void
Indexer_Add_Index_IMP(Indexer *self, Obj *index) {
IndexerIVARS *const ivars = Indexer_IVARS(self);
Folder *other_folder = NULL;
IndexReader *reader = NULL;
if (Obj_is_a(index, FOLDER)) {
other_folder = (Folder*)INCREF(index);
}
else if (Obj_is_a(index, STRING)) {
other_folder = (Folder*)FSFolder_new((String*)index);
}
else {
THROW(ERR, "Invalid type for 'index': %o", Obj_get_class_name(index));
}
reader = IxReader_open((Obj*)other_folder, NULL, NULL);
if (reader == NULL) {
THROW(ERR, "Index doesn't seem to contain any data");
}
else {
Schema *schema = ivars->schema;
Schema *other_schema = IxReader_Get_Schema(reader);
Vector *other_fields = Schema_All_Fields(other_schema);
Vector *seg_readers = IxReader_Seg_Readers(reader);
// Validate schema compatibility and add fields.
Schema_Eat(schema, other_schema);
// Add fields to Segment.
for (size_t i = 0, max = Vec_Get_Size(other_fields); i < max; i++) {
String *other_field = (String*)Vec_Fetch(other_fields, i);
Seg_Add_Field(ivars->segment, other_field);
}
DECREF(other_fields);
// Add all segments.
for (size_t i = 0, max = Vec_Get_Size(seg_readers); i < max; i++) {
SegReader *seg_reader = (SegReader*)Vec_Fetch(seg_readers, i);
DeletionsReader *del_reader
= (DeletionsReader*)SegReader_Fetch(
seg_reader, Class_Get_Name(DELETIONSREADER));
Matcher *deletions = del_reader
? DelReader_Iterator(del_reader)
: NULL;
I32Array *doc_map = DelWriter_Generate_Doc_Map(
ivars->del_writer, deletions,
SegReader_Doc_Max(seg_reader),
(int32_t)Seg_Get_Count(ivars->segment));
SegWriter_Add_Segment(ivars->seg_writer, seg_reader, doc_map);
DECREF(deletions);
DECREF(doc_map);
}
DECREF(seg_readers);
}
DECREF(reader);
DECREF(other_folder);
}
Vector*
HeatMap_Flatten_Spans_IMP(HeatMap *self, Vector *spans) {
const size_t num_spans = Vec_Get_Size(spans);
UNUSED_VAR(self);
if (!num_spans) {
return Vec_new(0);
}
else {
Vector *flattened = S_flattened_but_empty_spans(spans);
const size_t num_raw_flattened = Vec_Get_Size(flattened);
// Iterate over each of the source spans, contributing their scores to
// any destination span that falls within range.
size_t dest_tick = 0;
for (size_t i = 0; i < num_spans; i++) {
Span *source_span = (Span*)Vec_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*)Vec_Fetch(flattened, dest_tick);
if (Span_Get_Offset(dest_span) == source_span_offset) {
break;
}
}
// Fill in scores.
for (size_t j = dest_tick; j < num_raw_flattened; j++) {
Span *dest_span = (Span*)Vec_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 (size_t i = 0; i < num_raw_flattened; i++) {
Span *span = (Span*)Vec_Fetch(flattened, i);
if (Span_Get_Weight(span)) {
Vec_Store(flattened, dest_tick++, INCREF(span));
}
}
Vec_Excise(flattened, dest_tick, num_raw_flattened - dest_tick);
return flattened;
}
}
static Obj*
S_dump_array(Vector *array) {
Vector *dump = Vec_new(Vec_Get_Size(array));
for (size_t i = 0, max = Vec_Get_Size(array); i < max; i++) {
Obj *elem = Vec_Fetch(array, i);
if (elem) {
Vec_Store(dump, i, Freezer_dump(elem));
}
}
return (Obj*)dump;
}
Obj*
S_load_from_array(Vector *dump) {
Vector *loaded = Vec_new(Vec_Get_Size(dump));
for (size_t i = 0, max = Vec_Get_Size(dump); i < max; i++) {
Obj *elem_dump = Vec_Fetch(dump, i);
if (elem_dump) {
Vec_Store(loaded, i, Freezer_load(elem_dump));
}
}
return (Obj*)loaded;
}
QueryParser*
QParser_init(QueryParser *self, Schema *schema, Analyzer *analyzer,
String *default_boolop, Vector *fields) {
QueryParserIVARS *const ivars = QParser_IVARS(self);
// Init.
ivars->heed_colons = false;
ivars->lexer = QueryLexer_new();
// Assign.
ivars->schema = (Schema*)INCREF(schema);
ivars->analyzer = (Analyzer*)INCREF(analyzer);
ivars->default_boolop = default_boolop
? Str_Clone(default_boolop)
: Str_new_from_trusted_utf8("OR", 2);
if (fields) {
ivars->fields = Vec_Clone(fields);
for (uint32_t i = 0, max = Vec_Get_Size(fields); i < max; i++) {
CERTIFY(Vec_Fetch(fields, i), STRING);
}
Vec_Sort(ivars->fields);
}
else {
Vector *all_fields = Schema_All_Fields(schema);
uint32_t num_fields = Vec_Get_Size(all_fields);
ivars->fields = Vec_new(num_fields);
for (uint32_t i = 0; i < num_fields; i++) {
String *field = (String*)Vec_Fetch(all_fields, i);
FieldType *type = Schema_Fetch_Type(schema, field);
if (type && FType_Indexed(type)) {
Vec_Push(ivars->fields, INCREF(field));
}
}
DECREF(all_fields);
}
Vec_Sort(ivars->fields);
// Derive default "occur" from default boolean operator.
if (Str_Equals_Utf8(ivars->default_boolop, "OR", 2)) {
ivars->default_occur = SHOULD;
}
else if (Str_Equals_Utf8(ivars->default_boolop, "AND", 3)) {
ivars->default_occur = MUST;
}
else {
THROW(ERR, "Invalid value for default_boolop: %o", ivars->default_boolop);
}
return self;
}
Compiler*
PhraseQuery_Make_Compiler_IMP(PhraseQuery *self, Searcher *searcher,
float boost, bool subordinate) {
PhraseQueryIVARS *const ivars = PhraseQuery_IVARS(self);
if (Vec_Get_Size(ivars->terms) == 1) {
// Optimize for one-term "phrases".
Obj *term = Vec_Fetch(ivars->terms, 0);
TermQuery *term_query = TermQuery_new(ivars->field, term);
TermCompiler *term_compiler;
TermQuery_Set_Boost(term_query, ivars->boost);
term_compiler
= (TermCompiler*)TermQuery_Make_Compiler(term_query, searcher,
boost, subordinate);
DECREF(term_query);
return (Compiler*)term_compiler;
}
else {
PhraseCompiler *compiler
= PhraseCompiler_new(self, searcher, boost);
if (!subordinate) {
PhraseCompiler_Normalize(compiler);
}
return (Compiler*)compiler;
}
}
void
PolyLex_Reset_IMP(PolyLexicon *self) {
PolyLexiconIVARS *const ivars = PolyLex_IVARS(self);
Vector *seg_lexicons = ivars->seg_lexicons;
uint32_t num_segs = Vec_Get_Size(seg_lexicons);
SegLexQueue *lex_q = ivars->lex_q;
// Empty out the queue.
while (1) {
SegLexicon *seg_lex = (SegLexicon*)SegLexQ_Pop(lex_q);
if (seg_lex == NULL) { break; }
DECREF(seg_lex);
}
// Fill the queue with valid SegLexicons.
for (uint32_t i = 0; i < num_segs; i++) {
SegLexicon *const seg_lexicon
= (SegLexicon*)Vec_Fetch(seg_lexicons, i);
SegLex_Reset(seg_lexicon);
if (SegLex_Next(seg_lexicon)) {
SegLexQ_Insert(ivars->lex_q, INCREF(seg_lexicon));
}
}
if (ivars->term != NULL) {
DECREF(ivars->term);
ivars->term = NULL;
}
}
PolyLexicon*
PolyLex_init(PolyLexicon *self, String *field, Vector *sub_readers) {
uint32_t num_sub_readers = Vec_Get_Size(sub_readers);
Vector *seg_lexicons = Vec_new(num_sub_readers);
// Init.
Lex_init((Lexicon*)self, field);
PolyLexiconIVARS *const ivars = PolyLex_IVARS(self);
ivars->term = NULL;
ivars->lex_q = SegLexQ_new(num_sub_readers);
// Derive.
for (uint32_t i = 0; i < num_sub_readers; i++) {
LexiconReader *lex_reader = (LexiconReader*)Vec_Fetch(sub_readers, i);
if (lex_reader && CERTIFY(lex_reader, LEXICONREADER)) {
Lexicon *seg_lexicon = LexReader_Lexicon(lex_reader, field, NULL);
if (seg_lexicon != NULL) {
Vec_Push(seg_lexicons, (Obj*)seg_lexicon);
}
}
}
ivars->seg_lexicons = seg_lexicons;
PolyLex_Reset(self);
return self;
}
PolySearcher*
PolySearcher_init(PolySearcher *self, Schema *schema, Vector *searchers) {
const uint32_t num_searchers = Vec_Get_Size(searchers);
int32_t *starts_array = (int32_t*)MALLOCATE(num_searchers * sizeof(int32_t));
int32_t doc_max = 0;
Searcher_init((Searcher*)self, schema);
PolySearcherIVARS *const ivars = PolySearcher_IVARS(self);
ivars->searchers = (Vector*)INCREF(searchers);
ivars->starts = NULL; // Safe cleanup.
for (uint32_t i = 0; i < num_searchers; i++) {
Searcher *searcher
= (Searcher*)CERTIFY(Vec_Fetch(searchers, i), SEARCHER);
Schema *candidate = Searcher_Get_Schema(searcher);
Class *orig_class = Schema_Get_Class(schema);
Class *candidate_class = Schema_Get_Class(candidate);
// Confirm that searchers all use the same schema.
if (orig_class != candidate_class) {
THROW(ERR, "Conflicting schemas: '%o', '%o'",
Schema_Get_Class_Name(schema),
Schema_Get_Class_Name(candidate));
}
// Derive doc_max and relative start offsets.
starts_array[i] = (int32_t)doc_max;
doc_max += Searcher_Doc_Max(searcher);
}
ivars->doc_max = doc_max;
ivars->starts = I32Arr_new_steal(starts_array, num_searchers);
return self;
}
// Create all the spans needed by HeatMap_Flatten_Spans, based on the source
// offsets and lengths... but leave the scores at 0.
static Vector*
S_flattened_but_empty_spans(Vector *spans) {
const size_t num_spans = Vec_Get_Size(spans);
int32_t *bounds = (int32_t*)MALLOCATE((num_spans * 2) * sizeof(int32_t));
// Assemble a list of all unique start/end boundaries.
for (size_t i = 0; i < num_spans; i++) {
Span *span = (Span*)Vec_Fetch(spans, i);
bounds[i] = Span_Get_Offset(span);
bounds[i + num_spans] = Span_Get_Offset(span) + Span_Get_Length(span);
}
qsort(bounds, num_spans * 2, sizeof(int32_t), S_compare_i32);
size_t num_bounds = 0;
int32_t last = INT32_MAX;
for (size_t i = 0; i < num_spans * 2; i++) {
if (bounds[i] != last) {
bounds[num_bounds++] = bounds[i];
last = bounds[i];
}
}
// Create one Span for each zone between two bounds.
Vector *flattened = Vec_new(num_bounds - 1);
for (size_t i = 0; i < num_bounds - 1; i++) {
int32_t start = bounds[i];
int32_t length = bounds[i + 1] - start;
Vec_Push(flattened, (Obj*)Span_new(start, length, 0.0f));
}
FREEMEM(bounds);
return flattened;
}
void
SortWriter_Add_Inverted_Doc_IMP(SortWriter *self, Inverter *inverter,
int32_t doc_id) {
SortWriterIVARS *const ivars = SortWriter_IVARS(self);
int32_t field_num;
Inverter_Iterate(inverter);
while (0 != (field_num = Inverter_Next(inverter))) {
FieldType *type = Inverter_Get_Type(inverter);
if (FType_Sortable(type)) {
SortFieldWriter *field_writer
= S_lazy_init_field_writer(self, field_num);
SortFieldWriter_Add(field_writer, doc_id,
Inverter_Get_Value(inverter));
}
}
// If our SortFieldWriters have collectively passed the memory threshold,
// flush all of them, then reset the counter which tracks memory
// consumption.
if ((size_t)Counter_Get_Value(ivars->counter) > ivars->mem_thresh) {
for (size_t i = 0; i < Vec_Get_Size(ivars->field_writers); i++) {
SortFieldWriter *const field_writer
= (SortFieldWriter*)Vec_Fetch(ivars->field_writers, i);
if (field_writer) { SortFieldWriter_Flush(field_writer); }
}
Counter_Reset(ivars->counter);
ivars->flush_at_finish = true;
}
}
void
DefDelWriter_Delete_By_Query_IMP(DefaultDeletionsWriter *self, Query *query) {
DefaultDeletionsWriterIVARS *const ivars = DefDelWriter_IVARS(self);
Compiler *compiler = Query_Make_Compiler(query, (Searcher*)ivars->searcher,
Query_Get_Boost(query), false);
for (size_t i = 0, max = Vec_Get_Size(ivars->seg_readers); i < max; i++) {
SegReader *seg_reader = (SegReader*)Vec_Fetch(ivars->seg_readers, i);
BitVector *bit_vec = (BitVector*)Vec_Fetch(ivars->bit_vecs, i);
Matcher *matcher = Compiler_Make_Matcher(compiler, seg_reader, false);
if (matcher) {
int32_t doc_id;
int32_t num_zapped = 0;
// Iterate through matches, marking each doc as deleted.
while (0 != (doc_id = Matcher_Next(matcher))) {
num_zapped += !BitVec_Get(bit_vec, (size_t)doc_id);
BitVec_Set(bit_vec, (size_t)doc_id);
}
if (num_zapped) { ivars->updated[i] = true; }
DECREF(matcher);
}
}
DECREF(compiler);
}
void
DefDelWriter_Delete_By_Term_IMP(DefaultDeletionsWriter *self,
String *field, Obj *term) {
DefaultDeletionsWriterIVARS *const ivars = DefDelWriter_IVARS(self);
for (size_t i = 0, max = Vec_Get_Size(ivars->seg_readers); i < max; i++) {
SegReader *seg_reader = (SegReader*)Vec_Fetch(ivars->seg_readers, i);
PostingListReader *plist_reader
= (PostingListReader*)SegReader_Fetch(
seg_reader, Class_Get_Name(POSTINGLISTREADER));
BitVector *bit_vec = (BitVector*)Vec_Fetch(ivars->bit_vecs, i);
PostingList *plist = plist_reader
? PListReader_Posting_List(plist_reader, field, term)
: NULL;
int32_t doc_id;
int32_t num_zapped = 0;
// Iterate through postings, marking each doc as deleted.
if (plist) {
while (0 != (doc_id = PList_Next(plist))) {
num_zapped += !BitVec_Get(bit_vec, (size_t)doc_id);
BitVec_Set(bit_vec, (size_t)doc_id);
}
if (num_zapped) { ivars->updated[i] = true; }
DECREF(plist);
}
}
}
Hash*
DefDelWriter_Metadata_IMP(DefaultDeletionsWriter *self) {
DefaultDeletionsWriterIVARS *const ivars = DefDelWriter_IVARS(self);
DefDelWriter_Metadata_t super_meta
= (DefDelWriter_Metadata_t)SUPER_METHOD_PTR(DEFAULTDELETIONSWRITER,
LUCY_DefDelWriter_Metadata);
Hash *const metadata = super_meta(self);
Hash *const files = Hash_new(0);
for (size_t i = 0, max = Vec_Get_Size(ivars->seg_readers); i < max; i++) {
SegReader *seg_reader = (SegReader*)Vec_Fetch(ivars->seg_readers, i);
if (ivars->updated[i]) {
BitVector *deldocs = (BitVector*)Vec_Fetch(ivars->bit_vecs, i);
Segment *segment = SegReader_Get_Segment(seg_reader);
Hash *mini_meta = Hash_new(2);
Hash_Store_Utf8(mini_meta, "count", 5,
(Obj*)Str_newf("%u32", (uint32_t)BitVec_Count(deldocs)));
Hash_Store_Utf8(mini_meta, "filename", 8,
(Obj*)S_del_filename(self, seg_reader));
Hash_Store(files, Seg_Get_Name(segment), (Obj*)mini_meta);
}
}
Hash_Store_Utf8(metadata, "files", 5, (Obj*)files);
return metadata;
}
void
DefDelWriter_Finish_IMP(DefaultDeletionsWriter *self) {
DefaultDeletionsWriterIVARS *const ivars = DefDelWriter_IVARS(self);
Folder *const folder = ivars->folder;
for (size_t i = 0, max = Vec_Get_Size(ivars->seg_readers); i < max; i++) {
SegReader *seg_reader = (SegReader*)Vec_Fetch(ivars->seg_readers, i);
if (ivars->updated[i]) {
BitVector *deldocs = (BitVector*)Vec_Fetch(ivars->bit_vecs, i);
int32_t doc_max = SegReader_Doc_Max(seg_reader);
size_t byte_size = (((size_t)doc_max + 1) + 7) / 8;
size_t new_max = byte_size * 8 - 1;
String *filename = S_del_filename(self, seg_reader);
OutStream *outstream = Folder_Open_Out(folder, filename);
if (!outstream) { RETHROW(INCREF(Err_get_error())); }
// Ensure that we have 1 bit for each doc in segment.
BitVec_Grow(deldocs, new_max);
// Write deletions data and clean up.
OutStream_Write_Bytes(outstream,
(char*)BitVec_Get_Raw_Bits(deldocs),
byte_size);
OutStream_Close(outstream);
DECREF(outstream);
DECREF(filename);
}
}
Seg_Store_Metadata_Utf8(ivars->segment, "deletions", 9,
(Obj*)DefDelWriter_Metadata(self));
}
Vector*
HeatMap_Generate_Proximity_Boosts_IMP(HeatMap *self, Vector *spans) {
Vector *boosts = Vec_new(0);
const size_t num_spans = Vec_Get_Size(spans);
if (num_spans > 1) {
for (size_t i = 0, max = num_spans - 1; i < max; i++) {
Span *span1 = (Span*)Vec_Fetch(spans, i);
for (size_t j = i + 1; j <= max; j++) {
Span *span2 = (Span*)Vec_Fetch(spans, j);
float prox_score
= HeatMap_Calc_Proximity_Boost(self, span1, span2);
if (prox_score == 0) {
break;
}
else {
int32_t length = Span_Get_Offset(span2)
- Span_Get_Offset(span1)
+ Span_Get_Length(span2);
Vec_Push(boosts,
(Obj*)Span_new(Span_Get_Offset(span1), length,
prox_score));
}
}
}
}
return boosts;
}
void
SortWriter_Add_Segment_IMP(SortWriter *self, SegReader *reader,
I32Array *doc_map) {
SortWriterIVARS *const ivars = SortWriter_IVARS(self);
Vector *fields = Schema_All_Fields(ivars->schema);
// Proceed field-at-a-time, rather than doc-at-a-time.
for (size_t i = 0, max = Vec_Get_Size(fields); i < max; i++) {
String *field = (String*)Vec_Fetch(fields, i);
SortReader *sort_reader = (SortReader*)SegReader_Fetch(
reader, Class_Get_Name(SORTREADER));
SortCache *cache = sort_reader
? SortReader_Fetch_Sort_Cache(sort_reader, field)
: NULL;
if (cache) {
int32_t field_num = Seg_Field_Num(ivars->segment, field);
SortFieldWriter *field_writer
= S_lazy_init_field_writer(self, field_num);
SortFieldWriter_Add_Segment(field_writer, reader, doc_map, cache);
ivars->flush_at_finish = true;
}
}
DECREF(fields);
}
PhraseCompiler*
PhraseCompiler_init(PhraseCompiler *self, PhraseQuery *parent,
Searcher *searcher, float boost) {
PhraseCompilerIVARS *const ivars = PhraseCompiler_IVARS(self);
PhraseQueryIVARS *const parent_ivars = PhraseQuery_IVARS(parent);
Schema *schema = Searcher_Get_Schema(searcher);
Similarity *sim = Schema_Fetch_Sim(schema, parent_ivars->field);
Vector *terms = parent_ivars->terms;
// Try harder to find a Similarity if necessary.
if (!sim) { sim = Schema_Get_Similarity(schema); }
// Init.
Compiler_init((Compiler*)self, (Query*)parent, searcher, sim, boost);
// Store IDF for the phrase.
ivars->idf = 0;
for (uint32_t i = 0, max = Vec_Get_Size(terms); i < max; i++) {
Obj *term = Vec_Fetch(terms, i);
int32_t doc_max = Searcher_Doc_Max(searcher);
int32_t doc_freq = Searcher_Doc_Freq(searcher, parent_ivars->field, term);
ivars->idf += Sim_IDF(sim, doc_freq, doc_max);
}
// Calculate raw weight.
ivars->raw_weight = ivars->idf * ivars->boost;
return self;
}
static void
S_compose_inner_queries(QueryParser *self, Vector *elems,
String *default_field) {
const int32_t default_occur = QParser_IVARS(self)->default_occur;
// Generate all queries. Apply any fields.
for (uint32_t i = Vec_Get_Size(elems); i--;) {
String *field = default_field;
ParserElem *elem = (ParserElem*)Vec_Fetch(elems, i);
// Apply field.
if (i > 0) {
// Field specifier must immediately precede any query.
ParserElem* maybe_field_elem
= (ParserElem*)Vec_Fetch(elems, i - 1);
if (ParserElem_Get_Type(maybe_field_elem) == TOKEN_FIELD) {
field = (String*)ParserElem_As(maybe_field_elem, STRING);
}
}
if (ParserElem_Get_Type(elem) == TOKEN_STRING) {
String *text = (String*)ParserElem_As(elem, STRING);
LeafQuery *query = LeafQuery_new(field, text);
ParserElem *new_elem
= ParserElem_new(TOKEN_QUERY, (Obj*)query);
if (default_occur == MUST) {
ParserElem_Require(new_elem);
}
Vec_Store(elems, i, (Obj*)new_elem);
}
}
}
static void
S_discard_elems(Vector *elems, uint32_t type) {
for (size_t i = Vec_Get_Size(elems); i--;) {
ParserElem *elem = (ParserElem*)Vec_Fetch(elems, i);
if (ParserElem_Get_Type(elem) == type) { Vec_Excise(elems, i, 1); }
}
}
static void
S_balance_parens(QueryParser *self, Vector *elems) {
UNUSED_VAR(self);
// Count paren balance, eliminate unbalanced right parens.
int64_t paren_depth = 0;
size_t i = 0;
while (i < Vec_Get_Size(elems)) {
ParserElem *elem = (ParserElem*)Vec_Fetch(elems, i);
if (ParserElem_Get_Type(elem) == TOKEN_OPEN_PAREN) {
paren_depth++;
}
else if (ParserElem_Get_Type(elem) == TOKEN_CLOSE_PAREN) {
if (paren_depth > 0) {
paren_depth--;
}
else {
Vec_Excise(elems, i, 1);
continue;
}
}
i++;
}
// Insert implicit parens.
while (paren_depth--) {
ParserElem *elem = ParserElem_new(TOKEN_CLOSE_PAREN, NULL);
Vec_Push(elems, (Obj*)elem);
}
}
static ORMatcher*
S_ormatcher_init2(ORMatcher *self, ORMatcherIVARS *ivars, Vector *children,
Similarity *sim) {
// Init.
PolyMatcher_init((PolyMatcher*)self, children, sim);
ivars->size = 0;
// Derive.
ivars->max_size = (uint32_t)Vec_Get_Size(children);
// Allocate.
ivars->heap = (HeapedMatcherDoc**)CALLOCATE(ivars->max_size + 1, sizeof(HeapedMatcherDoc*));
// Create a pool of HMDs. Encourage CPU cache hits by using a single
// allocation for all of them.
size_t amount_to_malloc = (ivars->max_size + 1) * sizeof(HeapedMatcherDoc);
ivars->blob = (char*)MALLOCATE(amount_to_malloc);
ivars->pool = (HeapedMatcherDoc**)CALLOCATE(ivars->max_size + 1, sizeof(HeapedMatcherDoc*));
for (uint32_t i = 1; i <= ivars->max_size; i++) {
size_t offset = i * sizeof(HeapedMatcherDoc);
HeapedMatcherDoc *hmd = (HeapedMatcherDoc*)(ivars->blob + offset);
ivars->pool[i] = hmd;
}
// Prime queue.
for (uint32_t i = 0; i < ivars->max_size; i++) {
Matcher *matcher = (Matcher*)Vec_Fetch(children, i);
if (matcher) {
S_add_element(self, ivars, (Matcher*)INCREF(matcher), 0);
}
}
return self;
}
Inversion*
PolyAnalyzer_Transform_Text_IMP(PolyAnalyzer *self, String *text) {
Vector *const analyzers = PolyAnalyzer_IVARS(self)->analyzers;
const uint32_t num_analyzers = Vec_Get_Size(analyzers);
Inversion *retval;
if (num_analyzers == 0) {
size_t token_len = Str_Get_Size(text);
const char *buf = Str_Get_Ptr8(text);
Token *seed = Token_new(buf, token_len, 0, token_len, 1.0f, 1);
retval = Inversion_new(seed);
DECREF(seed);
}
else {
Analyzer *first_analyzer = (Analyzer*)Vec_Fetch(analyzers, 0);
retval = Analyzer_Transform_Text(first_analyzer, text);
for (uint32_t i = 1; i < num_analyzers; i++) {
Analyzer *analyzer = (Analyzer*)Vec_Fetch(analyzers, i);
Inversion *new_inversion = Analyzer_Transform(analyzer, retval);
DECREF(retval);
retval = new_inversion;
}
}
return retval;
}
void
SortEx_Shrink_IMP(SortExternal *self) {
SortExternalIVARS *const ivars = SortEx_IVARS(self);
if (ivars->buf_max - ivars->buf_tick > 0) {
size_t buf_count = SortEx_Buffer_Count(self);
size_t size = buf_count * sizeof(Obj*);
if (ivars->buf_tick > 0) {
Obj **start = ivars->buffer + ivars->buf_tick;
memmove(ivars->buffer, start, size);
}
ivars->buffer = (Obj**)REALLOCATE(ivars->buffer, size);
ivars->buf_tick = 0;
ivars->buf_max = buf_count;
ivars->buf_cap = buf_count;
}
else {
FREEMEM(ivars->buffer);
ivars->buffer = NULL;
ivars->buf_tick = 0;
ivars->buf_max = 0;
ivars->buf_cap = 0;
}
ivars->scratch_cap = 0;
FREEMEM(ivars->scratch);
ivars->scratch = NULL;
for (uint32_t i = 0, max = Vec_Get_Size(ivars->runs); i < max; i++) {
SortExternal *run = (SortExternal*)Vec_Fetch(ivars->runs, i);
SortEx_Shrink(run);
}
}
static Vector*
S_clean_segment_contents(Vector *orig) {
// Since Snapshot format 2, no DataReader has depended on individual files
// within segment directories being listed. Filter these files because
// they cause a problem with FilePurger.
Vector *cleaned = Vec_new(Vec_Get_Size(orig));
for (uint32_t i = 0, max = Vec_Get_Size(orig); i < max; i++) {
String *name = (String*)Vec_Fetch(orig, i);
if (!Seg_valid_seg_name(name)) {
if (Str_Starts_With_Utf8(name, "seg_", 4)) {
continue; // Skip this file.
}
}
Vec_Push(cleaned, INCREF(name));
}
return cleaned;
}
static void
S_compose_or_queries(QueryParser *self, Vector *elems) {
const int32_t default_occur = QParser_IVARS(self)->default_occur;
for (uint32_t i = 0; i + 2 < Vec_Get_Size(elems); i++) {
ParserElem *elem = (ParserElem*)Vec_Fetch(elems, i + 1);
if (ParserElem_Get_Type(elem) == TOKEN_OR) {
ParserElem *preceding = (ParserElem*)Vec_Fetch(elems, i);
Vector *children = Vec_new(2);
uint32_t num_to_zap = 0;
// Add first clause.
Query *preceding_query = (Query*)ParserElem_As(preceding, QUERY);
Vec_Push(children, INCREF(preceding_query));
// Add following clauses.
for (uint32_t j = i + 1, jmax = Vec_Get_Size(elems);
j < jmax;
j += 2, num_to_zap += 2
) {
ParserElem *maybe_or = (ParserElem*)Vec_Fetch(elems, j);
ParserElem *following = (ParserElem*)Vec_Fetch(elems, j + 1);
if (ParserElem_Get_Type(maybe_or) != TOKEN_OR) {
break;
}
else if (ParserElem_Get_Type(following) == TOKEN_QUERY) {
Query *next = (Query*)ParserElem_As(following, QUERY);
Vec_Push(children, INCREF(next));
}
else {
THROW(ERR, "Unexpected type: %u32",
ParserElem_Get_Type(following));
}
}
Query *or_query = QParser_Make_OR_Query(self, children);
ParserElem_Set_Value(preceding, (Obj*)or_query);
if (default_occur == MUST) {
ParserElem_Require(preceding);
}
DECREF(or_query);
DECREF(children);
Vec_Excise(elems, i + 1, num_to_zap);
}
}
}
bool
DefDelWriter_Updated_IMP(DefaultDeletionsWriter *self) {
DefaultDeletionsWriterIVARS *const ivars = DefDelWriter_IVARS(self);
for (size_t i = 0, max = Vec_Get_Size(ivars->seg_readers); i < max; i++) {
if (ivars->updated[i]) { return true; }
}
return false;
}
static void
S_modify_doc_ids(Vector *match_docs, int32_t base) {
for (uint32_t i = 0, max = Vec_Get_Size(match_docs); i < max; i++) {
MatchDoc *match_doc = (MatchDoc*)Vec_Fetch(match_docs, i);
int32_t new_doc_id = MatchDoc_Get_Doc_ID(match_doc) + base;
MatchDoc_Set_Doc_ID(match_doc, new_doc_id);
}
}
void
PolyCompiler_Apply_Norm_Factor_IMP(PolyCompiler *self, float factor) {
PolyCompilerIVARS *const ivars = PolyCompiler_IVARS(self);
for (size_t i = 0, max = Vec_Get_Size(ivars->children); i < max; i++) {
Compiler *child = (Compiler*)Vec_Fetch(ivars->children, i);
Compiler_Apply_Norm_Factor(child, factor);
}
}
void
Indexer_Prepare_Commit_IMP(Indexer *self) {
IndexerIVARS *const ivars = Indexer_IVARS(self);
Vector *seg_readers = PolyReader_Get_Seg_Readers(ivars->polyreader);
size_t num_seg_readers = Vec_Get_Size(seg_readers);
bool merge_happened = false;
if (!ivars->write_lock || ivars->prepared) {
THROW(ERR, "Can't call Prepare_Commit() more than once");
}
// Merge existing index data.
if (num_seg_readers) {
merge_happened = S_maybe_merge(self, seg_readers);
}
// Add a new segment and write a new snapshot file if...
if (Seg_Get_Count(ivars->segment) // Docs/segs added.
|| merge_happened // Some segs merged.
|| !Snapshot_Num_Entries(ivars->snapshot) // Initializing index.
|| DelWriter_Updated(ivars->del_writer)
) {
Folder *folder = ivars->folder;
Schema *schema = ivars->schema;
Snapshot *snapshot = ivars->snapshot;
// Derive snapshot and schema file names.
DECREF(ivars->snapfile);
String *snapfile = IxManager_Make_Snapshot_Filename(ivars->manager);
ivars->snapfile = Str_Cat_Trusted_Utf8(snapfile, ".temp", 5);
DECREF(snapfile);
uint64_t schema_gen = IxFileNames_extract_gen(ivars->snapfile);
char base36[StrHelp_MAX_BASE36_BYTES];
StrHelp_to_base36(schema_gen, &base36);
String *new_schema_name = Str_newf("schema_%s.json", base36);
// Finish the segment, write schema file.
SegWriter_Finish(ivars->seg_writer);
Schema_Write(schema, folder, new_schema_name);
String *old_schema_name = S_find_schema_file(snapshot);
if (old_schema_name) {
Snapshot_Delete_Entry(snapshot, old_schema_name);
}
Snapshot_Add_Entry(snapshot, new_schema_name);
DECREF(new_schema_name);
// Write temporary snapshot file.
Folder_Delete(folder, ivars->snapfile);
Snapshot_Write_File(snapshot, folder, ivars->snapfile);
ivars->needs_commit = true;
}
// Close reader, so that we can delete its files if appropriate.
PolyReader_Close(ivars->polyreader);
ivars->prepared = true;
}
