void
LexWriter_start_field(LexiconWriter *self, i32_t field_num)
{
CharBuf *const seg_name = Seg_Get_Name(self->segment);
Folder *const folder = self->folder;
Snapshot *const snapshot = LexWriter_Get_Snapshot(self);
/* Open outstreams. */
CB_setf(self->dat_file, "%o/lexicon-%i32.dat", seg_name, field_num);
CB_setf(self->ix_file, "%o/lexicon-%i32.ix", seg_name, field_num);
CB_setf(self->ixix_file, "%o/lexicon-%i32.ixix", seg_name, field_num);
Snapshot_Add_Entry(snapshot, self->dat_file);
Snapshot_Add_Entry(snapshot, self->ix_file);
Snapshot_Add_Entry(snapshot, self->ixix_file);
self->dat_out = Folder_Open_Out(folder, self->dat_file);
self->ix_out = Folder_Open_Out(folder, self->ix_file);
self->ixix_out = Folder_Open_Out(folder, self->ixix_file);
if (!self->dat_out) { THROW("Can't open %o", self->dat_file); }
if (!self->ix_out) { THROW("Can't open %o", self->ix_file); }
if (!self->ixix_out) { THROW("Can't open %o", self->ixix_file); }
/* Initialize count and ix_count, TermInfo and last term text. */
self->count = 0;
self->ix_count = 0;
TInfo_Reset(self->last_tinfo);
CB_Set_Size(self->last_text, 0);
}
void
LexWriter_enter_temp_mode(LexiconWriter *self, OutStream *temp_outstream)
{
/* Assign outstream. */
if (self->dat_out != NULL)
THROW("Can't enter temp mode (filename: %o) ", self->dat_file);
self->dat_out = (OutStream*)INCREF(temp_outstream);
/* Initialize count and ix_count, TermInfo and last term text. */
self->count = 0;
self->ix_count = 0;
TInfo_Reset(self->last_tinfo);
CB_Set_Size(self->last_text, 0);
/* Remember that we're in temp mode. */
self->temp_mode = true;
}
void
MatchTInfoStepper_Reset_IMP(MatchTermInfoStepper *self) {
MatchTermInfoStepperIVARS *const ivars = MatchTInfoStepper_IVARS(self);
TInfo_Reset((TermInfo*)ivars->value);
}
static void
S_write_terms_and_postings(PostingPool *self, PostingWriter *post_writer,
OutStream *skip_stream) {
PostingPoolIVARS *const ivars = PostPool_IVARS(self);
TermInfo *const tinfo = TInfo_new(0);
TermInfo *const skip_tinfo = TInfo_new(0);
TermInfoIVARS *const tinfo_ivars = TInfo_IVARS(tinfo);
TermInfoIVARS *const skip_tinfo_ivars = TInfo_IVARS(skip_tinfo);
LexiconWriter *const lex_writer = ivars->lex_writer;
SkipStepper *const skip_stepper = ivars->skip_stepper;
SkipStepperIVARS *const skip_stepper_ivars
= SkipStepper_IVARS(skip_stepper);
int32_t last_skip_doc = 0;
int64_t last_skip_filepos = 0;
const int32_t skip_interval
= Arch_Skip_Interval(Schema_Get_Architecture(ivars->schema));
// Prime heldover variables.
RawPosting *posting
= (RawPosting*)CERTIFY(PostPool_Fetch(self), RAWPOSTING);
RawPostingIVARS *post_ivars = RawPost_IVARS(posting);
CharBuf *last_term_text
= CB_new_from_trusted_utf8(post_ivars->blob, post_ivars->content_len);
const char *last_text_buf = CB_Get_Ptr8(last_term_text);
uint32_t last_text_size = CB_Get_Size(last_term_text);
SkipStepper_Set_ID_And_Filepos(skip_stepper, 0, 0);
// Initialize sentinel to be used on the last iter, using an empty string
// in order to make LexiconWriter Do The Right Thing.
size_t sentinel_size = Class_Get_Obj_Alloc_Size(RAWPOSTING)
+ 20; // blob length + cushion
char empty_string[] = "";
RawPosting *sentinel = RawPost_new(alloca(sentinel_size), 0, 1,
empty_string, 0);
while (1) {
bool same_text_as_last = true;
if (posting == NULL) {
// On the last iter, use an empty string to make LexiconWriter
// DTRT.
posting = sentinel;
post_ivars = RawPost_IVARS(posting);
same_text_as_last = false;
}
else {
// Compare once.
if (post_ivars->content_len != last_text_size
|| memcmp(&post_ivars->blob, last_text_buf, last_text_size) != 0
) {
same_text_as_last = false;
}
}
// If the term text changes, process the last term.
if (!same_text_as_last) {
// Hand off to LexiconWriter.
LexWriter_Add_Term(lex_writer, (Obj*)last_term_text, tinfo);
// Start each term afresh.
TInfo_Reset(tinfo);
PostWriter_Start_Term(post_writer, tinfo);
// Init skip data in preparation for the next term.
skip_stepper_ivars->doc_id = 0;
skip_stepper_ivars->filepos = tinfo_ivars->post_filepos;
last_skip_doc = 0;
last_skip_filepos = tinfo_ivars->post_filepos;
// Remember the term_text so we can write string diffs.
CB_Mimic_Utf8(last_term_text, post_ivars->blob,
post_ivars->content_len);
last_text_buf = CB_Get_Ptr8(last_term_text);
last_text_size = CB_Get_Size(last_term_text);
}
// Bail on last iter before writing invalid posting data.
if (posting == sentinel) { break; }
// Write posting data.
PostWriter_Write_Posting(post_writer, posting);
// Doc freq lags by one iter.
tinfo_ivars->doc_freq++;
// Write skip data.
if (skip_stream != NULL
&& same_text_as_last
&& tinfo_ivars->doc_freq % skip_interval == 0
&& tinfo_ivars->doc_freq != 0
) {
// If first skip group, save skip stream pos for term info.
if (tinfo_ivars->doc_freq == skip_interval) {
tinfo_ivars->skip_filepos = OutStream_Tell(skip_stream);
}
// Write deltas.
last_skip_doc = skip_stepper_ivars->doc_id;
last_skip_filepos = skip_stepper_ivars->filepos;
skip_stepper_ivars->doc_id = post_ivars->doc_id;
PostWriter_Update_Skip_Info(post_writer, skip_tinfo);
skip_stepper_ivars->filepos = skip_tinfo_ivars->post_filepos;
SkipStepper_Write_Record(skip_stepper, skip_stream,
last_skip_doc, last_skip_filepos);
}
// Retrieve the next posting from the sort pool.
// DECREF(posting); // No!! DON'T destroy!!!
posting = (RawPosting*)PostPool_Fetch(self);
post_ivars = RawPost_IVARS(posting);
}
// Clean up.
DECREF(last_term_text);
DECREF(skip_tinfo);
DECREF(tinfo);
}
void
MatchTInfoStepper_reset(MatchTermInfoStepper *self) {
TInfo_Reset((TermInfo*)self->value);
}
