId; + unsigned prevTextId = textId; // Cache textId value. + --textId; + /** + * At first c == '\0' it holds that (prevTextId == numberOfTexts), thus, + * we have to search for the first text that is actually *indexed* + * to get correct prevTextId. + */ + if (prevTextId == numberOfTexts) + { + prevTextId = 0; + while (notIndexed.isSet(prevTextId)) + ++ prevTextId; + // Now prevTextId points to the first indexed Doc ID. + } + + /** + * Insert non-indexed texts + */ + while (notIndexed.isSet(textId)) + { + do { + tsbuilder[tsb_i] = *nit_i; + -- tsb_i; + ++ nit_i; + } while (nit_i != niText.rend() && *nit_i != '\0'); + + tsbuilder[tsb_i] = '\0'; + + if (textId == 0) + break; + --textId; + } // Record the order of end-markers in BWT: ulong endmarkerRank = alphabetrank_i_tmp - 1; - (*endmarkerDocId)[endmarkerRank] = textId; + //set_field(endmarkerDocId, logNumberOfTexts, endmarkerRank, (textId + 1) % numberOfTexts); + (*endmarkerDocId)[endmarkerRank] = prevTextId % numberOfTexts; // Store text length and text start position: if (textId < (DocId)numberOfTexts - 1) { - //(*textLength)[textId + 1] = posOfSuccEndmarker - x; - (*textStartPos)[textId + 1] = x; // x is the position of end-marker. - posOfSuccEndmarker = x; +// (*textStartPos)[textId + 1] = x; // x-1 is text position of end-marker. + + posOfSuccEndmarker = i; } - // LF-mapping from '\0' does not work with this (pseudo) BWT (see details from Wolfgang's thesis). - p = textId; // Correct LF-mapping to the last char of the previous text. + // LF-mapping from '\0' does not work with this (pseudo) BWT. + // Correct LF-mapping to the last char of the previous text: + p = textId - notIndexed.rank(textId); } - else + else // Now c != '\0', do LF-mapping: p = C[c]+alphabetrank_i_tmp-1; } + while (textId > 0 && notIndexed.isSet(textId-1)) + { + do { + -- tsb_i; + tsbuilder[tsb_i] = *nit_i; + ++ nit_i; + } while (nit_i != niText.rend() && *nit_i != '\0'); + --textId; + } assert(textId == 0); - - sampled = new BSGAP(sampledpositions,n,true); - sampleLength = sampled->rank(n-1); + assert(tsb_i == 0); + assert(nit_i == niText.rend()); + +#ifdef DEBUG_MEMUSAGE + std::cerr << "heap usage before tsbuilder init: " << HeapProfiler::GetHeapConsumption()/(1024*1024) << " / " << HeapProfiler::GetMaxHeapConsumption()/(1024*1024) << " Mbytes, " << HeapProfiler::GetHeapConsumption() << " / " << HeapProfiler::GetMaxHeapConsumption() << std::endl; + HeapProfiler::ResetMaxHeapConsumption(); +#endif + + textStorage = tsbuilder.InitTextStorage(tsType); + +#ifdef DEBUG_MEMUSAGE + std::cerr << "heap usage after tsbuilder init: " << HeapProfiler::GetHeapConsumption()/(1024*1024) << " / " << HeapProfiler::GetMaxHeapConsumption()/(1024*1024) << " Mbytes, " << HeapProfiler::GetHeapConsumption() << " / " << HeapProfiler::GetMaxHeapConsumption() << std::endl; + HeapProfiler::ResetMaxHeapConsumption(); +#endif + + sampled = new static_bitsequence_rrr02(sampledpositions, n, 16); + delete [] sampledpositions; assert(sampleCount == sampleLength); + assert(sampled->rank1(n-1) == sampleLength); + +#ifdef DEBUG_MEMUSAGE + std::cerr << "heap usage after sampled bit vector: " << HeapProfiler::GetHeapConsumption()/(1024*1024) << " / " << HeapProfiler::GetMaxHeapConsumption()/(1024*1024) << " Mbytes, " << HeapProfiler::GetHeapConsumption() << " / " << HeapProfiler::GetMaxHeapConsumption() << std::endl; + HeapProfiler::ResetMaxHeapConsumption(); +#endif // Suffixes store an offset from the text start position suffixes = new BlockArray(sampleLength, Tools::CeilLog2(maxTextLength)); suffixDocId = new BlockArray(sampleLength, Tools::CeilLog2(numberOfTexts)); - for(ulong i=0; iIsEndmarker(x)) + { + posOfSuccEndmarker = x--; + -- textId; + } + } assert((*positions)[i] < n); - ulong j = sampled->rank((*positions)[i]); - if (j==0) j=sampleLength; - TextPosition textPos = (*tmpSuffix)[i]; - (*suffixDocId)[j-1] = DocIdAtTextPos(textStartPos, textPos); + ulong j = sampled->rank1((*positions)[i]); + + assert(j != 0); // if (j==0) j=sampleLength; + + TextPosition textPos = (x==n-1)?0:x+1; + (*suffixDocId)[j-1] = textId; // textStorage->DocIdAtTextPos(textPos); + assert(textStorage->DocIdAtTextPos(textPos) == textId); - assert((unsigned)DocIdAtTextPos(textStartPos, textPos) < numberOfTexts); assert((*suffixDocId)[j-1] < numberOfTexts); // calculate offset from text start: - (*suffixes)[j-1] = textPos - (*textStartPos)[(*suffixDocId)[j-1]]; + (*suffixes)[j-1] = textPos - textStorage->TextStartPos((*suffixDocId)[j-1]); + --x; + if (x != ~0lu && textStorage->IsEndmarker(x)) + { + posOfSuccEndmarker = x--; + -- textId; + } } - // FIXME Temp, remove - delete tmpSuffix; + delete positions; -// delete textLength; - delete textStartPos; + +#ifdef DEBUG_MEMUSAGE + std::cerr << "heap usage after sampled arrays: " << HeapProfiler::GetHeapConsumption()/(1024*1024) << " / " << HeapProfiler::GetMaxHeapConsumption()/(1024*1024) << " Mbytes, " << HeapProfiler::GetHeapConsumption() << " / " << HeapProfiler::GetMaxHeapConsumption() << std::endl; + HeapProfiler::ResetMaxHeapConsumption(); +#endif + +#ifdef DEBUG_MEMUSAGE + std::cerr << "max heap usage before Doc: " << HeapProfiler::GetMaxHeapConsumption()/(1024*1024) << " Mbytes" << std::endl; + HeapProfiler::ResetMaxHeapConsumption(); +#endif + + /*alphabet_mapper * am = new alphabet_mapper_none(); + static_bitsequence_builder * bmb = new static_bitsequence_builder_rrr02(32); // FIXME samplerate? + Doc = new static_sequence_wvtree_noptrs(endmarkerDocId, numberOfTexts, logNumberOfTexts, bmb, am, true); + delete bmb;*/ + // delete [] endmarkerDocId; // already deleted in static_sequence_wvtree_noptrs! + + Doc = new ArrayDoc(endmarkerDocId); + +#ifdef DEBUG_MEMUSAGE + std::cerr << "max heap usage after Doc: " << HeapProfiler::GetMaxHeapConsumption()/(1024*1024) << " Mbytes" << std::endl; +#endif }