Int4
Blast_GetQueryIndexFromContext(Int4 context, EBlastProgramType program)
{
if (program == eBlastTypePsiTblastn || Blast_QueryIsProtein(program)) {
return context;
} else if (Blast_QueryIsTranslated(program)) {
return context / NUM_FRAMES;
} else {
return context / NUM_STRANDS;
}
}
BlastHSPStream*
BlastHSPStreamNew(EBlastProgramType program,
const BlastExtensionOptions* extn_opts,
Boolean sort_on_read,
Int4 num_queries,
BlastHSPWriter *writer)
{
BlastHSPStream* hsp_stream =
(BlastHSPStream*) malloc(sizeof(BlastHSPStream));
hsp_stream->program = program;
hsp_stream->num_hsplists = 0;
hsp_stream->num_hsplists_alloc = 100;
hsp_stream->sorted_hsplists = (BlastHSPList **)malloc(
hsp_stream->num_hsplists_alloc *
sizeof(BlastHSPList *));
hsp_stream->results = Blast_HSPResultsNew(num_queries);
hsp_stream->results_sorted = FALSE;
/* This is needed to meet a pre-condition of the composition-based
* statistics code */
if ((Blast_QueryIsProtein(program) || Blast_QueryIsPssm(program)) &&
extn_opts->compositionBasedStats != 0) {
hsp_stream->sort_by_score =
(SSortByScoreStruct*)calloc(1, sizeof(SSortByScoreStruct));
hsp_stream->sort_by_score->sort_on_read = sort_on_read;
hsp_stream->sort_by_score->first_query_index = 0;
} else {
hsp_stream->sort_by_score = NULL;
}
hsp_stream->x_lock = NULL;
hsp_stream->writer = writer;
hsp_stream->writer_initialized = FALSE;
hsp_stream->writer_finalized = FALSE;
hsp_stream->pre_pipe = NULL;
hsp_stream->tback_pipe = NULL;
return hsp_stream;
}
void
CQuerySplitter::x_ComputeContextOffsets_TranslatedQueries()
{
_ASSERT( !m_QueryChunkFactories.empty() );
const EBlastProgramType kProgram = m_Options->GetProgramType();
_ASSERT(Blast_QueryIsTranslated(kProgram));
const BlastQueryInfo* global_qinfo = m_LocalQueryData->GetQueryInfo();
#ifdef DEBUG_COMPARE_SEQUENCES
const BLAST_SequenceBlk* global_seq = m_LocalQueryData->GetSequenceBlk();
#endif
const size_t kOverlap =
SplitQuery_GetOverlapChunkSize(kProgram) / CODON_LENGTH;
CContextTranslator ctx_translator(*m_SplitBlk, &m_QueryChunkFactories,
m_Options);
CQueryDataPerChunk qdpc(*m_SplitBlk, kProgram, m_LocalQueryData);
vector<const BlastQueryInfo*> chunk_qinfo(m_NumChunks, 0);
for (size_t chunk_num = 0; chunk_num < m_NumChunks; chunk_num++) {
CRef<IQueryFactory> chunk_qf(m_QueryChunkFactories[chunk_num]);
CRef<ILocalQueryData> chunk_qd(chunk_qf->MakeLocalQueryData(m_Options));
#ifdef DEBUG_COMPARE_SEQUENCES
const BLAST_SequenceBlk* chunk_seq = chunk_qd->GetSequenceBlk();
#endif
// BlastQueryInfo structure corresponding to chunk number chunk_num
chunk_qinfo[chunk_num] = chunk_qd->GetQueryInfo();
_ASSERT(chunk_qinfo[chunk_num]);
// In case the first context differs from 0, for consistency with the
// other data returned by this class...
for (Int4 ctx = 0; ctx < chunk_qinfo[chunk_num]->first_context; ctx++) {
m_SplitBlk->AddContextOffsetToChunk(chunk_num, INT4_MAX);
}
for (Int4 ctx = chunk_qinfo[chunk_num]->first_context;
ctx <= chunk_qinfo[chunk_num]->last_context;
ctx++) {
size_t correction = 0;
const int starting_chunk =
ctx_translator.GetStartingChunk(chunk_num, ctx);
const int absolute_context =
ctx_translator.GetAbsoluteContext(chunk_num, ctx);
const int last_query_chunk = qdpc.GetLastChunk(chunk_num, ctx);
if (absolute_context == kInvalidContext ||
starting_chunk == kInvalidContext) {
_ASSERT( !chunk_qinfo[chunk_num]->contexts[ctx].is_valid );
// INT4_MAX is the sentinel value for invalid contexts
m_SplitBlk->AddContextOffsetToChunk(chunk_num, INT4_MAX);
continue;
}
// The corrections for the contexts corresponding to the negative
// strand in the last chunk of a query sequence are all 0
if (!s_IsPlusStrand(chunk_qinfo[chunk_num], ctx) &&
(chunk_num == (size_t)last_query_chunk) &&
(ctx % NUM_FRAMES >= 3)) {
correction = 0;
goto error_check;
}
// The corrections for the contexts corresponding to the plus
// strand are always the same, so only calculate the first one
if (s_IsPlusStrand(chunk_qinfo[chunk_num], ctx) &&
(ctx % NUM_FRAMES == 1 || ctx % NUM_FRAMES == 2)) {
correction = m_SplitBlk->GetContextOffsets(chunk_num).back();
goto error_check;
}
// If the query length is divisible by CODON_LENGTH, the
// corrections for all contexts corresponding to a given strand are
// the same, so only calculate the first one
if ((qdpc.GetQueryLength(chunk_num, ctx) % CODON_LENGTH == 0) &&
(ctx % NUM_FRAMES != 0) && (ctx % NUM_FRAMES != 3)) {
correction = m_SplitBlk->GetContextOffsets(chunk_num).back();
goto error_check;
}
// If the query length % CODON_LENGTH == 1, the corrections for the
// first two contexts of the negative strand are the same, and the
// correction for the last context is one more than that.
if ((qdpc.GetQueryLength(chunk_num, ctx) % CODON_LENGTH == 1) &&
!s_IsPlusStrand(chunk_qinfo[chunk_num], ctx)) {
if (ctx % NUM_FRAMES == 4) {
correction =
m_SplitBlk->GetContextOffsets(chunk_num).back();
goto error_check;
} else if (ctx % NUM_FRAMES == 5) {
correction =
m_SplitBlk->GetContextOffsets(chunk_num).back() + 1;
goto error_check;
}
}
// If the query length % CODON_LENGTH == 2, the corrections for the
// last two contexts of the negative strand are the same, which is
// one more that the first context on the negative strand.
if ((qdpc.GetQueryLength(chunk_num, ctx) % CODON_LENGTH == 2) &&
!s_IsPlusStrand(chunk_qinfo[chunk_num], ctx)) {
if (ctx % NUM_FRAMES == 4) {
correction =
m_SplitBlk->GetContextOffsets(chunk_num).back() + 1;
goto error_check;
} else if (ctx % NUM_FRAMES == 5) {
correction =
m_SplitBlk->GetContextOffsets(chunk_num).back();
goto error_check;
}
}
if (s_IsPlusStrand(chunk_qinfo[chunk_num], ctx)) {
for (int c = chunk_num; c != starting_chunk; c--) {
size_t prev_len = s_GetAbsoluteContextLength(chunk_qinfo,
c - 1,
ctx_translator,
absolute_context);
size_t curr_len = s_GetAbsoluteContextLength(chunk_qinfo, c,
ctx_translator,
absolute_context);
size_t overlap = min(kOverlap, curr_len);
correction += prev_len - min(overlap, prev_len);
}
} else {
size_t subtrahend = 0;
for (int c = chunk_num; c >= starting_chunk && c >= 0; c--) {
size_t prev_len = s_GetAbsoluteContextLength(chunk_qinfo,
c - 1,
ctx_translator,
absolute_context);
size_t curr_len = s_GetAbsoluteContextLength(chunk_qinfo,
c,
ctx_translator,
absolute_context);
size_t overlap = min(kOverlap, curr_len);
subtrahend += (curr_len - min(overlap, prev_len));
}
correction =
global_qinfo->contexts[absolute_context].query_length -
subtrahend;
}
error_check:
_ASSERT((chunk_qinfo[chunk_num]->contexts[ctx].is_valid));
m_SplitBlk->AddContextOffsetToChunk(chunk_num, correction);
#ifdef DEBUG_COMPARE_SEQUENCES
{
int global_offset = global_qinfo->contexts[absolute_context].query_offset +
correction;
int chunk_offset = chunk_qinfo[chunk_num]->contexts[ctx].query_offset;
int num_bases2compare =
min(10, chunk_qinfo[chunk_num]->contexts[ctx].query_length);
if (!cmp_sequence(&global_seq->sequence[global_offset],
&chunk_seq->sequence[chunk_offset],
num_bases2compare, Blast_QueryIsProtein(kProgram))) {
cerr << "Failed to compare sequence data for chunk " << chunk_num
<< ", context " << ctx << endl;
}
}
#endif
}
}
_TRACE("CContextTranslator contents: " << ctx_translator);
}
// Record the correction needed to synchronize with the complete query all the
// query offsets within HSPs that fall within this context
void
CQuerySplitter::x_ComputeContextOffsets_NonTranslatedQueries()
{
_ASSERT( !m_QueryChunkFactories.empty() );
const EBlastProgramType kProgram = m_Options->GetProgramType();
_ASSERT( !Blast_QueryIsTranslated(kProgram) );
const BlastQueryInfo* global_qinfo = m_LocalQueryData->GetQueryInfo();
#ifdef DEBUG_COMPARE_SEQUENCES
const BLAST_SequenceBlk* global_seq = m_LocalQueryData->GetSequenceBlk();
#endif
const size_t kOverlap = SplitQuery_GetOverlapChunkSize(kProgram);
CContextTranslator ctx_translator(*m_SplitBlk, &m_QueryChunkFactories,
m_Options);
vector<const BlastQueryInfo*> chunk_qinfo(m_NumChunks, 0);
for (size_t chunk_num = 0; chunk_num < m_NumChunks; chunk_num++) {
CRef<IQueryFactory> chunk_qf(m_QueryChunkFactories[chunk_num]);
CRef<ILocalQueryData> chunk_qd(chunk_qf->MakeLocalQueryData(m_Options));
#ifdef DEBUG_COMPARE_SEQUENCES
const BLAST_SequenceBlk* chunk_seq = chunk_qd->GetSequenceBlk();
#endif
// BlastQueryInfo structure corresponding to chunk number chunk_num
chunk_qinfo[chunk_num] = chunk_qd->GetQueryInfo();
_ASSERT(chunk_qinfo[chunk_num]);
// In case the first context differs from 0, for consistency with the
// other data returned by this class...
for (Int4 ctx = 0; ctx < chunk_qinfo[chunk_num]->first_context; ctx++) {
m_SplitBlk->AddContextOffsetToChunk(chunk_num, INT4_MAX);
}
for (Int4 ctx = chunk_qinfo[chunk_num]->first_context;
ctx <= chunk_qinfo[chunk_num]->last_context;
ctx++) {
size_t correction = 0;
const int starting_chunk =
ctx_translator.GetStartingChunk(chunk_num, ctx);
const int absolute_context =
ctx_translator.GetAbsoluteContext(chunk_num, ctx);
if (absolute_context == kInvalidContext ||
starting_chunk == kInvalidContext) {
_ASSERT( !chunk_qinfo[chunk_num]->contexts[ctx].is_valid );
// INT4_MAX is the sentinel value for invalid contexts
m_SplitBlk->AddContextOffsetToChunk(chunk_num, INT4_MAX);
continue;
}
if (s_IsPlusStrand(chunk_qinfo[chunk_num], ctx)) {
for (int c = chunk_num; c != starting_chunk; c--) {
size_t prev_len = s_GetAbsoluteContextLength(chunk_qinfo,
c - 1,
ctx_translator,
absolute_context);
size_t curr_len = s_GetAbsoluteContextLength(chunk_qinfo, c,
ctx_translator,
absolute_context);
size_t overlap = min(kOverlap, curr_len);
correction += prev_len - min(overlap, prev_len);
}
} else {
size_t subtrahend = 0;
for (int c = chunk_num; c >= starting_chunk && c >= 0; c--) {
size_t prev_len = s_GetAbsoluteContextLength(chunk_qinfo,
c - 1,
ctx_translator,
absolute_context);
size_t curr_len = s_GetAbsoluteContextLength(chunk_qinfo,
c,
ctx_translator,
absolute_context);
size_t overlap = min(kOverlap, curr_len);
subtrahend += (curr_len - min(overlap, prev_len));
}
correction =
global_qinfo->contexts[absolute_context].query_length -
subtrahend;
}
_ASSERT((chunk_qinfo[chunk_num]->contexts[ctx].is_valid));
m_SplitBlk->AddContextOffsetToChunk(chunk_num, correction);
#ifdef DEBUG_COMPARE_SEQUENCES
{
int global_offset = global_qinfo->contexts[absolute_context].query_offset +
correction;
int chunk_offset = chunk_qinfo[chunk_num]->contexts[ctx].query_offset;
if (!cmp_sequence(&global_seq->sequence[global_offset],
&chunk_seq->sequence[chunk_offset], 10,
Blast_QueryIsProtein(kProgram))) {
cerr << "Failed to compare sequence data!" << endl;
}
}
#endif
}
}
_TRACE("CContextTranslator contents: " << ctx_translator);
}
void
CQuerySplitter::x_ComputeQueryContextsForChunks()
{
const EBlastProgramType kProgram = m_Options->GetProgramType();
const unsigned int kNumContexts = GetNumberOfContexts(kProgram);
const ENa_strand kStrandOption = m_Options->GetStrandOption();
auto_ptr<CQueryDataPerChunk> qdpc;
if (Blast_QueryIsTranslated(kProgram)) {
qdpc.reset(new CQueryDataPerChunk(*m_SplitBlk, kProgram,
m_LocalQueryData));
}
for (size_t chunk_num = 0; chunk_num < m_NumChunks; chunk_num++) {
vector<size_t> queries = m_SplitBlk->GetQueryIndices(chunk_num);
for (size_t i = 0; i < queries.size(); i++) {
CConstRef<CSeq_loc> sl = m_LocalQueryData->GetSeq_loc(queries[i]);
const ENa_strand kStrand =
BlastSetup_GetStrand(*sl, kProgram, kStrandOption);
if (Blast_QueryIsTranslated(kProgram)) {
size_t qlength = qdpc->GetQueryLength(queries[i]);
int last_query_chunk = qdpc->GetLastChunk(queries[i]);
_ASSERT(last_query_chunk != -1);
int shift = s_GetShiftForTranslatedNegStrand(qlength);
for (unsigned int ctx = 0; ctx < kNumContexts; ctx++) {
// handle the plus strand...
if (ctx % NUM_FRAMES < CODON_LENGTH) {
if (kStrand == eNa_strand_minus) {
m_SplitBlk->AddContextToChunk(chunk_num,
kInvalidContext);
} else {
m_SplitBlk->AddContextToChunk(chunk_num,
kNumContexts*queries[i]+ctx);
}
} else { // handle the negative strand
if (kStrand == eNa_strand_plus) {
m_SplitBlk->AddContextToChunk(chunk_num,
kInvalidContext);
} else {
if (chunk_num == (size_t)last_query_chunk) {
// last chunk doesn't have shift
m_SplitBlk->AddContextToChunk(chunk_num,
kNumContexts*queries[i]+ctx);
} else {
m_SplitBlk->AddContextToChunk(chunk_num,
kNumContexts*queries[i]+
s_AddShift(ctx, shift));
}
}
}
}
} else if (Blast_QueryIsNucleotide(kProgram)) {
for (unsigned int ctx = 0; ctx < kNumContexts; ctx++) {
// handle the plus strand...
if (ctx % NUM_STRANDS == 0) {
if (kStrand == eNa_strand_minus) {
m_SplitBlk->AddContextToChunk(chunk_num,
kInvalidContext);
} else {
m_SplitBlk->AddContextToChunk(chunk_num,
kNumContexts*queries[i]+ctx);
}
} else { // handle the negative strand
if (kStrand == eNa_strand_plus) {
m_SplitBlk->AddContextToChunk(chunk_num,
kInvalidContext);
} else {
m_SplitBlk->AddContextToChunk(chunk_num,
kNumContexts*queries[i]+ctx);
}
}
}
} else if (Blast_QueryIsProtein(kProgram)) {
m_SplitBlk->AddContextToChunk(chunk_num,
kNumContexts*queries[i]);
} else {
abort();
}
}
}
}
int CBlastDemoApplication::Run(void)
{
// Get arguments
const CArgs& args = GetArgs();
EProgram program = ProgramNameToEnum(args["program"].AsString());
bool db_is_aa = (program == eBlastp || program == eBlastx ||
program == eRPSBlast || program == eRPSTblastn);
CRef<CBlastOptionsHandle> opts(CBlastOptionsFactory::Create(program, CBlastOptions::eRemote));
ProcessCommandLineArgs(opts);
opts->Validate(); // Can throw CBlastException::eInvalidOptions for invalid option.
// This will dump the options to stderr.
// opts->GetOptions().DebugDumpText(cerr, "opts", 1);
CRef<CObjectManager> objmgr = CObjectManager::GetInstance();
if (!objmgr) {
throw std::runtime_error("Could not initialize object manager");
}
const bool is_protein =
!!Blast_QueryIsProtein(opts->GetOptions().GetProgramType());
SDataLoaderConfig dlconfig(is_protein);
CBlastInputSourceConfig iconfig(dlconfig, objects::eNa_strand_other, false,
args["parse"].AsBoolean());
CBlastFastaInputSource fasta_input(args["in"].AsInputFile(), iconfig);
CScope scope(*objmgr);
CBlastInput blast_input(&fasta_input);
TSeqLocVector query_loc = blast_input.GetAllSeqLocs(scope);
CRef<IQueryFactory> query_factory(new CObjMgr_QueryFactory(query_loc));
const CSearchDatabase target_db(args["db"].AsString(),
db_is_aa ? CSearchDatabase::eBlastDbIsProtein : CSearchDatabase::eBlastDbIsNucleotide);
CRemoteBlast blaster(query_factory, opts, target_db);
// This will dump a lot of stuff to stderr.
// blaster.SetVerbose();
bool status = blaster.SubmitSync();
if (status == false)
throw std::runtime_error("No results returned by SubmitSync");
cerr << "RID: " << blaster.GetRID() << '\n';
CSearchResultSet results = *blaster.GetResultSet();
CNcbiOstream& out = args["out"].AsOutputFile();
for (unsigned int i = 0; i < results.GetNumResults(); i++) {
CConstRef<CSeq_align_set> sas = results[i].GetSeqAlign();
out << MSerial_AsnText << *sas;
}
return 0;
}