void
PostPool_Add_Segment_IMP(PostingPool *self, SegReader *reader,
I32Array *doc_map, int32_t doc_base) {
PostingPoolIVARS *const ivars = PostPool_IVARS(self);
LexiconReader *lex_reader = (LexiconReader*)SegReader_Fetch(
reader, Class_Get_Name(LEXICONREADER));
Lexicon *lexicon = lex_reader
? LexReader_Lexicon(lex_reader, ivars->field, NULL)
: NULL;
if (lexicon) {
PostingListReader *plist_reader
= (PostingListReader*)SegReader_Fetch(
reader, Class_Get_Name(POSTINGLISTREADER));
PostingList *plist = plist_reader
? PListReader_Posting_List(plist_reader, ivars->field, NULL)
: NULL;
if (!plist) {
THROW(ERR, "Got a Lexicon but no PostingList for '%o' in '%o'",
ivars->field, SegReader_Get_Seg_Name(reader));
}
PostingPool *run
= PostPool_new(ivars->schema, ivars->snapshot, ivars->segment,
ivars->polyreader, ivars->field, ivars->lex_writer,
ivars->mem_pool, ivars->lex_temp_out,
ivars->post_temp_out, ivars->skip_out);
PostingPoolIVARS *const run_ivars = PostPool_IVARS(run);
run_ivars->lexicon = lexicon;
run_ivars->plist = plist;
run_ivars->doc_base = doc_base;
run_ivars->doc_map = (I32Array*)INCREF(doc_map);
PostPool_Add_Run(self, (SortExternal*)run);
}
}
Obj*
Err_certify(Obj *obj, Class *klass, const char *file, int line,
const char *func) {
if (!obj) {
Err_throw_at(ERR, file, line, func, "Object isn't a %o, it's NULL",
Class_Get_Name(klass));
}
else if (!SI_obj_is_a(obj, klass)) {
Err_throw_at(ERR, file, line, func, "Can't downcast from %o to %o",
Obj_get_class_name(obj), Class_Get_Name(klass));
}
return obj;
}
Matcher*
RangeCompiler_Make_Matcher_IMP(RangeCompiler *self, SegReader *reader,
bool need_score) {
RangeQuery *parent = (RangeQuery*)RangeCompiler_IVARS(self)->parent;
String *field = RangeQuery_IVARS(parent)->field;
SortReader *sort_reader
= (SortReader*)SegReader_Fetch(reader, Class_Get_Name(SORTREADER));
SortCache *sort_cache = sort_reader
? SortReader_Fetch_Sort_Cache(sort_reader, field)
: NULL;
UNUSED_VAR(need_score);
if (!sort_cache) {
return NULL;
}
else {
int32_t lower = S_find_lower_bound(self, sort_cache);
int32_t upper = S_find_upper_bound(self, sort_cache);
int32_t max_ord = SortCache_Get_Cardinality(sort_cache) + 1;
if (lower > max_ord || upper < 0) {
return NULL;
}
else {
int32_t doc_max = SegReader_Doc_Max(reader);
return (Matcher*)RangeMatcher_new(lower, upper, sort_cache,
doc_max);
}
}
}
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);
}
void
SortColl_Set_Reader_IMP(SortCollector *self, SegReader *reader) {
SortCollectorIVARS *const ivars = SortColl_IVARS(self);
SortReader *sort_reader
= (SortReader*)SegReader_Fetch(reader, Class_Get_Name(SORTREADER));
// Reset threshold variables and trigger auto-action behavior.
MatchDocIVARS *const bumped_ivars = MatchDoc_IVARS(ivars->bumped);
bumped_ivars->doc_id = INT32_MAX;
ivars->bubble_doc = INT32_MAX;
bumped_ivars->score = ivars->need_score ? CHY_F32_NEGINF : CHY_F32_NAN;
ivars->bubble_score = ivars->need_score ? CHY_F32_NEGINF : CHY_F32_NAN;
ivars->actions = ivars->auto_actions;
// Obtain sort caches. Derive actions array for this segment.
if (ivars->need_values && sort_reader) {
for (uint32_t i = 0, max = ivars->num_rules; i < max; i++) {
SortRule *rule = (SortRule*)Vec_Fetch(ivars->rules, i);
String *field = SortRule_Get_Field(rule);
SortCache *cache = field
? SortReader_Fetch_Sort_Cache(sort_reader, field)
: NULL;
ivars->sort_caches[i] = cache;
ivars->derived_actions[i] = S_derive_action(rule, cache);
if (cache) { ivars->ord_arrays[i] = SortCache_Get_Ords(cache); }
else { ivars->ord_arrays[i] = NULL; }
}
}
ivars->seg_doc_max = reader ? SegReader_Doc_Max(reader) : 0;
SortColl_Set_Reader_t super_set_reader
= (SortColl_Set_Reader_t)SUPER_METHOD_PTR(SORTCOLLECTOR,
LUCY_SortColl_Set_Reader);
super_set_reader(self, reader);
}
Matcher*
DefDelWriter_Seg_Deletions_IMP(DefaultDeletionsWriter *self,
SegReader *seg_reader) {
DefaultDeletionsWriterIVARS *const ivars = DefDelWriter_IVARS(self);
Matcher *deletions = NULL;
Segment *segment = SegReader_Get_Segment(seg_reader);
String *seg_name = Seg_Get_Name(segment);
Integer32 *tick_obj = (Integer32*)Hash_Fetch(ivars->name_to_tick,
(Obj*)seg_name);
int32_t tick = tick_obj ? Int32_Get_Value(tick_obj) : 0;
SegReader *candidate = tick_obj
? (SegReader*)VA_Fetch(ivars->seg_readers, tick)
: NULL;
if (tick_obj) {
DeletionsReader *del_reader
= (DeletionsReader*)SegReader_Obtain(
candidate, Class_Get_Name(DELETIONSREADER));
if (ivars->updated[tick] || DelReader_Del_Count(del_reader)) {
BitVector *deldocs = (BitVector*)VA_Fetch(ivars->bit_vecs, tick);
deletions = (Matcher*)BitVecMatcher_new(deldocs);
}
}
else { // Sanity check.
THROW(ERR, "Couldn't find SegReader %o", seg_reader);
}
return deletions;
}
void
DefDelWriter_Delete_By_Term_IMP(DefaultDeletionsWriter *self,
String *field, Obj *term) {
DefaultDeletionsWriterIVARS *const ivars = DefDelWriter_IVARS(self);
for (uint32_t i = 0, max = VA_Get_Size(ivars->seg_readers); i < max; i++) {
SegReader *seg_reader = (SegReader*)VA_Fetch(ivars->seg_readers, i);
PostingListReader *plist_reader
= (PostingListReader*)SegReader_Fetch(
seg_reader, Class_Get_Name(POSTINGLISTREADER));
BitVector *bit_vec = (BitVector*)VA_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, doc_id);
BitVec_Set(bit_vec, doc_id);
}
if (num_zapped) { ivars->updated[i] = true; }
DECREF(plist);
}
}
}
static Obj*
S_load_via_load_method(Class *klass, Obj *dump) {
Obj *dummy = Class_Make_Obj(klass);
Obj *loaded = NULL;
if (Obj_is_a(dummy, ANALYZER)) {
loaded = Analyzer_Load((Analyzer*)dummy, dump);
}
else if (Obj_is_a(dummy, DOC)) {
loaded = (Obj*)Doc_Load((Doc*)dummy, dump);
}
else if (Obj_is_a(dummy, SIMILARITY)) {
loaded = (Obj*)Sim_Load((Similarity*)dummy, dump);
}
else if (Obj_is_a(dummy, FIELDTYPE)) {
loaded = FType_Load((FieldType*)dummy, dump);
}
else if (Obj_is_a(dummy, SCHEMA)) {
loaded = (Obj*)Schema_Load((Schema*)dummy, dump);
}
else if (Obj_is_a(dummy, QUERY)) {
loaded = Query_Load((Query*)dummy, dump);
}
else {
DECREF(dummy);
THROW(ERR, "Don't know how to load '%o'", Class_Get_Name(klass));
}
DECREF(dummy);
return loaded;
}
Matcher*
TermCompiler_Make_Matcher_IMP(TermCompiler *self, SegReader *reader,
bool need_score) {
TermCompilerIVARS *const ivars = TermCompiler_IVARS(self);
TermQueryIVARS *const parent_ivars
= TermQuery_IVARS((TermQuery*)ivars->parent);
PostingListReader *plist_reader
= (PostingListReader*)SegReader_Fetch(
reader, Class_Get_Name(POSTINGLISTREADER));
PostingList *plist = plist_reader
? PListReader_Posting_List(plist_reader,
parent_ivars->field,
parent_ivars->term)
: NULL;
if (plist == NULL || PList_Get_Doc_Freq(plist) == 0) {
DECREF(plist);
return NULL;
}
else {
Matcher *retval = PList_Make_Matcher(plist, ivars->sim,
(Compiler*)self, need_score);
DECREF(plist);
return retval;
}
}
SegReader*
SegReader_init(SegReader *self, Schema *schema, Folder *folder,
Snapshot *snapshot, Vector *segments, int32_t seg_tick) {
Segment *segment;
IxReader_init((IndexReader*)self, schema, folder, snapshot, segments,
seg_tick, NULL);
SegReaderIVARS *const ivars = SegReader_IVARS(self);
segment = SegReader_Get_Segment(self);
ivars->doc_max = (int32_t)Seg_Get_Count(segment);
ivars->seg_name = (String*)INCREF(Seg_Get_Name(segment));
ivars->seg_num = Seg_Get_Number(segment);
Err *error = Err_trap(S_try_init_components, self);
if (error) {
// An error occurred, so clean up self and rethrow the exception.
DECREF(self);
RETHROW(error);
}
DeletionsReader *del_reader
= (DeletionsReader*)Hash_Fetch(
ivars->components, Class_Get_Name(DELETIONSREADER));
ivars->del_count = del_reader ? DelReader_Del_Count(del_reader) : 0;
return self;
}
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);
}
Obj*
Err_downcast(Obj *obj, Class *klass, const char *file, int line,
const char *func) {
if (obj && !SI_obj_is_a(obj, klass)) {
Err_throw_at(ERR, file, line, func, "Can't downcast from %o to %o",
Obj_get_class_name(obj), Class_Get_Name(klass));
}
return obj;
}
static void
test_Is_A(TestBatchRunner *runner) {
String *string = Str_new_from_trusted_utf8("", 0);
Class *str_class = Str_Get_Class(string);
String *class_name = Str_Get_Class_Name(string);
TEST_TRUE(runner, Str_Is_A(string, STRING), "String Is_A String.");
TEST_TRUE(runner, Str_Is_A(string, OBJ), "String Is_A Obj.");
TEST_TRUE(runner, str_class == STRING, "Get_Class");
TEST_TRUE(runner, Str_Equals(Class_Get_Name(STRING), (Obj*)class_name),
"Get_Class_Name");
DECREF(string);
}
Matcher*
PhraseCompiler_Make_Matcher_IMP(PhraseCompiler *self, SegReader *reader,
bool need_score) {
UNUSED_VAR(need_score);
PhraseCompilerIVARS *const ivars = PhraseCompiler_IVARS(self);
PhraseQueryIVARS *const parent_ivars
= PhraseQuery_IVARS((PhraseQuery*)ivars->parent);
Vector *const terms = parent_ivars->terms;
uint32_t num_terms = Vec_Get_Size(terms);
// Bail if there are no terms.
if (!num_terms) { return NULL; }
// Bail unless field is valid and posting type supports positions.
Similarity *sim = PhraseCompiler_Get_Similarity(self);
Posting *posting = Sim_Make_Posting(sim);
if (posting == NULL || !Obj_is_a((Obj*)posting, SCOREPOSTING)) {
DECREF(posting);
return NULL;
}
DECREF(posting);
// Bail if there's no PostingListReader for this segment.
PostingListReader *const plist_reader
= (PostingListReader*)SegReader_Fetch(
reader, Class_Get_Name(POSTINGLISTREADER));
if (!plist_reader) { return NULL; }
// Look up each term.
Vector *plists = Vec_new(num_terms);
for (uint32_t i = 0; i < num_terms; i++) {
Obj *term = Vec_Fetch(terms, i);
PostingList *plist
= PListReader_Posting_List(plist_reader, parent_ivars->field, term);
// Bail if any one of the terms isn't in the index.
if (!plist || !PList_Get_Doc_Freq(plist)) {
DECREF(plist);
DECREF(plists);
return NULL;
}
Vec_Push(plists, (Obj*)plist);
}
Matcher *retval
= (Matcher*)PhraseMatcher_new(sim, plists, (Compiler*)self);
DECREF(plists);
return retval;
}
void
DefDelWriter_Merge_Segment_IMP(DefaultDeletionsWriter *self,
SegReader *reader, I32Array *doc_map) {
DefaultDeletionsWriterIVARS *const ivars = DefDelWriter_IVARS(self);
UNUSED_VAR(doc_map);
Segment *segment = SegReader_Get_Segment(reader);
Hash *del_meta = (Hash*)Seg_Fetch_Metadata_Utf8(segment, "deletions", 9);
if (del_meta) {
Vector *seg_readers = ivars->seg_readers;
Hash *files = (Hash*)Hash_Fetch_Utf8(del_meta, "files", 5);
if (files) {
HashIterator *iter = HashIter_new(files);
while (HashIter_Next(iter)) {
String *seg = HashIter_Get_Key(iter);
Hash *mini_meta = (Hash*)HashIter_Get_Value(iter);
/* Find the segment the deletions from the SegReader
* we're adding correspond to. If it's gone, we don't
* need to worry about losing deletions files that point
* at it. */
for (size_t i = 0, max = Vec_Get_Size(seg_readers); i < max; i++) {
SegReader *candidate
= (SegReader*)Vec_Fetch(seg_readers, i);
String *candidate_name
= Seg_Get_Name(SegReader_Get_Segment(candidate));
if (Str_Equals(seg, (Obj*)candidate_name)) {
/* If the count hasn't changed, we're about to
* merge away the most recent deletions file
* pointing at this target segment -- so force a
* new file to be written out. */
int32_t count = (int32_t)Json_obj_to_i64(Hash_Fetch_Utf8(mini_meta, "count", 5));
DeletionsReader *del_reader
= (DeletionsReader*)SegReader_Obtain(
candidate, Class_Get_Name(DELETIONSREADER));
if (count == DelReader_Del_Count(del_reader)) {
ivars->updated[i] = true;
}
break;
}
}
}
DECREF(iter);
}
}
}
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;
}
void
HLWriter_Add_Segment_IMP(HighlightWriter *self, SegReader *reader,
I32Array *doc_map) {
HighlightWriterIVARS *const ivars = HLWriter_IVARS(self);
int32_t doc_max = SegReader_Doc_Max(reader);
if (doc_max == 0) {
// Bail if the supplied segment is empty.
return;
}
else {
DefaultHighlightReader *hl_reader
= (DefaultHighlightReader*)CERTIFY(
SegReader_Obtain(reader, Class_Get_Name(HIGHLIGHTREADER)),
DEFAULTHIGHLIGHTREADER);
OutStream *dat_out = S_lazy_init(self);
OutStream *ix_out = ivars->ix_out;
int32_t orig;
ByteBuf *bb = BB_new(0);
for (orig = 1; orig <= doc_max; orig++) {
// Skip deleted docs.
if (doc_map && !I32Arr_Get(doc_map, orig)) {
continue;
}
// Write file pointer.
OutStream_Write_I64(ix_out, OutStream_Tell(dat_out));
// Copy the raw record.
DefHLReader_Read_Record(hl_reader, orig, bb);
OutStream_Write_Bytes(dat_out, BB_Get_Buf(bb), BB_Get_Size(bb));
BB_Set_Size(bb, 0);
}
DECREF(bb);
}
}
static Class*
S_simple_subclass(Class *parent, String *name) {
if (parent->flags & CFISH_fFINAL) {
THROW(ERR, "Can't subclass final class %o", Class_Get_Name(parent));
}
Class *subclass
= (Class*)Memory_wrapped_calloc(parent->class_alloc_size, 1);
Class_Init_Obj(parent->klass, subclass);
subclass->parent = parent;
subclass->flags = parent->flags;
subclass->obj_alloc_size = parent->obj_alloc_size;
subclass->class_alloc_size = parent->class_alloc_size;
subclass->methods = (Method**)CALLOCATE(1, sizeof(Method*));
S_set_name(subclass, Str_Get_Ptr8(name), Str_Get_Size(name));
memcpy(subclass->vtable, parent->vtable,
parent->class_alloc_size - offsetof(Class, vtable));
return subclass;
}
void
DocWriter_Add_Segment_IMP(DocWriter *self, SegReader *reader,
I32Array *doc_map) {
DocWriterIVARS *const ivars = DocWriter_IVARS(self);
int32_t doc_max = SegReader_Doc_Max(reader);
if (doc_max == 0) {
// Bail if the supplied segment is empty.
return;
}
else {
OutStream *const dat_out = S_lazy_init(self);
OutStream *const ix_out = ivars->ix_out;
ByteBuf *const buffer = BB_new(0);
DefaultDocReader *const doc_reader
= (DefaultDocReader*)CERTIFY(
SegReader_Obtain(reader, Class_Get_Name(DOCREADER)),
DEFAULTDOCREADER);
for (int32_t i = 1, max = SegReader_Doc_Max(reader); i <= max; i++) {
if (I32Arr_Get(doc_map, (size_t)i)) {
int64_t start = OutStream_Tell(dat_out);
// Copy record over.
DefDocReader_Read_Record(doc_reader, buffer, i);
const char *buf = BB_Get_Buf(buffer);
size_t size = BB_Get_Size(buffer);
OutStream_Write_Bytes(dat_out, buf, size);
// Write file pointer.
OutStream_Write_I64(ix_out, start);
}
}
DECREF(buffer);
}
}
void
Err_abstract_method_call(Obj *obj, Class *klass, const char *method_name) {
String *class_name = obj ? Obj_get_class_name(obj) : Class_Get_Name(klass);
THROW(ERR, "Abstract method '%s' not defined by %o", method_name,
class_name);
}
static bool
S_merge_updated_deletions(BackgroundMerger *self) {
BackgroundMergerIVARS *const ivars = BGMerger_IVARS(self);
Hash *updated_deletions = NULL;
PolyReader *new_polyreader
= PolyReader_open((Obj*)ivars->folder, NULL, NULL);
Vector *new_seg_readers
= PolyReader_Get_Seg_Readers(new_polyreader);
Vector *old_seg_readers
= PolyReader_Get_Seg_Readers(ivars->polyreader);
Hash *new_segs = Hash_new(Vec_Get_Size(new_seg_readers));
for (uint32_t i = 0, max = Vec_Get_Size(new_seg_readers); i < max; i++) {
SegReader *seg_reader = (SegReader*)Vec_Fetch(new_seg_readers, i);
String *seg_name = SegReader_Get_Seg_Name(seg_reader);
Hash_Store(new_segs, seg_name, INCREF(seg_reader));
}
for (uint32_t i = 0, max = Vec_Get_Size(old_seg_readers); i < max; i++) {
SegReader *seg_reader = (SegReader*)Vec_Fetch(old_seg_readers, i);
String *seg_name = SegReader_Get_Seg_Name(seg_reader);
// If this segment was merged away...
if (Hash_Fetch(ivars->doc_maps, seg_name)) {
SegReader *new_seg_reader
= (SegReader*)CERTIFY(
Hash_Fetch(new_segs, seg_name),
SEGREADER);
int32_t old_del_count = SegReader_Del_Count(seg_reader);
int32_t new_del_count = SegReader_Del_Count(new_seg_reader);
// ... were any new deletions applied against it?
if (old_del_count != new_del_count) {
DeletionsReader *del_reader
= (DeletionsReader*)SegReader_Obtain(
new_seg_reader,
Class_Get_Name(DELETIONSREADER));
if (!updated_deletions) {
updated_deletions = Hash_new(max);
}
Hash_Store(updated_deletions, seg_name,
(Obj*)DelReader_Iterator(del_reader));
}
}
}
DECREF(new_polyreader);
DECREF(new_segs);
if (!updated_deletions) {
return false;
}
else {
PolyReader *merge_polyreader
= PolyReader_open((Obj*)ivars->folder, ivars->snapshot, NULL);
Vector *merge_seg_readers
= PolyReader_Get_Seg_Readers(merge_polyreader);
Snapshot *latest_snapshot
= Snapshot_Read_File(Snapshot_new(), ivars->folder, NULL);
int64_t new_seg_num
= IxManager_Highest_Seg_Num(ivars->manager, latest_snapshot) + 1;
Segment *new_segment = Seg_new(new_seg_num);
SegWriter *seg_writer = SegWriter_new(ivars->schema, ivars->snapshot,
new_segment, merge_polyreader);
DeletionsWriter *del_writer = SegWriter_Get_Del_Writer(seg_writer);
int64_t merge_seg_num = Seg_Get_Number(ivars->segment);
uint32_t seg_tick = INT32_MAX;
int32_t offset = INT32_MAX;
SegWriter_Prep_Seg_Dir(seg_writer);
for (uint32_t i = 0, max = Vec_Get_Size(merge_seg_readers); i < max; i++) {
SegReader *seg_reader
= (SegReader*)Vec_Fetch(merge_seg_readers, i);
if (SegReader_Get_Seg_Num(seg_reader) == merge_seg_num) {
I32Array *offsets = PolyReader_Offsets(merge_polyreader);
seg_tick = i;
offset = I32Arr_Get(offsets, seg_tick);
DECREF(offsets);
}
}
if (offset == INT32_MAX) { THROW(ERR, "Failed sanity check"); }
HashIterator *iter = HashIter_new(updated_deletions);
while (HashIter_Next(iter)) {
String *seg_name = HashIter_Get_Key(iter);
Matcher *deletions = (Matcher*)HashIter_Get_Value(iter);
I32Array *doc_map
= (I32Array*)CERTIFY(
Hash_Fetch(ivars->doc_maps, seg_name),
I32ARRAY);
int32_t del;
while (0 != (del = Matcher_Next(deletions))) {
// Find the slot where the deleted doc resides in the
// rewritten segment. If the doc was already deleted when we
// were merging, do nothing.
int32_t remapped = I32Arr_Get(doc_map, del);
if (remapped) {
// It's a new deletion, so carry it forward and zap it in
// the rewritten segment.
DelWriter_Delete_By_Doc_ID(del_writer, remapped + offset);
}
}
}
DECREF(iter);
// Finish the segment and clean up.
DelWriter_Finish(del_writer);
SegWriter_Finish(seg_writer);
DECREF(seg_writer);
DECREF(new_segment);
DECREF(latest_snapshot);
DECREF(merge_polyreader);
DECREF(updated_deletions);
}
return true;
}
void
LexIndex_Seek_IMP(LexIndex *self, Obj *target) {
LexIndexIVARS *const ivars = LexIndex_IVARS(self);
TermStepper *term_stepper = ivars->term_stepper;
InStream *ix_in = ivars->ix_in;
FieldType *type = ivars->field_type;
int32_t lo = 0;
int32_t hi = ivars->size - 1;
int32_t result = -100;
if (target == NULL || ivars->size == 0) {
ivars->tick = 0;
return;
}
else {
if (!Obj_is_a(target, STRING)) {
THROW(ERR, "Target is a %o, and not comparable to a %o",
Obj_get_class_name(target), Class_Get_Name(STRING));
}
/* TODO:
Obj *first_obj = Vec_Fetch(terms, 0);
if (!Obj_is_a(target, Obj_get_class(first_obj))) {
THROW(ERR, "Target is a %o, and not comparable to a %o",
Obj_get_class_name(target), Obj_get_class_name(first_obj));
}
*/
}
// Divide and conquer.
while (hi >= lo) {
const int32_t mid = lo + ((hi - lo) / 2);
const int64_t offset
= (int64_t)NumUtil_decode_bigend_u64(ivars->offsets + mid);
InStream_Seek(ix_in, offset);
TermStepper_Read_Key_Frame(term_stepper, ix_in);
// Compare values. There is no need for a NULL-check because the term
// number is alway between 0 and ivars->size - 1.
Obj *value = TermStepper_Get_Value(term_stepper);
int32_t comparison = FType_Compare_Values(type, target, value);
if (comparison < 0) {
hi = mid - 1;
}
else if (comparison > 0) {
lo = mid + 1;
}
else {
result = mid;
break;
}
}
// Record the index of the entry we've seeked to, then read entry.
ivars->tick = hi == -1 // indicating that target lt first entry
? 0
: result == -100 // if result is still -100, it wasn't set
? hi
: result;
S_read_entry(self);
}
