Susumu Yata
null+****@clear*****
Mon Feb 22 15:43:46 JST 2016
Susumu Yata 2016-02-22 15:43:46 +0900 (Mon, 22 Feb 2016) New Revision: 1c5b275dd1a085cbe5d0c83218ee1184c1eb6553 https://github.com/groonga/groonga/commit/1c5b275dd1a085cbe5d0c83218ee1184c1eb6553 Message: ii: add comments Modified files: lib/ii.c Modified: lib/ii.c (+31 -3) =================================================================== --- lib/ii.c 2016-02-23 01:19:45 +0900 (78594c2) +++ lib/ii.c 2016-02-22 15:43:46 +0900 (c31ea31) @@ -2931,6 +2931,7 @@ buffer_merge(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h, GETNEXTC(); MERGE_BC(1); if (rc) { + // FIXME: cinfo may not be freed. datavec_fin(ctx, dv); datavec_fin(ctx, rdv); return rc; @@ -7624,6 +7625,7 @@ struct _grn_ii_buffer { size_t total_chunk_size; }; +/* block_new returns a new ii_buffer_block to store block information. */ static ii_buffer_block * block_new(grn_ctx *ctx, grn_ii_buffer *ii_buffer) { @@ -7645,6 +7647,7 @@ block_new(grn_ctx *ctx, grn_ii_buffer *ii_buffer) return block; } +/* allocate_outbuf allocates memory to flush a block. */ static uint8_t * allocate_outbuf(grn_ctx *ctx, grn_ii_buffer *ii_buffer) { @@ -7834,6 +7837,11 @@ encode_last_tf(grn_ctx *ctx, grn_ii_buffer *ii_buffer, uint8_t *outbuf) } } +/* + * grn_ii_buffer_flush flushes the current block (ii_buffer->buffer, counters + * and tmp_lexicon) to a temporary file (ii_buffer->tmpfd). + * Also, block information is stored into ii_buffer->blocks. + */ static void grn_ii_buffer_flush(grn_ctx *ctx, grn_ii_buffer *ii_buffer) { @@ -7853,6 +7861,7 @@ grn_ii_buffer_flush(grn_ctx *ctx, grn_ii_buffer *ii_buffer) ERR(GRN_INPUT_OUTPUT_ERROR, "write returned %" GRN_FMT_LLD " != %" GRN_FMT_LLU, (long long int)r, (unsigned long long int)encsize); + // FIXME: outbuf may not be freed. return; } ii_buffer->filepos += r; @@ -7872,6 +7881,12 @@ grn_ii_buffer_flush(grn_ctx *ctx, grn_ii_buffer *ii_buffer) const uint32_t PAT_CACHE_SIZE = 1<<20; +/* + * get_tmp_lexicon returns a temporary lexicon. + * + * Note that a lexicon is created for each block and ii_buffer->tmp_lexicon is + * closed in grn_ii_buffer_flush. + */ static grn_obj * get_tmp_lexicon(grn_ctx *ctx, grn_ii_buffer *ii_buffer) { @@ -7903,6 +7918,7 @@ get_tmp_lexicon(grn_ctx *ctx, grn_ii_buffer *ii_buffer) return tmp_lexicon; } +/* get_buffer_counter returns a counter associated with tid. */ static ii_buffer_counter * get_buffer_counter(grn_ctx *ctx, grn_ii_buffer *ii_buffer, grn_obj *tmp_lexicon, grn_id tid) @@ -7922,6 +7938,12 @@ get_buffer_counter(grn_ctx *ctx, grn_ii_buffer *ii_buffer, return &ii_buffer->counters[tid - 1]; } +/* + * grn_ii_buffer_tokenize_value tokenizes a value. + * + * The result is written into the current block (ii_buffer->tmp_lexicon, + * ii_buffer->block_buf, ii_buffer->counters, etc.). + */ static void grn_ii_buffer_tokenize_value(grn_ctx *ctx, grn_ii_buffer *ii_buffer, grn_id rid, const ii_buffer_value *value) @@ -7995,11 +8017,17 @@ grn_ii_buffer_tokenize_value(grn_ctx *ctx, grn_ii_buffer *ii_buffer, } } +/* + * grn_ii_buffer_tokenize tokenizes ii_buffer->values. + * + * grn_ii_buffer_tokenize estimates the size of tokenized values. + * If the remaining space of the current block is not enough to store the new + * tokenized values, the current block is flushed. + * Then, grn_ii_buffer_tokenize tokenizes values. + */ static void grn_ii_buffer_tokenize(grn_ctx *ctx, grn_ii_buffer *ii_buffer, grn_id rid) { - // Estimate the size of tokenized values and resize the internal buffer if - // the buffer size is not enough. unsigned int i; uint32_t est_len = 0; for (i = 0; i < ii_buffer->nvalues; i++) { @@ -8016,7 +8044,7 @@ grn_ii_buffer_tokenize(grn_ctx *ctx, grn_ii_buffer *ii_buffer, grn_id rid) ii_buffer->block_buf_size = est_len; } } - // Tokenize values. + for (i = 0; i < ii_buffer->nvalues; i++) { const ii_buffer_value *value = &ii_buffer->values[i]; if (value->len) { -------------- next part -------------- HTML����������������������������...다운로드
