Kouhei Sutou 2019-02-14 19:05:07 +0900 (Thu, 14 Feb 2019) Revision: 4c4446e9d69f717e2d39ccf89153a1189fcf38c9 https://github.com/groonga/groonga/commit/4c4446e9d69f717e2d39ccf89153a1189fcf38c9 Message: ii: use meaningful name for source chunk related variables Modified files: lib/ii.c Modified: lib/ii.c (+150 -108) =================================================================== --- lib/ii.c 2019-02-14 18:14:51 +0900 (d6ad0cded) +++ lib/ii.c 2019-02-14 19:05:07 +0900 (3a08cf1ab) @@ -3186,10 +3186,26 @@ typedef struct { } merger_buffer_data; typedef struct { + uint8_t *data_start; + uint8_t *data_end; + uint8_t *data; + docinfo id; + uint32_t n_documents; + uint32_t *record_id_gaps; + uint32_t *section_id_gaps; + uint32_t *tfs; + uint32_t *weights; + uint32_t *position_gaps; + uint32_t n_position_gaps; +} merger_chunk_data; + +typedef struct { grn_ii *ii; docinfo last_id; + uint64_t position; struct { merger_buffer_data buffer; + merger_chunk_data chunk; } source; struct { uint32_t *record_id_gaps; @@ -3198,19 +3214,6 @@ typedef struct { uint32_t *weights; uint32_t *position_gaps; } dest; - uint8_t *sc; - uint8_t *scp; - uint8_t *sce; - docinfo cid; - grn_id sid; - uint32_t *snp; - uint32_t snn; - uint32_t *srp; - uint32_t *stp; - uint32_t *sop; - uint32_t *ssp; - uint32_t sdf; - uint64_t spos; } merger_data; grn_inline static void @@ -3221,6 +3224,7 @@ merger_report_error(grn_ctx *ctx, docinfo *posting2) { merger_buffer_data *buffer_data = &(data->source.buffer); + merger_chunk_data *chunk_data = &(data->source.chunk); grn_obj term; DEFINE_NAME(data->ii); GRN_TEXT_INIT(&term, 0); @@ -3244,33 +3248,40 @@ merger_report_error(grn_ctx *ctx, posting1->rid, posting1->sid); } GRN_OBJ_FIN(ctx, &term); - merge_dump_source(ctx, data->ii, buffer_data->buffer, data->sc, GRN_LOG_CRIT); + merge_dump_source(ctx, + data->ii, + buffer_data->buffer, + chunk_data->data_start, + GRN_LOG_CRIT); } grn_inline static void merger_get_next_chunk(grn_ctx *ctx, merger_data *data) { - if (data->sdf) { - uint32_t dgap = *data->srp; - data->srp++; - data->cid.rid += dgap; - if (dgap) { data->cid.sid = 0; } - data->snp += data->cid.tf; - data->cid.tf = 1 + *data->stp; - data->stp++; + merger_chunk_data *chunk_data = &(data->source.chunk); + if (chunk_data->n_documents == 0) { + chunk_data->id.rid = 0; + } else { + uint32_t record_id_gap = *chunk_data->record_id_gaps; + chunk_data->record_id_gaps++; + chunk_data->id.rid += record_id_gap; + if (record_id_gap > 0) { + chunk_data->id.sid = 0; + } + chunk_data->position_gaps += chunk_data->id.tf; + chunk_data->id.tf = 1 + *chunk_data->tfs; + chunk_data->tfs++; if ((data->ii->header->flags & GRN_OBJ_WITH_WEIGHT)) { - data->cid.weight = *data->sop; - data->sop++; + chunk_data->id.weight = *chunk_data->weights; + chunk_data->weights++; } if ((data->ii->header->flags & GRN_OBJ_WITH_SECTION)) { - data->cid.sid += 1 + *data->ssp; - data->ssp++; + chunk_data->id.sid += 1 + *chunk_data->section_id_gaps; + chunk_data->section_id_gaps++; } else { - data->cid.sid = 1; + chunk_data->id.sid = 1; } - data->sdf--; - } else { - data->cid.rid = 0; + chunk_data->n_documents--; } } @@ -3303,33 +3314,34 @@ merger_put_next_id(grn_ctx *ctx, merger_data *data, docinfo *id) grn_inline static void merger_put_next_chunk(grn_ctx *ctx, merger_data *data) { - if (data->cid.rid) { - if (data->cid.tf) { - if (data->last_id.rid > data->cid.rid || - (data->last_id.rid == data->cid.rid && - data->last_id.sid >= data->cid.sid)) { + merger_chunk_data *chunk_data = &(data->source.chunk); + if (chunk_data->id.rid > 0) { + if (chunk_data->id.tf > 0) { + if (data->last_id.rid > chunk_data->id.rid || + (data->last_id.rid == chunk_data->id.rid && + data->last_id.sid >= chunk_data->id.sid)) { merger_report_error(ctx, data, "the last posting is larger than " "the next posting in chunk", &(data->last_id), - &(data->cid)); + &(chunk_data->id)); return; } - merger_put_next_id(ctx, data, &(data->cid)); + merger_put_next_id(ctx, data, &(chunk_data->id)); if ((data->ii->header->flags & GRN_OBJ_WITH_POSITION)) { uint32_t i; - for (i = 0; i < data->cid.tf; i++) { - *(data->dest.position_gaps) = data->snp[i]; + for (i = 0; i < chunk_data->id.tf; i++) { + *(data->dest.position_gaps) = chunk_data->position_gaps[i]; data->dest.position_gaps++; - data->spos += data->snp[i]; + data->position += chunk_data->position_gaps[i]; } } } else { merger_report_error(ctx, data, "invalid posting in chunk", - &(data->cid), + &(chunk_data->id), NULL); return; } @@ -3396,7 +3408,7 @@ merger_put_next_buffer(grn_ctx *ctx, merger_data *data) if ((data->ii->header->flags & GRN_OBJ_WITH_POSITION)) { while (buffer_data->id.tf--) { GRN_B_DEC(*(data->dest.position_gaps), buffer_data->data); - data->spos += *(data->dest.position_gaps); + data->position += *(data->dest.position_gaps); data->dest.position_gaps++; } } @@ -3409,22 +3421,23 @@ grn_inline static grn_bool merger_merge(grn_ctx *ctx, merger_data *data) { merger_buffer_data *buffer_data = &(data->source.buffer); - if (buffer_data->id.rid) { - if (data->cid.rid) { - if (data->cid.rid < buffer_data->id.rid) { + merger_chunk_data *chunk_data = &(data->source.chunk); + if (buffer_data->id.rid > 0) { + if (chunk_data->id.rid > 0) { + if (chunk_data->id.rid < buffer_data->id.rid) { merger_put_next_chunk(ctx, data); if (ctx->rc != GRN_SUCCESS) { return GRN_FALSE; } } else { - if (buffer_data->id.rid < data->cid.rid) { + if (buffer_data->id.rid < chunk_data->id.rid) { merger_put_next_buffer(ctx, data); if (ctx->rc != GRN_SUCCESS) { return GRN_FALSE; } } else { - if (buffer_data->id.sid) { - if (data->cid.sid < buffer_data->id.sid) { + if (buffer_data->id.sid > 0) { + if (chunk_data->id.sid < buffer_data->id.sid) { merger_put_next_chunk(ctx, data); if (ctx->rc != GRN_SUCCESS) { return GRN_FALSE; } } else { - if (buffer_data->id.sid == data->cid.sid) { + if (buffer_data->id.sid == chunk_data->id.sid) { merger_get_next_chunk(ctx, data); if (ctx->rc != GRN_SUCCESS) { return GRN_FALSE; } } @@ -3442,7 +3455,7 @@ merger_merge(grn_ctx *ctx, merger_data *data) if (ctx->rc != GRN_SUCCESS) { return GRN_FALSE; } } } else { - if (data->cid.rid) { + if (chunk_data->id.rid > 0) { merger_put_next_chunk(ctx, data); if (ctx->rc != GRN_SUCCESS) { return GRN_FALSE; } } else { @@ -3494,6 +3507,7 @@ chunk_merge(grn_ctx *ctx, int32_t *balance) { merger_buffer_data *buffer_data = &(data->source.buffer); + merger_chunk_data *chunk_data = &(data->source.chunk); grn_ii *ii = data->ii; grn_io_win sw; uint32_t segno = cinfo->segno; @@ -3523,9 +3537,8 @@ chunk_merge(grn_ctx *ctx, } data->last_id = null_docinfo; - data->snn = 0; - data->ssp = NULL; - data->sop = NULL; + chunk_data->section_id_gaps = NULL; + chunk_data->weights = NULL; datavec_init(ctx, rdv, ii->n_elements, 0, 0); if (ctx->rc != GRN_SUCCESS) { @@ -3568,22 +3581,26 @@ chunk_merge(grn_ctx *ctx, } bufsize += decoded_size; } - /* (df in chunk list) = a[1] - sdf; */ + /* (df in chunk list) = a[1] - chunk_data->n_documents; */ { int j = 0; - data->sdf = rdv[j].data_size; - data->srp = rdv[j++].data; + chunk_data->n_documents = rdv[j].data_size; + chunk_data->record_id_gaps = rdv[j++].data; if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { - data->ssp = rdv[j++].data; + chunk_data->section_id_gaps = rdv[j++].data; } - data->stp = rdv[j++].data; + chunk_data->tfs = rdv[j++].data; if ((ii->header->flags & GRN_OBJ_WITH_WEIGHT)) { - data->sop = rdv[j++].data; + chunk_data->weights = rdv[j++].data; } - data->snn = rdv[j].data_size; - data->snp = rdv[j].data; + chunk_data->position_gaps = rdv[j].data; + chunk_data->n_position_gaps = rdv[j].data_size; } - datavec_reset(ctx, dv, ii->n_elements, data->sdf + S_SEGMENT, bufsize); + datavec_reset(ctx, + dv, + ii->n_elements, + chunk_data->n_documents + S_SEGMENT, + bufsize); if (ctx->rc != GRN_SUCCESS) { grn_obj term; DEFINE_NAME(ii); @@ -3631,7 +3648,7 @@ chunk_merge(grn_ctx *ctx, if (!merger_merge(ctx, data)) { break; } - } while (buffer_data->id.rid <= rid || data->cid.rid); + } while (buffer_data->id.rid <= rid || chunk_data->id.rid); if (ctx->rc != GRN_SUCCESS) { grn_obj term; DEFINE_NAME(ii); @@ -3665,7 +3682,12 @@ chunk_merge(grn_ctx *ctx, dv[j].data_size = ndf; dv[j++].flags = f_s; } if ((ii->header->flags & GRN_OBJ_WITH_POSITION)) { - uint32_t f_p = ((np < 32) || (np <= (data->spos >> 13))) ? 0 : USE_P_ENC; + uint32_t f_p; + if ((np < 32) || (np <= (data->position >> 13))) { + f_p = 0; + } else { + f_p = USE_P_ENC; + } dv[j].data_size = np; dv[j].flags = f_p|ODD; } encsize_probably_enough = (ndf * 4 + np) * 3; @@ -3696,7 +3718,7 @@ chunk_merge(grn_ctx *ctx, } chunk_free(ctx, ii, segno, 0, size); } - *balance += (ndf - data->sdf); + *balance += (ndf - chunk_data->n_documents); exit : datavec_fin(ctx, rdv); @@ -3800,6 +3822,7 @@ buffer_merge(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h, for (bt = db->terms; n; n--, bt++) { merger_data data = {0}; merger_buffer_data *buffer_data = &(data.source.buffer); + merger_chunk_data *chunk_data = &(data.source.chunk); int32_t balance = 0; uint32_t nchunks = 0; uint32_t nvchunks = 0; @@ -3808,9 +3831,9 @@ buffer_merge(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h, uint32_t ndf; data.ii = ii; - data.sc = sc; buffer_data->buffer = sb; buffer_data->term = bt; + chunk_data->data_start = sc; if (!bt->tid) { nterms_void++; @@ -3821,7 +3844,7 @@ buffer_merge(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h, GRN_ASSERT(!bt->size_in_buffer); if (bt->size_in_chunk) { grn_memcpy(dcp, - data.sc + bt->pos_in_chunk, + chunk_data->data_start + bt->pos_in_chunk, bt->size_in_chunk); bt->pos_in_chunk = (uint32_t)(dcp - dc); dcp += bt->size_in_chunk; @@ -3829,19 +3852,19 @@ buffer_merge(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h, continue; } - data.spos = 0; + data.position = 0; buffer_data->next_position = bt->pos_in_buffer; merger_get_next_buffer(ctx, &data); if (ctx->rc != GRN_SUCCESS) { goto exit; } - if (data.sc && bt->size_in_chunk) { - data.scp = data.sc + bt->pos_in_chunk; - data.sce = data.scp + bt->size_in_chunk; + if (chunk_data->data_start && bt->size_in_chunk) { + chunk_data->data = chunk_data->data_start + bt->pos_in_chunk; + chunk_data->data_end = chunk_data->data + bt->size_in_chunk; size_t size = S_SEGMENT * ii->n_elements; if ((bt->tid & CHUNK_SPLIT)) { int i; - GRN_B_DEC(nchunks, data.scp); + GRN_B_DEC(nchunks, chunk_data->data); if (!(cinfo = GRN_MALLOCN(chunk_info, nchunks + 1))) { grn_obj term; DEFINE_NAME(ii); @@ -3865,9 +3888,9 @@ buffer_merge(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h, goto exit; } for (i = 0; i < nchunks; i++) { - GRN_B_DEC(cinfo[i].segno, data.scp); - GRN_B_DEC(cinfo[i].size, data.scp); - GRN_B_DEC(cinfo[i].dgap, data.scp); + GRN_B_DEC(cinfo[i].segno, chunk_data->data); + GRN_B_DEC(cinfo[i].size, chunk_data->data); + GRN_B_DEC(cinfo[i].dgap, chunk_data->data); crid += cinfo[i].dgap; if (buffer_data->id.rid <= crid) { chunk_merge(ctx, &data, &cinfo[i], crid, dv, &balance); @@ -3902,13 +3925,13 @@ buffer_merge(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h, } } } - if (data.sce > data.scp) { + if (chunk_data->data_end > chunk_data->data) { int decoded_size; decoded_size = grn_p_decv(ctx, ii, bt->tid & GRN_ID_MAX, - data.scp, - data.sce - data.scp, + chunk_data->data, + chunk_data->data_end - chunk_data->data, rdv, ii->n_elements); if (decoded_size == 0) { @@ -3938,19 +3961,23 @@ buffer_merge(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h, size += decoded_size; { int j = 0; - data.sdf = rdv[j].data_size; - data.srp = rdv[j++].data; + chunk_data->n_documents = rdv[j].data_size; + chunk_data->record_id_gaps = rdv[j++].data; if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { - data.ssp = rdv[j++].data; + chunk_data->section_id_gaps = rdv[j++].data; } - data.stp = rdv[j++].data; + chunk_data->tfs = rdv[j++].data; if ((ii->header->flags & GRN_OBJ_WITH_WEIGHT)) { - data.sop = rdv[j++].data; + chunk_data->weights = rdv[j++].data; } - data.snn = rdv[j].data_size; - data.snp = rdv[j].data; - } - datavec_reset(ctx, dv, ii->n_elements, data.sdf + S_SEGMENT, size); + chunk_data->position_gaps = rdv[j].data; + chunk_data->n_position_gaps = rdv[j].data_size; + } + datavec_reset(ctx, + dv, + ii->n_elements, + chunk_data->n_documents + S_SEGMENT, + size); if (ctx->rc != GRN_SUCCESS) { if (cinfo) { GRN_FREE(cinfo); } { @@ -3967,7 +3994,7 @@ buffer_merge(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h, name_size, name, (int)GRN_TEXT_LEN(&term), GRN_TEXT_VALUE(&term), bt->tid, - (size_t)(data.sdf + S_SEGMENT), + (size_t)(chunk_data->n_documents + S_SEGMENT), size); GRN_OBJ_FIN(ctx, &term); } @@ -4073,11 +4100,19 @@ buffer_merge(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h, } if ((ii->header->flags & GRN_OBJ_WITH_POSITION)) { uint32_t np = data.dest.position_gaps - dv[ii->n_elements - 1].data; - uint32_t f_p = ((np < 32) || (np <= (data.spos >> 13))) ? 0 : USE_P_ENC; + uint32_t f_p; + if ((np < 32) || (np <= (data.position >> 13))) { + f_p = 0; + } else { + f_p = USE_P_ENC; + } dv[j].data_size = np; dv[j].flags = f_p|ODD; } dcp0 = dcp; - a[1] = (bt->size_in_chunk ? a[1] : 0) + (ndf - data.sdf) + balance; + a[1] = + (bt->size_in_chunk ? a[1] : 0) + + (ndf - chunk_data->n_documents) + + balance; if (nvchunks) { int i; GRN_B_ENC(nvchunks, dcp); @@ -4363,6 +4398,7 @@ grn_ii_buffer_check(grn_ctx *ctx, grn_ii *ii, uint32_t seg) chunk_info *cinfo = NULL; grn_id crid = GRN_ID_NIL; merger_data data = {0}; + merger_chunk_data *chunk_data = &(data.source.chunk); if (!bt->tid && !bt->pos_in_buffer && !bt->size_in_buffer) { nterms_void++; @@ -4389,47 +4425,53 @@ grn_ii_buffer_check(grn_ctx *ctx, grn_ii *ii, uint32_t seg) data.source.buffer.term = bt; data.source.buffer.buffer = sb; data.source.buffer.next_position = bt->pos_in_buffer; - data.sc = sc; + chunk_data->data_start = sc; merger_get_next_buffer(ctx, &data); if (sc && bt->size_in_chunk) { - uint8_t *scp = sc + bt->pos_in_chunk; - uint8_t *sce = scp + bt->size_in_chunk; size_t size = S_SEGMENT * ii->n_elements; + chunk_data->data = chunk_data->data_start + bt->pos_in_chunk; + chunk_data->data_end = chunk_data->data + bt->size_in_chunk; if ((bt->tid & CHUNK_SPLIT)) { int i; - GRN_B_DEC(nchunks, scp); + GRN_B_DEC(nchunks, chunk_data->data); if (!(cinfo = GRN_MALLOCN(chunk_info, nchunks + 1))) { datavec_fin(ctx, rdv); GRN_OBJ_FIN(ctx, &buf); return; } for (i = 0; i < nchunks; i++) { - GRN_B_DEC(cinfo[i].segno, scp); - GRN_B_DEC(cinfo[i].size, scp); - GRN_B_DEC(cinfo[i].dgap, scp); + GRN_B_DEC(cinfo[i].segno, chunk_data->data); + GRN_B_DEC(cinfo[i].size, chunk_data->data); + GRN_B_DEC(cinfo[i].dgap, chunk_data->data); crid += cinfo[i].dgap; } } - if (sce > scp) { - size += grn_p_decv(ctx, ii, crid, scp, sce - scp, rdv, ii->n_elements); + if (chunk_data->data_end > chunk_data->data) { + size += grn_p_decv(ctx, + ii, + crid, + chunk_data->data, + chunk_data->data_end - chunk_data->data, + rdv, + ii->n_elements); { int j = 0; - data.sdf = rdv[j].data_size; - GRN_OUTPUT_INT64(data.sdf); - data.srp = rdv[j++].data; + chunk_data->n_documents = rdv[j].data_size; + GRN_OUTPUT_INT64(chunk_data->n_documents); + chunk_data->record_id_gaps = rdv[j++].data; if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { - data.ssp = rdv[j++].data; + chunk_data->section_id_gaps = rdv[j++].data; } - if (data.sdf != rdv[j].data_size) { + if (chunk_data->n_documents != rdv[j].data_size) { nterms_with_corrupt_chunk++; } - data.stp = rdv[j++].data; + chunk_data->tfs = rdv[j++].data; if ((ii->header->flags & GRN_OBJ_WITH_WEIGHT)) { - data.sop = rdv[j++].data; + chunk_data->weights = rdv[j++].data; } GRN_OUTPUT_INT64(rdv[j].data_size); - data.snn = rdv[j].data_size; - data.snp = rdv[j].data; + chunk_data->position_gaps = rdv[j].data; + chunk_data->n_position_gaps = rdv[j].data_size; } nterm_with_chunk++; } -------------- next part -------------- An HTML attachment was scrubbed... URL: <https://lists.osdn.me/mailman/archives/groonga-commit/attachments/20190214/fdf21ef4/attachment-0001.html>
