static InverterEntry*
S_fetch_entry(InverterIVARS *ivars, CharBuf *field) {
Schema *const schema = ivars->schema;
int32_t field_num = Seg_Field_Num(ivars->segment, field);
if (!field_num) {
// This field seems not to be in the segment yet. Try to find it in
// the Schema.
if (Schema_Fetch_Type(schema, field)) {
// The field is in the Schema. Get a field num from the Segment.
field_num = Seg_Add_Field(ivars->segment, field);
}
else {
// We've truly failed to find the field. The user must
// not have spec'd it.
THROW(ERR, "Unknown field name: '%o'", field);
}
}
InverterEntry *entry
= (InverterEntry*)VA_Fetch(ivars->entry_pool, field_num);
if (!entry) {
entry = InvEntry_new(schema, (CharBuf*)field, field_num);
VA_Store(ivars->entry_pool, field_num, (Obj*)entry);
}
return entry;
}
VArray*
HeatMap_Flatten_Spans_IMP(HeatMap *self, VArray *spans) {
const uint32_t num_spans = VA_Get_Size(spans);
UNUSED_VAR(self);
if (!num_spans) {
return VA_new(0);
}
else {
VArray *flattened = S_flattened_but_empty_spans(spans);
const uint32_t num_raw_flattened = VA_Get_Size(flattened);
// Iterate over each of the source spans, contributing their scores to
// any destination span that falls within range.
uint32_t dest_tick = 0;
for (uint32_t i = 0; i < num_spans; i++) {
Span *source_span = (Span*)VA_Fetch(spans, i);
int32_t source_span_offset = Span_Get_Offset(source_span);
int32_t source_span_len = Span_Get_Length(source_span);
int32_t source_span_end = source_span_offset + source_span_len;
// Get the location of the flattened span that shares the source
// span's offset.
for (; dest_tick < num_raw_flattened; dest_tick++) {
Span *dest_span = (Span*)VA_Fetch(flattened, dest_tick);
if (Span_Get_Offset(dest_span) == source_span_offset) {
break;
}
}
// Fill in scores.
for (uint32_t j = dest_tick; j < num_raw_flattened; j++) {
Span *dest_span = (Span*)VA_Fetch(flattened, j);
if (Span_Get_Offset(dest_span) == source_span_end) {
break;
}
else {
float new_weight = Span_Get_Weight(dest_span)
+ Span_Get_Weight(source_span);
Span_Set_Weight(dest_span, new_weight);
}
}
}
// Leave holes instead of spans that don't have any score.
dest_tick = 0;
for (uint32_t i = 0; i < num_raw_flattened; i++) {
Span *span = (Span*)VA_Fetch(flattened, i);
if (Span_Get_Weight(span)) {
VA_Store(flattened, dest_tick++, INCREF(span));
}
}
VA_Excise(flattened, dest_tick, num_raw_flattened - dest_tick);
return flattened;
}
}
VArray*
VA_dump(VArray *self)
{
VArray *dump = VA_new(self->size);
u32_t i, max;
for (i = 0, max = self->size; i < max; i++) {
Obj *elem = VA_Fetch(self, i);
if (elem) { VA_Store(dump, i, Obj_Dump(elem)); }
}
return dump;
}
static void
S_init_sub_readers(PolyReader *self, VArray *sub_readers) {
PolyReaderIVARS *const ivars = PolyReader_IVARS(self);
uint32_t num_sub_readers = VA_Get_Size(sub_readers);
int32_t *starts = (int32_t*)MALLOCATE(num_sub_readers * sizeof(int32_t));
Hash *data_readers = Hash_new(0);
DECREF(ivars->sub_readers);
DECREF(ivars->offsets);
ivars->sub_readers = (VArray*)INCREF(sub_readers);
// Accumulate doc_max, subreader start offsets, and DataReaders.
ivars->doc_max = 0;
for (uint32_t i = 0; i < num_sub_readers; i++) {
SegReader *seg_reader = (SegReader*)VA_Fetch(sub_readers, i);
Hash *components = SegReader_Get_Components(seg_reader);
CharBuf *api;
DataReader *component;
starts[i] = ivars->doc_max;
ivars->doc_max += SegReader_Doc_Max(seg_reader);
Hash_Iterate(components);
while (Hash_Next(components, (Obj**)&api, (Obj**)&component)) {
VArray *readers = (VArray*)Hash_Fetch(data_readers, (Obj*)api);
if (!readers) {
readers = VA_new(num_sub_readers);
Hash_Store(data_readers, (Obj*)api, (Obj*)readers);
}
VA_Store(readers, i, INCREF(component));
}
}
ivars->offsets = I32Arr_new_steal(starts, num_sub_readers);
CharBuf *api;
VArray *readers;
Hash_Iterate(data_readers);
while (Hash_Next(data_readers, (Obj**)&api, (Obj**)&readers)) {
DataReader *datareader
= (DataReader*)CERTIFY(S_first_non_null(readers), DATAREADER);
DataReader *aggregator
= DataReader_Aggregator(datareader, readers, ivars->offsets);
if (aggregator) {
CERTIFY(aggregator, DATAREADER);
Hash_Store(ivars->components, (Obj*)api, (Obj*)aggregator);
}
}
DECREF(data_readers);
DeletionsReader *del_reader
= (DeletionsReader*)Hash_Fetch(
ivars->components, (Obj*)VTable_Get_Name(DELETIONSREADER));
ivars->del_count = del_reader ? DelReader_Del_Count(del_reader) : 0;
}
static void
S_round_trip_integer(TestBatch *batch, int64_t value) {
Integer64 *num = Int64_new(value);
VArray *array = VA_new(1);
VA_Store(array, 0, (Obj*)num);
CharBuf *json = Json_to_json((Obj*)array);
Obj *dump = Json_from_json(json);
TEST_TRUE(batch, VA_Equals(array, dump), "Round trip integer %ld",
(long)value);
DECREF(dump);
DECREF(json);
DECREF(array);
}
VArray*
IxManager_recycle(IndexManager *self, PolyReader *reader,
DeletionsWriter *del_writer, int64_t cutoff, bool_t optimize)
{
VArray *seg_readers = PolyReader_Get_Seg_Readers(reader);
VArray *candidates = VA_Gather(seg_readers, S_check_cutoff, &cutoff);
VArray *recyclables = VA_new(VA_Get_Size(candidates));
const uint32_t num_candidates = VA_Get_Size(candidates);
if (optimize) {
DECREF(recyclables);
return candidates;
}
// Sort by ascending size in docs, choose sparsely populated segments.
VA_Sort(candidates, S_compare_doc_count, NULL);
int32_t *counts = (int32_t*)MALLOCATE(num_candidates * sizeof(int32_t));
for (uint32_t i = 0; i < num_candidates; i++) {
SegReader *seg_reader = (SegReader*)CERTIFY(
VA_Fetch(candidates, i), SEGREADER);
counts[i] = SegReader_Doc_Count(seg_reader);
}
I32Array *doc_counts = I32Arr_new_steal(counts, num_candidates);
uint32_t threshold = IxManager_Choose_Sparse(self, doc_counts);
DECREF(doc_counts);
// Move SegReaders to be recycled.
for (uint32_t i = 0; i < threshold; i++) {
VA_Store(recyclables, i, VA_Delete(candidates, i));
}
// Find segments where at least 10% of all docs have been deleted.
for (uint32_t i = threshold; i < num_candidates; i++) {
SegReader *seg_reader = (SegReader*)VA_Delete(candidates, i);
CharBuf *seg_name = SegReader_Get_Seg_Name(seg_reader);
double doc_max = SegReader_Doc_Max(seg_reader);
double num_deletions = DelWriter_Seg_Del_Count(del_writer, seg_name);
double del_proportion = num_deletions / doc_max;
if (del_proportion >= 0.1) {
VA_Push(recyclables, (Obj*)seg_reader);
}
else {
DECREF(seg_reader);
}
}
DECREF(candidates);
return recyclables;
}
static void
S_round_trip_float(TestBatch *batch, double value, double max_diff) {
Float64 *num = Float64_new(value);
VArray *array = VA_new(1);
VA_Store(array, 0, (Obj*)num);
CharBuf *json = Json_to_json((Obj*)array);
Obj *dump = CERTIFY(Json_from_json(json), VARRAY);
Float64 *got = (Float64*)CERTIFY(VA_Fetch((VArray*)dump, 0), FLOAT64);
double diff = Float64_Get_Value(num) - Float64_Get_Value(got);
if (diff < 0) { diff = 0 - diff; }
TEST_TRUE(batch, diff <= max_diff, "Round trip float %f", value);
DECREF(dump);
DECREF(json);
DECREF(array);
}
VArray*
PriQ_Pop_All_IMP(PriorityQueue *self) {
PriorityQueueIVARS *const ivars = PriQ_IVARS(self);
VArray *retval = VA_new(ivars->size);
// Map the queue nodes onto the array in reverse order.
if (ivars->size) {
for (uint32_t i = ivars->size; i--;) {
Obj *const elem = PriQ_Pop(self);
VA_Store(retval, i, elem);
}
}
return retval;
}
VArray*
VA_load(VArray *self, Obj *dump)
{
VArray *source = (VArray*)ASSERT_IS_A(dump, VARRAY);
VArray *loaded = VA_new(source->size);
u32_t i, max;
UNUSED_VAR(self);
for (i = 0, max = source->size; i < max; i++) {
Obj *elem_dump = VA_Fetch(source, i);
if (elem_dump) {
VA_Store(loaded, i, Obj_Load(elem_dump, elem_dump));
}
}
return loaded;
}
void
MemPool_eat(MemoryPool *self, MemoryPool *other) {
i32_t i;
if (self->buf != NULL)
THROW("Memory pool is not empty");
/* Move active arenas from other to self. */
for (i = 0; i <= other->tick; i++) {
ByteBuf *arena = (ByteBuf*)VA_Shift(other->arenas);
/* Maybe displace existing arena. */
VA_Store(self->arenas, i, (Obj*)arena);
}
self->tick = other->tick;
self->last_buf = other->last_buf;
self->buf = other->buf;
self->limit = other->limit;
}
static SortFieldWriter*
S_lazy_init_field_writer(SortWriter *self, int32_t field_num) {
SortWriterIVARS *const ivars = SortWriter_IVARS(self);
SortFieldWriter *field_writer
= (SortFieldWriter*)VA_Fetch(ivars->field_writers, field_num);
if (!field_writer) {
// Open temp files.
if (!ivars->temp_ord_out) {
Folder *folder = ivars->folder;
CharBuf *seg_name = Seg_Get_Name(ivars->segment);
CharBuf *path = CB_newf("%o/sort_ord_temp", seg_name);
ivars->temp_ord_out = Folder_Open_Out(folder, path);
if (!ivars->temp_ord_out) {
DECREF(path);
RETHROW(INCREF(Err_get_error()));
}
CB_setf(path, "%o/sort_ix_temp", seg_name);
ivars->temp_ix_out = Folder_Open_Out(folder, path);
if (!ivars->temp_ix_out) {
DECREF(path);
RETHROW(INCREF(Err_get_error()));
}
CB_setf(path, "%o/sort_dat_temp", seg_name);
ivars->temp_dat_out = Folder_Open_Out(folder, path);
if (!ivars->temp_dat_out) {
DECREF(path);
RETHROW(INCREF(Err_get_error()));
}
DECREF(path);
}
CharBuf *field = Seg_Field_Name(ivars->segment, field_num);
field_writer
= SortFieldWriter_new(ivars->schema, ivars->snapshot, ivars->segment,
ivars->polyreader, field, ivars->mem_pool,
ivars->mem_thresh, ivars->temp_ord_out,
ivars->temp_ix_out, ivars->temp_dat_out);
VA_Store(ivars->field_writers, field_num, (Obj*)field_writer);
}
return field_writer;
}
DefaultDeletionsWriter*
DefDelWriter_init(DefaultDeletionsWriter *self, Schema *schema,
Snapshot *snapshot, Segment *segment,
PolyReader *polyreader) {
DataWriter_init((DataWriter*)self, schema, snapshot, segment, polyreader);
DefaultDeletionsWriterIVARS *const ivars = DefDelWriter_IVARS(self);
ivars->seg_readers = PolyReader_Seg_Readers(polyreader);
uint32_t num_seg_readers = VA_Get_Size(ivars->seg_readers);
ivars->seg_starts = PolyReader_Offsets(polyreader);
ivars->bit_vecs = VA_new(num_seg_readers);
ivars->updated = (bool*)CALLOCATE(num_seg_readers, sizeof(bool));
ivars->searcher = IxSearcher_new((Obj*)polyreader);
ivars->name_to_tick = Hash_new(num_seg_readers);
// Materialize a BitVector of deletions for each segment.
for (uint32_t i = 0; i < num_seg_readers; i++) {
SegReader *seg_reader = (SegReader*)VA_Fetch(ivars->seg_readers, i);
BitVector *bit_vec = BitVec_new(SegReader_Doc_Max(seg_reader));
DeletionsReader *del_reader
= (DeletionsReader*)SegReader_Fetch(
seg_reader, Class_Get_Name(DELETIONSREADER));
Matcher *seg_dels = del_reader
? DelReader_Iterator(del_reader)
: NULL;
if (seg_dels) {
int32_t del;
while (0 != (del = Matcher_Next(seg_dels))) {
BitVec_Set(bit_vec, del);
}
DECREF(seg_dels);
}
VA_Store(ivars->bit_vecs, i, (Obj*)bit_vec);
Hash_Store(ivars->name_to_tick,
(Obj*)SegReader_Get_Seg_Name(seg_reader),
(Obj*)Int32_new(i));
}
return self;
}
DefaultLexiconReader*
DefLexReader_init(DefaultLexiconReader *self, Schema *schema, Folder *folder,
Snapshot *snapshot, VArray *segments, int32_t seg_tick) {
// Init.
LexReader_init((LexiconReader*)self, schema, folder, snapshot, segments,
seg_tick);
DefaultLexiconReaderIVARS *const ivars = DefLexReader_IVARS(self);
Segment *segment = DefLexReader_Get_Segment(self);
// Build an array of SegLexicon objects.
ivars->lexicons = VA_new(Schema_Num_Fields(schema));
for (uint32_t i = 1, max = Schema_Num_Fields(schema) + 1; i < max; i++) {
String *field = Seg_Field_Name(segment, i);
if (field && S_has_data(schema, folder, segment, field)) {
SegLexicon *lexicon = SegLex_new(schema, folder, segment, field);
VA_Store(ivars->lexicons, i, (Obj*)lexicon);
}
}
return self;
}
DefaultLexiconReader*
DefLexReader_init(DefaultLexiconReader *self, Schema *schema, Folder *folder,
Snapshot *snapshot, VArray *segments, i32_t seg_tick)
{
Segment *segment;
u32_t i, max;
/* Init. */
LexReader_init((LexiconReader*)self, schema, folder, snapshot, segments,
seg_tick);
segment = DefLexReader_Get_Segment(self);
/* Build an array of SegLexicon objects. */
self->lexicons = VA_new(Schema_Num_Fields(schema));
for (i = 1, max = Schema_Num_Fields(schema) + 1; i < max; i++) {
CharBuf *field = Seg_Field_Name(segment, i);
if (field && S_has_data(schema, folder, segment, field)) {
SegLexicon *lexicon = SegLex_new(schema, folder, segment, field);
VA_Store(self->lexicons, i, (Obj*)lexicon);
}
}
return self;
}
void
NOTQuery_set_negated_query(NOTQuery *self, Query *negated_query)
{
VA_Store(self->children, 0, INCREF(negated_query));
}
void
NOTQuery_set_negated_query(NOTQuery *self, Query *negated_query) {
NOTQueryIVARS *const ivars = NOTQuery_IVARS(self);
VA_Store(ivars->children, 0, INCREF(negated_query));
}
void
SortColl_collect(SortCollector *self, int32_t doc_id) {
SortCollectorIVARS *const ivars = SortColl_IVARS(self);
// Add to the total number of hits.
ivars->total_hits++;
// Collect this hit if it's competitive.
if (SI_competitive(ivars, doc_id)) {
MatchDoc *const match_doc = ivars->bumped;
MatchDocIVARS *const match_doc_ivars = MatchDoc_IVARS(match_doc);
match_doc_ivars->doc_id = doc_id + ivars->base;
if (ivars->need_score && match_doc_ivars->score == F32_NEGINF) {
match_doc_ivars->score = Matcher_Score(ivars->matcher);
}
// Fetch values so that cross-segment sorting can work.
if (ivars->need_values) {
VArray *values = match_doc_ivars->values;
for (uint32_t i = 0, max = ivars->num_rules; i < max; i++) {
SortCache *cache = ivars->sort_caches[i];
Obj *old_val = (Obj*)VA_Delete(values, i);
if (cache) {
int32_t ord = SortCache_Ordinal(cache, doc_id);
Obj *blank = old_val
? old_val
: SortCache_Make_Blank(cache);
Obj *val = SortCache_Value(cache, ord, blank);
if (val) { VA_Store(values, i, (Obj*)val); }
else { DECREF(blank); }
}
}
}
// Insert the new MatchDoc.
ivars->bumped = (MatchDoc*)HitQ_Jostle(ivars->hit_q, (Obj*)match_doc);
if (ivars->bumped) {
if (ivars->bumped == match_doc) {
/* The queue is full, and we have established a threshold for
* this segment as to what sort of document is definitely not
* acceptable. Turn off AUTO_ACCEPT and start actually
* testing whether hits are competitive. */
ivars->bubble_score = match_doc_ivars->score;
ivars->bubble_doc = doc_id;
ivars->actions = ivars->derived_actions;
}
// Recycle.
MatchDoc_IVARS(ivars->bumped)->score = ivars->need_score
? F32_NEGINF
: F32_NAN;
}
else {
// The queue isn't full yet, so create a fresh MatchDoc.
VArray *values = ivars->need_values
? VA_new(ivars->num_rules)
: NULL;
float fake_score = ivars->need_score ? F32_NEGINF : F32_NAN;
ivars->bumped = MatchDoc_new(INT32_MAX, fake_score, values);
DECREF(values);
}
}
}
