Kouhei Sutou 2019-04-01 17:34:47 +0900 (Mon, 01 Apr 2019) Revision: e8c415cdd1d0e41d4716f91ccb12a0fa76c6e6de https://github.com/groonga/groonga/commit/e8c415cdd1d0e41d4716f91ccb12a0fa76c6e6de Message: ii: add support for GRN_OBJ_INDEX_LARGE TODO: * TEST!!! * Document Added files: test/command/suite/dump/schema/column/index/index_large.expected test/command/suite/dump/schema/column/index/index_large.test test/command/suite/schema/tables/columns/type/index_large.expected test/command/suite/schema/tables/columns/type/index_large.test Modified files: include/groonga/groonga.h lib/dump.c lib/grn_ii.h lib/ii.c lib/index_column.c lib/proc.c lib/proc/proc_column.c lib/proc/proc_object_inspect.c Modified: include/groonga/groonga.h (+1 -0) =================================================================== --- include/groonga/groonga.h 2019-03-29 10:45:34 +0900 (db54cc9d4) +++ include/groonga/groonga.h 2019-04-01 17:34:47 +0900 (a07419f59) @@ -359,6 +359,7 @@ typedef uint32_t grn_column_flags; #define GRN_OBJ_INDEX_SMALL (0x01<<16) #define GRN_OBJ_INDEX_MEDIUM (0x01<<17) +#define GRN_OBJ_INDEX_LARGE (0x01<<18) /* obj types */ Modified: lib/dump.c (+4 -0) =================================================================== --- lib/dump.c 2019-03-29 10:45:34 +0900 (0c6eb6801) +++ lib/dump.c 2019-04-01 17:34:47 +0900 (084b95f99) @@ -1,6 +1,7 @@ /* -*- c-basic-offset: 2 -*- */ /* Copyright(C) 2016 Brazil + Copyright(C) 2019 Kouhei Sutou <kou****@clear*****> This library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public @@ -89,6 +90,9 @@ grn_dump_column_create_flags(grn_ctx *ctx, if (flags & GRN_OBJ_INDEX_MEDIUM) { GRN_TEXT_PUTS(ctx, buffer, "|INDEX_MEDIUM"); } + if (flags & GRN_OBJ_INDEX_LARGE) { + GRN_TEXT_PUTS(ctx, buffer, "|INDEX_LARGE"); + } break; } switch (flags & GRN_OBJ_COMPRESS_MASK) { Modified: lib/grn_ii.h (+46 -20) =================================================================== --- lib/grn_ii.h 2019-03-29 10:45:34 +0900 (c83704404) +++ lib/grn_ii.h 2019-04-01 17:34:47 +0900 (3004913b7) @@ -1,6 +1,7 @@ /* -*- c-basic-offset: 2 -*- */ /* Copyright(C) 2009-2018 Brazil + Copyright(C) 2019 Kouhei Sutou <kou****@clear*****> This library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public @@ -29,6 +30,10 @@ extern "C" { #endif +typedef struct _grn_ii_header_common grn_ii_header_common; +typedef struct _grn_ii_header_normal grn_ii_header_normal; +typedef struct _grn_ii_header_large grn_ii_header_large; + struct _grn_ii { grn_db_obj obj; grn_io *seg; /* I/O for a variety of segments */ @@ -39,7 +44,11 @@ struct _grn_ii { /* This member is used for matching */ uint32_t n_elements; /* Number of elements in postings */ /* rid, [sid], tf, [weight] and [pos] */ - struct grn_ii_header *header; + union { + grn_ii_header_common *common; + grn_ii_header_normal *normal; + grn_ii_header_large *large; + } header; }; /* BGQ is buffer garbage queue? */ @@ -62,25 +71,38 @@ struct _grn_ii { #define GRN_II_PSEG_NOT_ASSIGNED 0xffffffff -struct grn_ii_header { - uint64_t total_chunk_size; - uint64_t bmax; - uint32_t flags; - uint32_t amax; - uint32_t smax; - uint32_t param1; - uint32_t param2; - uint32_t pnext; - uint32_t bgqhead; - uint32_t bgqtail; - uint32_t bgqbody[GRN_II_BGQSIZE]; - uint32_t reserved[288]; - uint32_t ainfo[GRN_II_MAX_LSEG]; /* array info */ - uint32_t binfo[GRN_II_MAX_LSEG]; /* buffer info */ - uint32_t free_chunks[GRN_II_N_CHUNK_VARIATION + 1]; - uint32_t garbages[GRN_II_N_CHUNK_VARIATION + 1]; - uint32_t ngarbages[GRN_II_N_CHUNK_VARIATION + 1]; - uint8_t chunks[GRN_II_MAX_CHUNK >> 3]; +#define GRN_II_HEADER_COMMON_FIELDS \ + uint64_t total_chunk_size; \ + uint64_t bmax; \ + uint32_t flags; \ + uint32_t amax; \ + uint32_t smax; \ + uint32_t param1; \ + uint32_t param2; \ + uint32_t pnext; \ + uint32_t bgqhead; \ + uint32_t bgqtail; \ + uint32_t bgqbody[GRN_II_BGQSIZE]; \ + uint32_t reserved[288]; \ + uint32_t ainfo[GRN_II_MAX_LSEG]; /* array info */ \ + uint32_t binfo[GRN_II_MAX_LSEG]; /* buffer info */ \ + uint32_t free_chunks[GRN_II_N_CHUNK_VARIATION + 1]; \ + uint32_t garbages[GRN_II_N_CHUNK_VARIATION + 1]; \ + uint32_t ngarbages[GRN_II_N_CHUNK_VARIATION + 1]; \ + uint8_t chunks[GRN_II_MAX_CHUNK >> 3] + +struct _grn_ii_header_common { + GRN_II_HEADER_COMMON_FIELDS; +}; + +struct _grn_ii_header_normal { + GRN_II_HEADER_COMMON_FIELDS; +}; + +struct _grn_ii_header_large { + GRN_II_HEADER_COMMON_FIELDS; + uint32_t ainfo_extend[GRN_II_MAX_LSEG]; /* array info (extended) */ + uint32_t binfo_extend[GRN_II_MAX_LSEG]; /* buffer info (extended) */ }; struct _grn_ii_pos { @@ -126,6 +148,10 @@ size_t grn_ii_get_disk_usage(grn_ctx *ctx, grn_ii *ii); grn_ii_cursor *grn_ii_cursor_openv1(grn_ii *ii, uint32_t key); grn_rc grn_ii_cursor_openv2(grn_ii_cursor **cursors, int ncursors); +uint32_t grn_ii_get_array_pseg(grn_ii *ii, uint32_t lseg); +uint32_t grn_ii_get_buffer_pseg(grn_ii *ii, uint32_t pseg); +uint32_t grn_ii_n_logical_segments(grn_ii *ii); + uint32_t grn_ii_max_section(grn_ii *ii); const char *grn_ii_path(grn_ii *ii); Modified: lib/ii.c (+299 -205) =================================================================== --- lib/ii.c 2019-03-29 10:45:34 +0900 (24b4e5582) +++ lib/ii.c 2019-04-01 17:34:47 +0900 (1383b05ac) @@ -46,6 +46,7 @@ # include "grn_onigmo.h" #endif /* GRN_II_SELECT_ENABLE_SEQUENTIAL_SEARCH */ +/* P is for physical? */ #define MAX_PSEG 0x20000 #define MAX_PSEG_SMALL 0x00200 /* MAX_PSEG_MEDIUM has enough space for the following source: @@ -55,6 +56,7 @@ * * Lexicon doesn't have tokenizer. */ #define MAX_PSEG_MEDIUM 0x10000 +#define MAX_PSEG_LARGE 0x40000 #define S_CHUNK (1 << GRN_II_W_CHUNK) #define W_SEGMENT 18 #define S_SEGMENT (1 << W_SEGMENT) @@ -85,6 +87,7 @@ } \ } while (GRN_FALSE) +/* L is for logical? */ #define LSEG(pos) ((pos) >> 16) #define LPOS(pos) (((pos) & 0xffff) << 2) #define SEG2POS(seg,pos) ((((uint32_t)(seg)) << 16) + (((uint32_t)(pos)) >> 2)) @@ -266,36 +269,96 @@ grn_ii_get_term(grn_ctx *ctx, grn_ii *ii, grn_id tid, grn_obj *term) /* segment */ grn_inline static uint32_t +grn_ii_get_array_pseg_inline(grn_ii *ii, + uint32_t lseg) +{ + if (lseg < GRN_II_MAX_LSEG) { + return ii->header.common->ainfo[lseg]; + } else { + return ii->header.large->ainfo_extend[lseg - GRN_II_MAX_LSEG]; + } +} + +grn_inline static void +grn_ii_set_array_pseg_inline(grn_ii *ii, + uint32_t lseg, + uint32_t pseg) +{ + if (lseg < GRN_II_MAX_LSEG) { + ii->header.common->ainfo[lseg] = pseg; + } else { + ii->header.large->ainfo[lseg - GRN_II_MAX_LSEG] = pseg; + } +} + +grn_inline static uint32_t +grn_ii_get_buffer_pseg_inline(grn_ii *ii, + uint32_t lseg) +{ + if (lseg < GRN_II_MAX_LSEG) { + return ii->header.common->binfo[lseg]; + } else { + return ii->header.large->binfo[lseg - GRN_II_MAX_LSEG]; + } +} + +grn_inline static void +grn_ii_set_buffer_pseg_inline(grn_ii *ii, + uint32_t lseg, + uint32_t pseg) +{ + if (lseg < GRN_II_MAX_LSEG) { + ii->header.common->binfo[lseg] = pseg; + } else { + ii->header.large->binfo[lseg - GRN_II_MAX_LSEG] = pseg; + } +} + +grn_inline static uint32_t +grn_ii_n_logical_segments_inline(grn_ii *ii) +{ + if (ii->header.common->flags & GRN_OBJ_INDEX_LARGE) { + return GRN_II_MAX_LSEG * 2; + } else { + return GRN_II_MAX_LSEG; + } +} + +grn_inline static uint32_t segment_get(grn_ctx *ctx, grn_ii *ii) { + grn_ii_header_common *header = ii->header.common; uint32_t pseg; - if (ii->header->bgqtail == ((ii->header->bgqhead + 1) & (GRN_II_BGQSIZE - 1))) { - pseg = ii->header->bgqbody[ii->header->bgqtail]; - ii->header->bgqtail = (ii->header->bgqtail + 1) & (GRN_II_BGQSIZE - 1); + if (header->bgqtail == ((header->bgqhead + 1) & (GRN_II_BGQSIZE - 1))) { + pseg = header->bgqbody[header->bgqtail]; + header->bgqtail = (header->bgqtail + 1) & (GRN_II_BGQSIZE - 1); } else { - pseg = ii->header->pnext; + pseg = header->pnext; if (!pseg) { int i; uint32_t pmax = 0; char *used; - uint32_t max_segment = ii->seg->header->max_segment; + const uint32_t max_segment = ii->seg->header->max_segment; used = GRN_CALLOC(max_segment); if (!used) { return max_segment; } - for (i = 0; i < GRN_II_MAX_LSEG && i < max_segment; i++) { - if ((pseg = ii->header->ainfo[i]) != GRN_II_PSEG_NOT_ASSIGNED) { + const uint32_t n_logical_segments = grn_ii_n_logical_segments_inline(ii); + for (i = 0; i < n_logical_segments; i++) { + pseg = grn_ii_get_array_pseg_inline(ii, i); + if (pseg != GRN_II_PSEG_NOT_ASSIGNED) { if (pseg > pmax) { pmax = pseg; } used[pseg] = 1; } - if ((pseg = ii->header->binfo[i]) != GRN_II_PSEG_NOT_ASSIGNED) { + pseg = grn_ii_get_buffer_pseg_inline(ii, i); + if (pseg != GRN_II_PSEG_NOT_ASSIGNED) { if (pseg > pmax) { pmax = pseg; } used[pseg] = 1; } } for (pseg = 0; pseg < max_segment && used[pseg]; pseg++) ; GRN_FREE(used); - ii->header->pnext = pmax + 1; - } else if (ii->header->pnext < ii->seg->header->max_segment) { - ii->header->pnext++; + header->pnext = pmax + 1; + } else if (ii->header.common->pnext < ii->seg->header->max_segment) { + header->pnext++; } } return pseg; @@ -322,22 +385,29 @@ grn_inline static grn_rc buffer_segment_new(grn_ctx *ctx, grn_ii *ii, uint32_t *segno) { uint32_t lseg, pseg; - if (*segno < GRN_II_MAX_LSEG) { - if (ii->header->binfo[*segno] != GRN_II_PSEG_NOT_ASSIGNED) { - return GRN_INVALID_ARGUMENT; + if (*segno == GRN_II_PSEG_NOT_ASSIGNED) { + const uint32_t n_logical_segments = grn_ii_n_logical_segments_inline(ii); + for (lseg = 0; lseg < n_logical_segments; lseg++) { + if (grn_ii_get_buffer_pseg_inline(ii, lseg) == GRN_II_PSEG_NOT_ASSIGNED) { + break; + } } - lseg = *segno; - } else { - for (lseg = 0; lseg < GRN_II_MAX_LSEG; lseg++) { - if (ii->header->binfo[lseg] == GRN_II_PSEG_NOT_ASSIGNED) { break; } + if (lseg == n_logical_segments) { + return GRN_NO_MEMORY_AVAILABLE; } - if (lseg == GRN_II_MAX_LSEG) { return GRN_NO_MEMORY_AVAILABLE; } *segno = lseg; + } else { + if (grn_ii_get_buffer_pseg_inline(ii, *segno) != GRN_II_PSEG_NOT_ASSIGNED) { + return GRN_INVALID_ARGUMENT; + } + lseg = *segno; } pseg = segment_get(ctx, ii); if (pseg < ii->seg->header->max_segment) { - ii->header->binfo[lseg] = pseg; - if (lseg >= ii->header->bmax) { ii->header->bmax = lseg + 1; } + grn_ii_set_buffer_pseg_inline(ii, lseg, pseg); + if (lseg >= ii->header.common->bmax) { + ii->header.common->bmax = lseg + 1; + } return GRN_SUCCESS; } else { return GRN_NO_MEMORY_AVAILABLE; @@ -350,30 +420,35 @@ buffer_segment_reserve(grn_ctx *ctx, grn_ii *ii, uint32_t *lseg1, uint32_t *pseg1) { uint32_t i = 0; + const uint32_t n_logical_segments = grn_ii_n_logical_segments_inline(ii); for (;; i++) { - if (i == GRN_II_MAX_LSEG) { + if (i == n_logical_segments) { DEFINE_NAME(ii); MERR("[ii][buffer][segment][reserve] " "couldn't find a free buffer: <%.*s>: max:<%u>", name_size, name, - GRN_II_MAX_LSEG); + n_logical_segments); return ctx->rc; } - if (ii->header->binfo[i] == GRN_II_PSEG_NOT_ASSIGNED) { break; } + if (grn_ii_get_buffer_pseg_inline(ii, i) == GRN_II_PSEG_NOT_ASSIGNED) { + break; + } } *lseg0 = i++; for (;; i++) { - if (i == GRN_II_MAX_LSEG) { + if (i == n_logical_segments) { DEFINE_NAME(ii); MERR("[ii][buffer][segment][reserve] " "couldn't find two free buffers: " "<%.*s>: " "found:<%u>, max:<%u>", name_size, name, - *lseg0, GRN_II_MAX_LSEG); + *lseg0, n_logical_segments); return ctx->rc; } - if (ii->header->binfo[i] == GRN_II_PSEG_NOT_ASSIGNED) { break; } + if (grn_ii_get_buffer_pseg_inline(ii, i) == GRN_II_PSEG_NOT_ASSIGNED) { + break; + } } *lseg1 = i; if ((*pseg0 = segment_get(ctx, ii)) == ii->seg->header->max_segment) { @@ -395,76 +470,50 @@ buffer_segment_reserve(grn_ctx *ctx, grn_ii *ii, *lseg0, *lseg1, ii->seg->header->max_segment); return ctx->rc; } - /* - { - uint32_t pseg; - char *used = GRN_CALLOC(ii->seg->header->max_segment); - if (!used) { return GRN_NO_MEMORY_AVAILABLE; } - for (i = 0; i < GRN_II_MAX_LSEG; i++) { - if ((pseg = ii->header->ainfo[i]) != GRN_II_PSEG_NOT_ASSIGNED) { - used[pseg] = 1; - } - if ((pseg = ii->header->binfo[i]) != GRN_II_PSEG_NOT_ASSIGNED) { - used[pseg] = 1; - } - } - for (pseg = 0;; pseg++) { - if (pseg == ii->seg->header->max_segment) { - GRN_FREE(used); - return GRN_NO_MEMORY_AVAILABLE; - } - if (!used[pseg]) { break; } - } - *pseg0 = pseg++; - for (;; pseg++) { - if (pseg == ii->seg->header->max_segment) { - GRN_FREE(used); - return GRN_NO_MEMORY_AVAILABLE; - } - if (!used[pseg]) { break; } - } - *pseg1 = pseg; - GRN_FREE(used); - } - */ return ctx->rc; } -#define BGQENQUE(lseg) do {\ - if (ii->header->binfo[lseg] != GRN_II_PSEG_NOT_ASSIGNED) {\ - ii->header->bgqbody[ii->header->bgqhead] = ii->header->binfo[lseg];\ - ii->header->bgqhead = (ii->header->bgqhead + 1) & (GRN_II_BGQSIZE - 1);\ - GRN_ASSERT(ii->header->bgqhead != ii->header->bgqtail);\ - }\ -} while (0) +grn_inline static void +buffer_garbage_enqueue(grn_ii *ii, uint32_t lseg) +{ + const uint32_t pseg = grn_ii_get_buffer_pseg_inline(ii, lseg); + if (pseg != GRN_II_PSEG_NOT_ASSIGNED) { + grn_ii_header_common *header = ii->header.common; + header->bgqbody[header->bgqhead] = header->binfo[lseg]; + header->bgqhead = (header->bgqhead + 1) & (GRN_II_BGQSIZE - 1); + GRN_ASSERT(header->bgqhead != header->bgqtail); + } +} grn_inline static void buffer_segment_update(grn_ii *ii, uint32_t lseg, uint32_t pseg) { - BGQENQUE(lseg); + buffer_garbage_enqueue(ii, lseg); // smb_wmb(); - ii->header->binfo[lseg] = pseg; - if (lseg >= ii->header->bmax) { ii->header->bmax = lseg + 1; } + grn_ii_set_buffer_pseg_inline(ii, lseg, pseg); + if (lseg >= ii->header.common->bmax) { + ii->header.common->bmax = lseg + 1; + } } grn_inline static void buffer_segment_clear(grn_ii *ii, uint32_t lseg) { - BGQENQUE(lseg); + buffer_garbage_enqueue(ii, lseg); // smb_wmb(); - ii->header->binfo[lseg] = GRN_II_PSEG_NOT_ASSIGNED; + grn_ii_set_buffer_pseg_inline(ii, lseg, GRN_II_PSEG_NOT_ASSIGNED); } /* chunk */ #define HEADER_CHUNK_AT(ii,offset) \ - ((((ii)->header->chunks[((offset) >> 3)]) >> ((offset) & 7)) & 1) + ((((ii)->header.common->chunks[((offset) >> 3)]) >> ((offset) & 7)) & 1) #define HEADER_CHUNK_ON(ii,offset) \ - (((ii)->header->chunks[((offset) >> 3)]) |= (1 << ((offset) & 7))) + (((ii)->header.common->chunks[((offset) >> 3)]) |= (1 << ((offset) & 7))) #define HEADER_CHUNK_OFF(ii,offset) \ - (((ii)->header->chunks[((offset) >> 3)]) &= ~(1 << ((offset) & 7))) + (((ii)->header.common->chunks[((offset) >> 3)]) &= ~(1 << ((offset) & 7))) #define N_GARBAGES_TH 1 @@ -537,12 +586,13 @@ chunk_new(grn_ctx *ctx, grn_ii *ii, uint32_t *res, uint32_t size) m = GRN_II_W_LEAST_CHUNK; } aligned_size = 1 << (m - GRN_II_W_LEAST_CHUNK); - if (ii->header->ngarbages[m - GRN_II_W_LEAST_CHUNK] > N_GARBAGES_TH) { + grn_ii_header_common *header = ii->header.common; + if (header->ngarbages[m - GRN_II_W_LEAST_CHUNK] > N_GARBAGES_TH) { grn_ii_ginfo *ginfo; uint32_t *gseg; grn_io_win iw, iw_; iw_.addr = NULL; - gseg = &ii->header->garbages[m - GRN_II_W_LEAST_CHUNK]; + gseg = &header->garbages[m - GRN_II_W_LEAST_CHUNK]; while (*gseg != GRN_II_PSEG_NOT_ASSIGNED) { ginfo = WIN_MAP(ii->chunk, ctx, &iw, *gseg, 0, S_GARBAGE, grn_io_rdwr); //GRN_IO_SEG_MAP2(ii->chunk, *gseg, ginfo); @@ -554,7 +604,7 @@ chunk_new(grn_ctx *ctx, grn_ii *ii, uint32_t *res, uint32_t size) "<%.*s>: " "n-garbages:<%u>, size:<%u>, n-chunks:<%u>", name_size, name, - ii->header->ngarbages[m - GRN_II_W_LEAST_CHUNK], + header->ngarbages[m - GRN_II_W_LEAST_CHUNK], size, n_chunks); } @@ -565,7 +615,7 @@ chunk_new(grn_ctx *ctx, grn_ii *ii, uint32_t *res, uint32_t size) *res = ginfo->recs[ginfo->tail]; if (++ginfo->tail == N_GARBAGES) { ginfo->tail = 0; } ginfo->nrecs--; - ii->header->ngarbages[m - GRN_II_W_LEAST_CHUNK]--; + header->ngarbages[m - GRN_II_W_LEAST_CHUNK]--; if (!ginfo->nrecs) { HEADER_CHUNK_OFF(ii, *gseg); *gseg = ginfo->next; @@ -580,7 +630,7 @@ chunk_new(grn_ctx *ctx, grn_ii *ii, uint32_t *res, uint32_t size) } if (iw_.addr) { grn_io_win_unmap(&iw_); } } - vp = &ii->header->free_chunks[m - GRN_II_W_LEAST_CHUNK]; + vp = &header->free_chunks[m - GRN_II_W_LEAST_CHUNK]; if (*vp == GRN_II_PSEG_NOT_ASSIGNED) { int i = 0; while (HEADER_CHUNK_AT(ii, i)) { @@ -636,7 +686,7 @@ chunk_free(grn_ctx *ctx, grn_ii *ii, } else { m = GRN_II_W_LEAST_CHUNK; } - gseg = &ii->header->garbages[m - GRN_II_W_LEAST_CHUNK]; + gseg = &ii->header.common->garbages[m - GRN_II_W_LEAST_CHUNK]; iw_.addr = NULL; while (*gseg != GRN_II_PSEG_NOT_ASSIGNED) { ginfo = WIN_MAP(ii->chunk, ctx, &iw, *gseg, 0, S_GARBAGE, grn_io_rdwr); @@ -682,7 +732,7 @@ chunk_free(grn_ctx *ctx, grn_ii *ii, if (++ginfo->head == N_GARBAGES) { ginfo->head = 0; } ginfo->nrecs++; grn_io_win_unmap(&iw); - ii->header->ngarbages[m - GRN_II_W_LEAST_CHUNK]++; + ii->header.common->ngarbages[m - GRN_II_W_LEAST_CHUNK]++; return GRN_SUCCESS; } @@ -2277,10 +2327,10 @@ typedef struct { grn_inline static uint32_t buffer_open(grn_ctx *ctx, grn_ii *ii, uint32_t pos, buffer_term **bt, buffer **b) { - byte *p = NULL; - uint16_t lseg = (uint16_t) (LSEG(pos)); - uint32_t pseg = ii->header->binfo[lseg]; + const uint16_t lseg = (uint16_t) (LSEG(pos)); + const uint32_t pseg = grn_ii_get_buffer_pseg_inline(ii, lseg); if (pseg != GRN_II_PSEG_NOT_ASSIGNED) { + byte *p = NULL; GRN_IO_SEG_REF(ii->seg, pseg, p); if (!p) { return GRN_II_PSEG_NOT_ASSIGNED; } if (b) { *b = (buffer *)p; } @@ -2307,7 +2357,7 @@ buffer_open_if_capable(grn_ctx *ctx, grn_ii *ii, int32_t seg, int size, buffer * if ((pseg = buffer_open(ctx, ii, pos, NULL, b)) != GRN_II_PSEG_NOT_ASSIGNED) { uint16_t nterms = (*b)->header.nterms - (*b)->header.nterms_void; if (!((nterms < 4096 || - (ii->header->total_chunk_size >> ((nterms >> 8) - 6)) + (ii->header.common->total_chunk_size >> ((nterms >> 8) - 6)) > (*b)->header.chunk_size) && ((*b)->header.buffer_free >= size + sizeof(buffer_term)))) { buffer_close(ctx, ii, pseg); @@ -2349,7 +2399,7 @@ buffer_term_dump(grn_ctx *ctx, grn_ii *ii, buffer *b, buffer_term *bt) r = BUFFER_REC_AT(b, pos); p = GRN_NEXT_ADDR(r); GRN_B_DEC(rid, p); - if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { GRN_B_DEC(sid, p); } else { sid = 1; @@ -2369,7 +2419,7 @@ check_jump(grn_ctx *ctx, grn_ii *ii, buffer *b, buffer_rec *r, int j) if (!j) { return GRN_SUCCESS; } p = GRN_NEXT_ADDR(r); GRN_B_DEC(id.rid, p); - if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { GRN_B_DEC(id.sid, p); } else { id.sid = 1; @@ -2381,7 +2431,7 @@ check_jump(grn_ctx *ctx, grn_ii *ii, buffer *b, buffer_rec *r, int j) r2 = BUFFER_REC_AT(b, j); p = GRN_NEXT_ADDR(r2); GRN_B_DEC(id2.rid, p); - if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { GRN_B_DEC(id2.sid, p); } else { id2.sid = 1; @@ -2468,7 +2518,7 @@ buffer_put(grn_ctx *ctx, grn_ii *ii, buffer *b, buffer_term *bt, r_curr = BUFFER_REC_AT(b, *lastp); p = GRN_NEXT_ADDR(r_curr); GRN_B_DEC(id_curr.rid, p); - if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { GRN_B_DEC(id_curr.sid, p); } else { id_curr.sid = 1; @@ -2502,7 +2552,7 @@ buffer_put(grn_ctx *ctx, grn_ii *ii, buffer *b, buffer_term *bt, r_curr = BUFFER_REC_AT(b, step); p = GRN_NEXT_ADDR(r_curr); GRN_B_DEC(id_curr.rid, p); - if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { GRN_B_DEC(id_curr.sid, p); } else { id_curr.sid = 1; @@ -2554,7 +2604,7 @@ buffer_put(grn_ctx *ctx, grn_ii *ii, buffer *b, buffer_term *bt, docid idj; p = GRN_NEXT_ADDR(rj); GRN_B_DEC(idj.rid, p); - if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { GRN_B_DEC(idj.sid, p); } else { idj.sid = 1; @@ -2577,13 +2627,13 @@ buffer_put(grn_ctx *ctx, grn_ii *ii, buffer *b, buffer_term *bt, grn_inline static uint32_t * array_at(grn_ctx *ctx, grn_ii *ii, uint32_t id) { - byte *p = NULL; - uint32_t seg, pseg; if (id > GRN_ID_MAX) { return NULL; } - seg = id >> W_ARRAY; - if ((pseg = ii->header->ainfo[seg]) == GRN_II_PSEG_NOT_ASSIGNED) { + const uint16_t seg = id >> W_ARRAY; + const uint32_t pseg = grn_ii_get_array_pseg_inline(ii, seg); + if (pseg == GRN_II_PSEG_NOT_ASSIGNED) { return NULL; } + byte *p = NULL; GRN_IO_SEG_REF(ii->seg, pseg, p); if (!p) { return NULL; } return (uint32_t *)(p + (id & ARRAY_MASK_IN_A_SEGMENT) * S_ARRAY_ELEMENT); @@ -2593,14 +2643,15 @@ grn_inline static uint32_t * array_get(grn_ctx *ctx, grn_ii *ii, uint32_t id) { byte *p = NULL; - uint16_t seg; - uint32_t pseg; if (id > GRN_ID_MAX) { return NULL; } - seg = id >> W_ARRAY; - if ((pseg = ii->header->ainfo[seg]) == GRN_II_PSEG_NOT_ASSIGNED) { + const uint16_t seg = id >> W_ARRAY; + uint32_t pseg = grn_ii_get_array_pseg_inline(ii, seg); + if (pseg == GRN_II_PSEG_NOT_ASSIGNED) { if (segment_get_clear(ctx, ii, &pseg)) { return NULL; } - ii->header->ainfo[seg] = pseg; - if (seg >= ii->header->amax) { ii->header->amax = seg + 1; } + grn_ii_set_array_pseg_inline(ii, seg, pseg); + if (seg >= ii->header.common->amax) { + ii->header.common->amax = seg + 1; + } } GRN_IO_SEG_REF(ii->seg, pseg, p); if (!p) { return NULL; } @@ -2610,7 +2661,8 @@ array_get(grn_ctx *ctx, grn_ii *ii, uint32_t id) grn_inline static void array_unref(grn_ii *ii, uint32_t id) { - GRN_IO_SEG_UNREF(ii->seg, ii->header->ainfo[id >> W_ARRAY]); + const uint32_t pseg = grn_ii_get_array_pseg_inline(ii, id >> W_ARRAY); + GRN_IO_SEG_UNREF(ii->seg, pseg); } /* updspec */ @@ -2702,14 +2754,16 @@ encode_rec(grn_ctx *ctx, grn_ii *ii, grn_ii_updspec *u, unsigned int *size, int } p = br; GRN_B_ENC(u->rid, p); - if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { GRN_B_ENC(u->sid, p); } else { u->sid = 1; } GRN_B_ENC(tf, p); - if ((ii->header->flags & GRN_OBJ_WITH_WEIGHT)) { GRN_B_ENC(weight, p); } - if ((ii->header->flags & GRN_OBJ_WITH_POSITION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_WEIGHT)) { + GRN_B_ENC(weight, p); + } + if ((ii->header.common->flags & GRN_OBJ_WITH_POSITION)) { for (lpos = 0, pp = u->pos; pp && tf--; lpos = pp->pos, pp = pp->next) { GRN_B_ENC(pp->pos - lpos, p); } @@ -2835,7 +2889,7 @@ merge_dump_source_add_entry(grn_ctx *ctx, info->sid, info->tf, info->weight); - if (data->ii->header->flags & GRN_OBJ_WITH_POSITION) { + if (data->ii->header.common->flags & GRN_OBJ_WITH_POSITION) { uint32_t i; uint32_t position = 0; @@ -2919,14 +2973,14 @@ merge_dump_source_chunk_raw(grn_ctx *ctx, int i = 0; n_documents = data->data_vector[i].data_size; record_id_gaps = data->data_vector[i++].data; - if (data->ii->header->flags & GRN_OBJ_WITH_SECTION) { + if (data->ii->header.common->flags & GRN_OBJ_WITH_SECTION) { section_id_gaps = data->data_vector[i++].data; } n_terms_list = data->data_vector[i++].data; - if (data->ii->header->flags & GRN_OBJ_WITH_WEIGHT) { + if (data->ii->header.common->flags & GRN_OBJ_WITH_WEIGHT) { weights = data->data_vector[i++].data; } - if (data->ii->header->flags & GRN_OBJ_WITH_POSITION) { + if (data->ii->header.common->flags & GRN_OBJ_WITH_POSITION) { position_gaps = data->data_vector[i++].data; } } @@ -2942,7 +2996,7 @@ merge_dump_source_chunk_raw(grn_ctx *ctx, record_id += record_id_gap; info.rid = record_id; - if (data->ii->header->flags & GRN_OBJ_WITH_SECTION) { + if (data->ii->header.common->flags & GRN_OBJ_WITH_SECTION) { if (record_id_gap > 0) { section_id = 0; } @@ -2952,7 +3006,7 @@ merge_dump_source_chunk_raw(grn_ctx *ctx, } info.sid = section_id; info.tf = 1 + n_terms_list[i]; - if (data->ii->header->flags & GRN_OBJ_WITH_WEIGHT) { + if (data->ii->header.common->flags & GRN_OBJ_WITH_WEIGHT) { info.weight = weights[i]; } else { info.weight = 0; @@ -3116,7 +3170,7 @@ merge_dump_source(grn_ctx *ctx, data.nth_chunk = 0; data.n_chunks = 0; datavec_init(ctx, data.data_vector, ii->n_elements, 0, 0); - if (ii->header->flags & GRN_OBJ_WITH_POSITION) { + if (ii->header.common->flags & GRN_OBJ_WITH_POSITION) { data.data_vector[ii->n_elements - 1].flags = ODD; } { @@ -3188,13 +3242,13 @@ merge_dump_source(grn_ctx *ctx, ':'); GRN_B_DEC(info.rid, record_data); - if (ii->header->flags & GRN_OBJ_WITH_SECTION) { + if (ii->header.common->flags & GRN_OBJ_WITH_SECTION) { GRN_B_DEC(info.sid, record_data); } else { info.sid = 1; } GRN_B_DEC(info.tf, record_data); - if (ii->header->flags & GRN_OBJ_WITH_WEIGHT) { + if (ii->header.common->flags & GRN_OBJ_WITH_WEIGHT) { GRN_B_DEC(info.weight, record_data); } else { info.weight = 0; @@ -3268,11 +3322,11 @@ merger_init_chunk_data(grn_ctx *ctx, memset(&(chunk_data->id), 0, sizeof(docinfo)); chunk_data->n_documents = rdv[j].data_size; chunk_data->record_id_gaps = rdv[j++].data; - if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { chunk_data->section_id_gaps = rdv[j++].data; } chunk_data->tfs = rdv[j++].data; - if ((ii->header->flags & GRN_OBJ_WITH_WEIGHT)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_WEIGHT)) { chunk_data->weights = rdv[j++].data; } chunk_data->position_gaps = rdv[j].data; @@ -3334,11 +3388,11 @@ merger_get_next_chunk(grn_ctx *ctx, merger_data *data) 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)) { + if ((data->ii->header.common->flags & GRN_OBJ_WITH_WEIGHT)) { chunk_data->id.weight = *chunk_data->weights; chunk_data->weights++; } - if ((data->ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((data->ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { chunk_data->id.sid += 1 + *chunk_data->section_id_gaps; chunk_data->section_id_gaps++; } else { @@ -3354,7 +3408,7 @@ merger_put_next_id(grn_ctx *ctx, merger_data *data, docinfo *id) uint32_t record_id_gap = id->rid - data->last_id.rid; *(data->dest.record_id_gaps) = record_id_gap; data->dest.record_id_gaps++; - if ((data->ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((data->ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { uint32_t section_id_gap; if (record_id_gap > 0) { section_id_gap = id->sid - 1; @@ -3366,7 +3420,7 @@ merger_put_next_id(grn_ctx *ctx, merger_data *data, docinfo *id) } *(data->dest.tfs) = id->tf - 1; data->dest.tfs++; - if ((data->ii->header->flags & GRN_OBJ_WITH_WEIGHT)) { + if ((data->ii->header.common->flags & GRN_OBJ_WITH_WEIGHT)) { *(data->dest.weights) = id->weight; data->dest.weights++; } @@ -3392,7 +3446,7 @@ merger_put_next_chunk(grn_ctx *ctx, merger_data *data) return; } merger_put_next_id(ctx, data, &(chunk_data->id)); - if ((data->ii->header->flags & GRN_OBJ_WITH_POSITION)) { + if ((data->ii->header.common->flags & GRN_OBJ_WITH_POSITION)) { uint32_t i; uint32_t rest_n_positions = (uint32_t)(chunk_data->position_gaps_end - @@ -3479,7 +3533,7 @@ merger_get_next_buffer(grn_ctx *ctx, merger_data *data) buffer_data->next_position); buffer_data->data = GRN_NEXT_ADDR(record); GRN_B_DEC(buffer_data->id.rid, buffer_data->data); - if ((data->ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((data->ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { GRN_B_DEC(buffer_data->id.sid, buffer_data->data); } else { buffer_data->id.sid = 1; @@ -3516,11 +3570,11 @@ merger_put_next_buffer(grn_ctx *ctx, merger_data *data) &(buffer_data->id)); return; } - if ((data->ii->header->flags & GRN_OBJ_WITH_WEIGHT)) { + if ((data->ii->header.common->flags & GRN_OBJ_WITH_WEIGHT)) { GRN_B_DEC(buffer_data->id.weight, buffer_data->data); } merger_put_next_id(ctx, data, &(buffer_data->id)); - if ((data->ii->header->flags & GRN_OBJ_WITH_POSITION)) { + if ((data->ii->header.common->flags & GRN_OBJ_WITH_POSITION)) { uint32_t rest_n_dest_positiongs = (uint32_t)(data->dest.position_gaps_end - data->dest.position_gaps); @@ -3694,7 +3748,7 @@ chunk_merge(grn_ctx *ctx, return ctx->rc; } - if ((ii->header->flags & GRN_OBJ_WITH_POSITION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_POSITION)) { rdv[ii->n_elements - 1].flags = ODD; } { @@ -3748,11 +3802,11 @@ chunk_merge(grn_ctx *ctx, { int j = 0; data->dest.record_id_gaps = dv[j++].data; - if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { data->dest.section_id_gaps = dv[j++].data; } data->dest.tfs = dv[j++].data; - if ((ii->header->flags & GRN_OBJ_WITH_WEIGHT)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_WEIGHT)) { data->dest.weights = dv[j++].data; } data->dest.position_gaps = dv[j++].data; @@ -3805,14 +3859,14 @@ chunk_merge(grn_ctx *ctx, uint32_t f_s = (ndf < 3) ? 0 : USE_P_ENC; uint32_t f_d = ((ndf < 16) || (ndf <= (data->last_id.rid >> 8))) ? 0 : USE_P_ENC; dv[j].data_size = ndf; dv[j++].flags = f_d; - if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { dv[j].data_size = ndf; dv[j++].flags = f_s; } dv[j].data_size = ndf; dv[j++].flags = f_s; - if ((ii->header->flags & GRN_OBJ_WITH_WEIGHT)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_WEIGHT)) { dv[j].data_size = ndf; dv[j++].flags = f_s; } - if ((ii->header->flags & GRN_OBJ_WITH_POSITION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_POSITION)) { uint32_t f_p; if ((np < 32) || (np <= (data->position >> 13))) { f_p = 0; @@ -3995,7 +4049,7 @@ buffer_merge(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h, name_size, name); return ctx->rc; } - if ((ii->header->flags & GRN_OBJ_WITH_POSITION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_POSITION)) { rdv[ii->n_elements - 1].flags = ODD; } @@ -4216,11 +4270,11 @@ buffer_merge(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h, { int j = 0; data.dest.record_id_gaps = dv[j++].data; - if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { data.dest.section_id_gaps = dv[j++].data; } data.dest.tfs = dv[j++].data; - if ((ii->header->flags & GRN_OBJ_WITH_WEIGHT)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_WEIGHT)) { data.dest.weights = dv[j++].data; } data.dest.position_gaps = dv[j++].data; @@ -4270,7 +4324,7 @@ buffer_merge(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h, lexicon_delete(ctx, ii, tid, h); memset(bt, 0, sizeof(buffer_term)); nterms_void++; - } else if ((ii->header->flags & GRN_OBJ_WITH_SECTION) && + } else if ((ii->header.common->flags & GRN_OBJ_WITH_SECTION) && !nvchunks && ndf == 1 && data.last_id.rid < 0x100000 && @@ -4278,20 +4332,20 @@ buffer_merge(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h, data.last_id.tf == 1 && data.last_id.weight == 0) { a[0] = (data.last_id.rid << 12) + (data.last_id.sid << 1) + 1; - if (ii->header->flags & GRN_OBJ_WITH_POSITION) { + if (ii->header.common->flags & GRN_OBJ_WITH_POSITION) { a[1] = data.dest.position_gaps[-1]; } else { a[1] = 0; } memset(bt, 0, sizeof(buffer_term)); nterms_void++; - } else if (!(ii->header->flags & GRN_OBJ_WITH_SECTION) && + } else if (!(ii->header.common->flags & GRN_OBJ_WITH_SECTION) && !nvchunks && ndf == 1 && data.last_id.tf == 1 && data.last_id.weight == 0) { a[0] = (data.last_id.rid << 1) + 1; - if (ii->header->flags & GRN_OBJ_WITH_POSITION) { + if (ii->header.common->flags & GRN_OBJ_WITH_POSITION) { a[1] = data.dest.position_gaps[-1]; } else { a[1] = 0; @@ -4304,14 +4358,14 @@ buffer_merge(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h, uint32_t f_s = (ndf < 3) ? 0 : USE_P_ENC; uint32_t f_d = ((ndf < 16) || (ndf <= (data.last_id.rid >> 8))) ? 0 : USE_P_ENC; dv[j].data_size = ndf; dv[j++].flags = f_d; - if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { dv[j].data_size = ndf; dv[j++].flags = f_s; } dv[j].data_size = ndf; dv[j++].flags = f_s; - if ((ii->header->flags & GRN_OBJ_WITH_WEIGHT)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_WEIGHT)) { dv[j].data_size = ndf; dv[j++].flags = f_s; } - if ((ii->header->flags & GRN_OBJ_WITH_POSITION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_POSITION)) { uint32_t np = data.dest.position_gaps - dv[ii->n_elements - 1].data; uint32_t f_p; if ((np < 32) || (np <= (data.position >> 13))) { @@ -4444,7 +4498,7 @@ buffer_flush(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h) grn_io_win sw, dw; buffer *sb, *db = NULL; uint32_t ds, pseg, scn, dcn = 0; - if (ii->header->binfo[seg] == GRN_II_PSEG_NOT_ASSIGNED) { + if (grn_ii_get_buffer_pseg_inline(ii, seg) == GRN_II_PSEG_NOT_ASSIGNED) { DEFINE_NAME(ii); CRIT(GRN_FILE_CORRUPT, "[ii][buffer][flush] invalid segment: " @@ -4504,11 +4558,11 @@ buffer_flush(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h) if (rc == GRN_SUCCESS) { dc = NULL; buffer_segment_update(ii, seg, ds); - ii->header->total_chunk_size += actual_chunk_size; + ii->header.common->total_chunk_size += actual_chunk_size; need_chunk_free = false; if (scn != GRN_II_PSEG_NOT_ASSIGNED) { chunk_free(ctx, ii, scn, 0, sb->header.chunk_size); - ii->header->total_chunk_size -= sb->header.chunk_size; + ii->header.common->total_chunk_size -= sb->header.chunk_size; } } else { DEFINE_NAME(ii); @@ -4571,7 +4625,7 @@ grn_ii_buffer_check(grn_ctx *ctx, grn_ii *ii, uint32_t seg) grn_obj buf; size_t lower_bound; int64_t nloops = 0, nviolations = 0; - if (ii->header->binfo[seg] == GRN_II_PSEG_NOT_ASSIGNED) { + if (grn_ii_get_buffer_pseg_inline(ii, seg) == GRN_II_PSEG_NOT_ASSIGNED) { GRN_OUTPUT_BOOL(GRN_FALSE); return; } @@ -4584,7 +4638,7 @@ grn_ii_buffer_check(grn_ctx *ctx, grn_ii *ii, uint32_t seg) (sb->header.buffer_free + sizeof(buffer_term) * sb->header.nterms) / sizeof(buffer_rec); datavec_init(ctx, rdv, ii->n_elements, 0, 0); - if ((ii->header->flags & GRN_OBJ_WITH_POSITION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_POSITION)) { rdv[ii->n_elements - 1].flags = ODD; } GRN_OUTPUT_MAP_OPEN("BUFFER", -1); @@ -4678,14 +4732,14 @@ grn_ii_buffer_check(grn_ctx *ctx, grn_ii *ii, uint32_t seg) { int j = 0; j++; - if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { j++; } if (chunk_data->n_documents != rdv[j].data_size) { nterms_with_corrupt_chunk++; } j++; - if ((ii->header->flags & GRN_OBJ_WITH_WEIGHT)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_WEIGHT)) { j++; } } @@ -4704,7 +4758,7 @@ grn_ii_buffer_check(grn_ctx *ctx, grn_ii *ii, uint32_t seg) r = BUFFER_REC_AT(sb, pos); p = GRN_NEXT_ADDR(r); GRN_B_DEC(rid, p); - if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { GRN_B_DEC(sid, p); } else { sid = 1; @@ -4840,7 +4894,7 @@ buffer_split(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h) grn_io_win sw, dw0, dw1; buffer *sb, *db0 = NULL, *db1 = NULL; uint32_t dps0 = 0, dps1 = 0, dls0 = 0, dls1 = 0, sps, scn, dcn0 = 0, dcn1 = 0; - if (ii->header->binfo[seg] == GRN_II_PSEG_NOT_ASSIGNED) { + if (grn_ii_get_buffer_pseg_inline(ii, seg) == GRN_II_PSEG_NOT_ASSIGNED) { DEFINE_NAME(ii); CRIT(GRN_FILE_CORRUPT, "[ii][buffer][split] invalid segment: " @@ -4924,13 +4978,15 @@ buffer_split(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h) array_update(ctx, ii, dls0, db0); array_update(ctx, ii, dls1, db1); buffer_segment_clear(ii, seg); - ii->header->total_chunk_size += actual_db0_chunk_size; - ii->header->total_chunk_size += actual_db1_chunk_size; + ii->header.common->total_chunk_size += + actual_db0_chunk_size + + actual_db1_chunk_size; need_db0_chunk_free = false; need_db1_chunk_free = false; if (scn != GRN_II_PSEG_NOT_ASSIGNED) { chunk_free(ctx, ii, scn, 0, sb->header.chunk_size); - ii->header->total_chunk_size -= sb->header.chunk_size; + ii->header.common->total_chunk_size -= + sb->header.chunk_size; } } else { DEFINE_NAME(ii); @@ -5031,7 +5087,7 @@ buffer_split(grn_ctx *ctx, grn_ii *ii, uint32_t seg, grn_hash *h) #define SPLIT_COND(ii, buffer)\ ((buffer)->header.nterms > 1024 ||\ ((buffer)->header.nterms > 1 &&\ - (buffer)->header.chunk_size * 100 > (ii)->header->total_chunk_size)) + (buffer)->header.chunk_size * 100 > (ii)->header.common->total_chunk_size)) grn_inline static void buffer_new_find_segment(grn_ctx *ctx, @@ -5061,14 +5117,14 @@ buffer_new_find_segment(grn_ctx *ctx, } buffer_close(ctx, ii, *pseg); if (SPLIT_COND(ii, (*b))) { - /* ((S_SEGMENT - sizeof(buffer_header) + ii->header->bmax - + /* ((S_SEGMENT - sizeof(buffer_header) + ii->header.common->bmax - (*b)->header.nterms * sizeof(buffer_term)) * 4 < (*b)->header.chunk_size) */ GRN_LOG(ctx, GRN_LOG_DEBUG, "nterms=%d chunk=%d total=%" GRN_FMT_INT64U, (*b)->header.nterms, (*b)->header.chunk_size, - ii->header->total_chunk_size >> 10); + ii->header.common->total_chunk_size >> 10); if (buffer_split(ctx, ii, LSEG(pos), h)) { break; } } else { if (S_SEGMENT - sizeof(buffer_header) @@ -5275,11 +5331,12 @@ static grn_ii * _grn_ii_create(grn_ctx *ctx, grn_ii *ii, const char *path, grn_obj *lexicon, uint32_t flags) { int i; + uint32_t header_size; uint32_t max_n_segments; uint32_t max_n_chunks; grn_io *seg, *chunk; char path2[PATH_MAX]; - struct grn_ii_header *header; + grn_ii_header_common *header; grn_table_flags lflags; grn_encoding encoding; grn_obj *tokenizer; @@ -5296,19 +5353,26 @@ _grn_ii_create(grn_ctx *ctx, grn_ii *ii, const char *path, grn_obj *lexicon, uin if (path && strlen(path) + 6 >= PATH_MAX) { return NULL; } if (flags & GRN_OBJ_INDEX_SMALL) { + header_size = sizeof(grn_ii_header_normal); max_n_segments = grn_ii_max_n_segments_small; max_n_chunks = grn_ii_max_n_chunks_small; } else if (flags & GRN_OBJ_INDEX_MEDIUM) { + header_size = sizeof(grn_ii_header_normal); max_n_segments = MAX_PSEG_MEDIUM; max_n_chunks = GRN_II_MAX_CHUNK_MEDIUM; + } else if (flags & GRN_OBJ_INDEX_LARGE) { + header_size = sizeof(grn_ii_header_large); + max_n_segments = MAX_PSEG_LARGE; + max_n_chunks = GRN_II_MAX_CHUNK; } else { + header_size = sizeof(grn_ii_header_normal); max_n_segments = MAX_PSEG; max_n_chunks = GRN_II_MAX_CHUNK; } seg = grn_io_create(ctx, path, - sizeof(struct grn_ii_header), + header_size, S_SEGMENT, max_n_segments, grn_io_auto, @@ -5333,6 +5397,13 @@ _grn_ii_create(grn_ctx *ctx, grn_ii *ii, const char *path, grn_obj *lexicon, uin header->ainfo[i] = GRN_II_PSEG_NOT_ASSIGNED; header->binfo[i] = GRN_II_PSEG_NOT_ASSIGNED; } + if (flags & GRN_OBJ_INDEX_LARGE) { + grn_ii_header_large *header_large = (grn_ii_header_large *)header; + for (i = 0; i < GRN_II_MAX_LSEG; i++) { + header_large->ainfo_extend[i] = GRN_II_PSEG_NOT_ASSIGNED; + header_large->binfo_extend[i] = GRN_II_PSEG_NOT_ASSIGNED; + } + } for (i = 0; i <= GRN_II_N_CHUNK_VARIATION; i++) { header->free_chunks[i] = GRN_II_PSEG_NOT_ASSIGNED; header->garbages[i] = GRN_II_PSEG_NOT_ASSIGNED; @@ -5343,7 +5414,7 @@ _grn_ii_create(grn_ctx *ctx, grn_ii *ii, const char *path, grn_obj *lexicon, uin ii->lexicon = lexicon; ii->lflags = lflags; ii->encoding = encoding; - ii->header = header; + ii->header.common = header; ii->n_elements = 2; if ((flags & GRN_OBJ_WITH_SECTION)) { ii->n_elements++; } if ((flags & GRN_OBJ_WITH_WEIGHT)) { ii->n_elements++; } @@ -5405,7 +5476,7 @@ grn_ii_truncate(grn_ctx *ctx, grn_ii *ii) segpath = NULL; } lexicon = ii->lexicon; - flags = ii->header->flags; + flags = ii->header.common->flags; if ((rc = grn_io_close(ctx, ii->seg))) { goto exit; } if ((rc = grn_io_close(ctx, ii->chunk))) { goto exit; } ii->seg = NULL; @@ -5427,7 +5498,7 @@ grn_ii_open(grn_ctx *ctx, const char *path, grn_obj *lexicon) grn_io *seg, *chunk; grn_ii *ii; char path2[PATH_MAX]; - struct grn_ii_header *header; + grn_ii_header_common *header; uint32_t io_type; grn_table_flags lflags; grn_encoding encoding; @@ -5467,7 +5538,7 @@ grn_ii_open(grn_ctx *ctx, const char *path, grn_obj *lexicon) ii->lexicon = lexicon; ii->lflags = lflags; ii->encoding = encoding; - ii->header = header; + ii->header.common = header; ii->n_elements = 2; if ((header->flags & GRN_OBJ_WITH_SECTION)) { ii->n_elements++; } if ((header->flags & GRN_OBJ_WITH_WEIGHT)) { ii->n_elements++; } @@ -5523,7 +5594,7 @@ grn_ii_get_flags(grn_ctx *ctx, grn_ii *ii) return 0; } - return ii->header->flags; + return ii->header.common->flags; } uint32_t @@ -5589,7 +5660,7 @@ grn_ii_update_one(grn_ctx *ctx, grn_ii *ii, grn_id tid, grn_ii_updspec *u, grn_h uint32_t pseg = 0, pos = 0, size, *a; if (!tid) { return ctx->rc; } if (!u->tf || !u->sid) { return grn_ii_delete_one(ctx, ii, tid, u, h); } - if (u->sid > ii->header->smax) { ii->header->smax = u->sid; } + if (u->sid > ii->header.common->smax) { ii->header.common->smax = u->sid; } if (!(a = array_get(ctx, ii, tid))) { grn_obj term; DEFINE_NAME(ii); @@ -5654,14 +5725,14 @@ grn_ii_update_one(grn_ctx *ctx, grn_ii *ii, grn_id tid, grn_ii_updspec *u, grn_h a[0], LSEG(a[0]), b, b->header.buffer_free); buffer_close(ctx, ii, pseg); if (SPLIT_COND(ii, b)) { - /*((S_SEGMENT - sizeof(buffer_header) + ii->header->bmax - + /*((S_SEGMENT - sizeof(buffer_header) + ii->header.common->bmax - b->header.nterms * sizeof(buffer_term)) * 4 < b->header.chunk_size)*/ GRN_LOG(ctx, GRN_LOG_DEBUG, "nterms=%d chunk=%d total=%" GRN_FMT_INT64U, b->header.nterms, b->header.chunk_size, - ii->header->total_chunk_size >> 10); + ii->header.common->total_chunk_size >> 10); buffer_split(ctx, ii, LSEG(pos), h); if (ctx->rc != GRN_SUCCESS) { grn_obj term; @@ -5789,7 +5860,7 @@ grn_ii_update_one(grn_ctx *ctx, grn_ii *ii, grn_id tid, grn_ii_updspec *u, grn_h pos2.pos = a[1]; pos2.next = NULL; u2.pos = &pos2; - if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { u2.rid = BIT31_12(v); u2.sid = BIT11_01(v); } else { @@ -5877,7 +5948,7 @@ grn_ii_update_one(grn_ctx *ctx, grn_ii *ii, grn_id tid, grn_ii_updspec *u, grn_h } if (!br) { if (u->tf == 1 && u->weight == 0) { - if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { if (u->rid < 0x100000 && u->sid < 0x800) { a[0] = (u->rid << 12) + (u->sid << 1) + 1; a[1] = u->pos->pos; @@ -5976,7 +6047,7 @@ grn_ii_delete_one(grn_ctx *ctx, grn_ii *ii, grn_id tid, grn_ii_updspec *u, grn_h for (;;) { if (!a[0]) { goto exit; } if (a[0] & 1) { - if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { uint32_t rid = BIT31_12(a[0]); uint32_t sid = BIT11_01(a[0]); if (u->rid == rid && (!u->sid || u->sid == sid)) { @@ -6166,9 +6237,9 @@ buffer_is_reused(grn_ctx *ctx, grn_ii *ii, grn_ii_cursor *c) { if (*c->ppseg != c->buffer_pseg) { uint32_t i; - for (i = ii->header->bgqtail; i != ii->header->bgqhead; + for (i = ii->header.common->bgqtail; i != ii->header.common->bgqhead; i = (i + 1) & (GRN_II_BGQSIZE - 1)) { - if (ii->header->bgqbody[i] == c->buffer_pseg) { return GRN_FALSE; } + if (ii->header.common->bgqbody[i] == c->buffer_pseg) { return GRN_FALSE; } } return GRN_TRUE; } @@ -6188,7 +6259,7 @@ chunk_is_reused(grn_ctx *ctx, grn_ii *ii, grn_ii_cursor *c, uint32_t offset, uin } else { m = GRN_II_W_LEAST_CHUNK; } - gseg = ii->header->garbages[m - GRN_II_W_LEAST_CHUNK]; + gseg = ii->header.common->garbages[m - GRN_II_W_LEAST_CHUNK]; while (gseg != GRN_II_PSEG_NOT_ASSIGNED) { grn_io_win iw; grn_ii_ginfo *ginfo = WIN_MAP(ii->chunk, ctx, &iw, gseg, 0, S_GARBAGE, @@ -6237,7 +6308,7 @@ grn_ii_cursor_open(grn_ctx *ctx, grn_ii *ii, grn_id tid, c->weight = 0; if (pos & 1) { c->stat = 0; - if ((ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { c->pb.rid = BIT31_12(pos); c->pb.sid = BIT11_01(pos); } else { @@ -6256,7 +6327,7 @@ grn_ii_cursor_open(grn_ctx *ctx, grn_ii *ii, grn_id tid, c = NULL; goto exit; } - c->ppseg = &ii->header->binfo[LSEG(pos)]; + c->ppseg = &ii->header.common->binfo[LSEG(pos)]; if (bt->size_in_chunk && (chunk = c->buf->header.chunk) != GRN_II_PSEG_NOT_ASSIGNED) { if (!(c->cp = WIN_MAP(ii->chunk, ctx, &c->iw, chunk, bt->pos_in_chunk, bt->size_in_chunk, grn_io_rdonly))) { @@ -6300,7 +6371,7 @@ grn_ii_cursor_open(grn_ctx *ctx, grn_ii *ii, grn_id tid, continue; } } - if ((ii->header->flags & GRN_OBJ_WITH_POSITION)) { + if ((ii->header.common->flags & GRN_OBJ_WITH_POSITION)) { c->rdv[ii->n_elements - 1].flags = ODD; } } @@ -6387,14 +6458,14 @@ grn_ii_cursor_next_internal(grn_ctx *ctx, grn_ii_cursor *c, uint32_t dgap = *c->crp++; c->pc.rid += dgap; if (dgap) { c->pc.sid = 0; } - if ((c->ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((c->ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { c->pc.sid += 1 + *c->csp++; } else { c->pc.sid = 1; } c->cpp += c->pc.rest; c->pc.rest = c->pc.tf = 1 + *c->ctp++; - if ((c->ii->header->flags & GRN_OBJ_WITH_WEIGHT)) { + if ((c->ii->header.common->flags & GRN_OBJ_WITH_WEIGHT)) { c->pc.weight = *c->cwp++; } else { c->pc.weight = 0; @@ -6531,14 +6602,14 @@ grn_ii_cursor_next_internal(grn_ctx *ctx, grn_ii_cursor *c, int j = 0; c->cdf = c->rdv[j].data_size; c->crp = c->cdp = c->rdv[j++].data; - if ((c->ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((c->ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { c->csp = c->rdv[j++].data; } c->ctp = c->rdv[j++].data; - if ((c->ii->header->flags & GRN_OBJ_WITH_WEIGHT)) { + if ((c->ii->header.common->flags & GRN_OBJ_WITH_WEIGHT)) { c->cwp = c->rdv[j++].data; } - if ((c->ii->header->flags & GRN_OBJ_WITH_POSITION)) { + if ((c->ii->header.common->flags & GRN_OBJ_WITH_POSITION)) { c->cpp = c->rdv[j].data; } } @@ -6576,7 +6647,7 @@ grn_ii_cursor_next_internal(grn_ctx *ctx, grn_ii_cursor *c, } c->bp = GRN_NEXT_ADDR(br); GRN_B_DEC(c->pb.rid, c->bp); - if ((c->ii->header->flags & GRN_OBJ_WITH_SECTION)) { + if ((c->ii->header.common->flags & GRN_OBJ_WITH_SECTION)) { GRN_B_DEC(c->pb.sid, c->bp); } else { c->pb.sid = 1; @@ -6618,7 +6689,7 @@ grn_ii_cursor_next_internal(grn_ctx *ctx, grn_ii_cursor *c, } c->nextb = br->step; GRN_B_DEC(c->pb.tf, c->bp); - if ((c->ii->header->flags & GRN_OBJ_WITH_WEIGHT)) { + if ((c->ii->header.common->flags & GRN_OBJ_WITH_WEIGHT)) { GRN_B_DEC(c->pb.weight, c->bp); } else { c->pb.weight = 0; @@ -6716,7 +6787,7 @@ grn_posting * grn_ii_cursor_next_pos(grn_ctx *ctx, grn_ii_cursor *c) { uint32_t gap; - if ((c->ii->header->flags & GRN_OBJ_WITH_POSITION)) { + if ((c->ii->header.common->flags & GRN_OBJ_WITH_POSITION)) { if (c->nelements == c->ii->n_elements) { if (c->buf) { if (c->post == &c->pc) { @@ -6862,6 +6933,26 @@ grn_ii_entry_info(grn_ctx *ctx, grn_ii *ii, grn_id tid, unsigned int *a, return 4; } +uint32_t +grn_ii_get_array_pseg(grn_ii *ii, + uint32_t lseg) +{ + return grn_ii_get_array_pseg_inline(ii, lseg); +} + +uint32_t +grn_ii_get_buffer_pseg(grn_ii *ii, + uint32_t lseg) +{ + return grn_ii_get_buffer_pseg_inline(ii, lseg); +} + +uint32_t +grn_ii_n_logical_segments(grn_ii *ii) +{ + return grn_ii_n_logical_segments_inline(ii); +} + const char * grn_ii_path(grn_ii *ii) { @@ -6871,7 +6962,7 @@ grn_ii_path(grn_ii *ii) uint32_t grn_ii_max_section(grn_ii *ii) { - return ii->header->smax; + return ii->header.common->smax; } grn_obj * @@ -9705,7 +9796,7 @@ grn_ii_quorum_match(grn_ctx *ctx, grn_ii *ii, grn_ii_select_data *data) int ii_cursor_flags = 0; record_key_size = sizeof(grn_id); - if (ii->header->flags & GRN_OBJ_WITH_SECTION) { + if (ii->header.common->flags & GRN_OBJ_WITH_SECTION) { record_key_size += sizeof(uint32_t); ii_cursor_flags |= GRN_OBJ_WITH_SECTION; } @@ -10832,7 +10923,7 @@ static uint8_t * allocate_outbuf(grn_ctx *ctx, grn_ii_buffer *ii_buffer) { size_t bufsize = 0, bufsize_ = 0; - uint32_t flags = ii_buffer->ii->header->flags; + uint32_t flags = ii_buffer->ii->header.common->flags; ii_buffer_counter *counter = ii_buffer->counters; grn_id tid, tid_max = grn_table_size(ctx, ii_buffer->tmp_lexicon); for (tid = 1; tid <= tid_max; counter++, tid++) { @@ -10894,7 +10985,7 @@ encode_terms(grn_ctx *ctx, grn_ii_buffer *ii_buffer, grn_table_cursor *tc; /* The first size is written into block->nextsize. */ uint8_t *pnext = (uint8_t *)&block->nextsize; - uint32_t flags = ii_buffer->ii->header->flags; + uint32_t flags = ii_buffer->ii->header.common->flags; tc = grn_table_cursor_open(ctx, ii_buffer->tmp_lexicon, NULL, 0, NULL, 0, 0, -1, II_BUFFER_ORDER); while ((tid = grn_table_cursor_next(ctx, tc)) != GRN_ID_NIL) { @@ -10957,7 +11048,7 @@ encode_postings(grn_ctx *ctx, grn_ii_buffer *ii_buffer, uint8_t *outbuf) uint32_t pos = 0; uint32_t rest; grn_id *bp = ii_buffer->block_buf; - uint32_t flags = ii_buffer->ii->header->flags; + uint32_t flags = ii_buffer->ii->header.common->flags; for (rest = ii_buffer->block_pos; rest; bp++, rest--) { grn_id id = *bp; switch (II_BUFFER_TYPE(id)) { @@ -11034,7 +11125,7 @@ encode_last_tf(grn_ctx *ctx, grn_ii_buffer *ii_buffer, uint8_t *outbuf) uint8_t *p = outbuf + counter->offset_tf; GRN_B_ENC(counter->last_tf - 1, p); } - if ((ii_buffer->ii->header->flags & GRN_OBJ_WITH_WEIGHT)) { + if ((ii_buffer->ii->header.common->flags & GRN_OBJ_WITH_WEIGHT)) { for (tid = 1; tid <= tid_max; counter++, tid++) { uint8_t *p = outbuf + counter->offset_weight; GRN_B_ENC(counter->last_weight, p); @@ -11182,7 +11273,7 @@ grn_ii_buffer_tokenize_value(grn_ctx *ctx, grn_ii_buffer *ii_buffer, grn_token_cursor *token_cursor; grn_id *buffer = ii_buffer->block_buf; uint32_t block_pos = ii_buffer->block_pos; - uint32_t ii_flags = ii_buffer->ii->header->flags; + uint32_t ii_flags = ii_buffer->ii->header.common->flags; buffer[block_pos++] = II_BUFFER_PACK(rid, II_BUFFER_TYPE_RID); if (ii_flags & GRN_OBJ_WITH_SECTION) { buffer[block_pos++] = value->sid; @@ -11373,13 +11464,13 @@ grn_ii_buffer_chunk_flush(grn_ctx *ctx, grn_ii_buffer *ii_buffer) ii_buffer->term_buffer->header.nterms * sizeof(buffer_term); ii_buffer->term_buffer->header.nterms_void = 0; buffer_segment_update(ii_buffer->ii, ii_buffer->lseg, ii_buffer->dseg); - ii_buffer->ii->header->total_chunk_size += ii_buffer->packed_len; + ii_buffer->ii->header.common->total_chunk_size += ii_buffer->packed_len; ii_buffer->total_chunk_size += ii_buffer->packed_len; GRN_LOG(ctx, GRN_LOG_DEBUG, "nterms=%d chunk=%d total=%" GRN_FMT_INT64U "KB", ii_buffer->term_buffer->header.nterms, ii_buffer->term_buffer->header.chunk_size, - ii_buffer->ii->header->total_chunk_size >> 10); + ii_buffer->ii->header.common->total_chunk_size >> 10); ii_buffer->term_buffer = NULL; ii_buffer->packed_buf = NULL; ii_buffer->packed_len = 0; @@ -11398,7 +11489,7 @@ merge_hit_blocks(grn_ctx *ctx, grn_ii_buffer *ii_buffer, uint64_t nrecs = 0; uint64_t nposts = 0; size_t max_size; - uint64_t flags = ii_buffer->ii->header->flags; + uint64_t flags = ii_buffer->ii->header.common->flags; int i; for (i = 0; i < nhits; i++) { ii_buffer_block *block = hits[i]; @@ -11499,12 +11590,14 @@ static buffer * get_term_buffer(grn_ctx *ctx, grn_ii_buffer *ii_buffer) { if (!ii_buffer->term_buffer) { + const uint32_t n_logical_segments = + grn_ii_n_logical_segments_inline(ii_buffer->ii); uint32_t lseg; - void *term_buffer; - for (lseg = 0; lseg < GRN_II_MAX_LSEG; lseg++) { - if (ii_buffer->ii->header->binfo[lseg] == GRN_II_PSEG_NOT_ASSIGNED) { break; } + for (lseg = 0; lseg < n_logical_segments; lseg++) { + const uint32_t pseg = grn_ii_get_buffer_pseg_inline(ii_buffer->ii, lseg); + if (pseg == GRN_II_PSEG_NOT_ASSIGNED) { break; } } - if (lseg == GRN_II_MAX_LSEG) { + if (lseg == n_logical_segments) { DEFINE_NAME(ii_buffer->ii); MERR("[ii][buffer][term-buffer] couldn't find a free buffer: " "<%.*s>", @@ -11513,6 +11606,7 @@ get_term_buffer(grn_ctx *ctx, grn_ii_buffer *ii_buffer) } ii_buffer->lseg = lseg; ii_buffer->dseg = segment_get(ctx, ii_buffer->ii); + void *term_buffer; GRN_IO_SEG_REF(ii_buffer->ii->seg, ii_buffer->dseg, term_buffer); ii_buffer->term_buffer = (buffer *)term_buffer; } @@ -11537,7 +11631,7 @@ try_in_place_packing(grn_ctx *ctx, grn_ii_buffer *ii_buffer, uint32_t sid = 1, tf, pos = 0, weight = 0; ii_buffer_block *block = hits[0]; uint8_t *p = block->bufcur; - uint32_t flags = ii_buffer->ii->header->flags; + uint32_t flags = ii_buffer->ii->header.common->flags; GRN_B_DEC(rid, p); if (flags & GRN_OBJ_WITH_SECTION) { GRN_B_DEC(sid, p); @@ -12478,7 +12572,7 @@ grn_ii_builder_buffer_assign(grn_ctx *ctx, grn_ii_builder_buffer *buf, } return rc; } - buf->buf_seg_id = buf->ii->header->binfo[buf->buf_id]; + buf->buf_seg_id = grn_ii_get_buffer_pseg_inline(buf->ii, buf->buf_id); GRN_IO_SEG_REF(buf->ii->seg, buf->buf_seg_id, seg); if (!seg) { if (ctx->rc == GRN_SUCCESS) { @@ -12520,7 +12614,7 @@ grn_ii_builder_buffer_assign(grn_ctx *ctx, grn_ii_builder_buffer *buf, buf->buf->header.buffer_free = S_SEGMENT - sizeof(buffer_header); buf->buf->header.nterms = 0; buf->buf->header.nterms_void = 0; - buf->ii->header->total_chunk_size += chunk_size; + buf->ii->header.common->total_chunk_size += chunk_size; return GRN_SUCCESS; } @@ -12538,7 +12632,7 @@ grn_ii_builder_buffer_flush(grn_ctx *ctx, grn_ii_builder_buffer *buf) buf->buf->header.nterms, buf->chunk_offset, buf->buf->header.chunk_size, - buf->ii->header->total_chunk_size >> 10); + buf->ii->header.common->total_chunk_size >> 10); ii = buf->ii; grn_ii_builder_buffer_fin(ctx, buf); @@ -13222,7 +13316,7 @@ grn_ii_builder_flush_term(grn_ctx *ctx, grn_ii_builder *builder, /* Append sentinels. */ if (term->rid != GRN_ID_NIL) { - if (builder->ii->header->flags & GRN_OBJ_WITH_POSITION) { + if (builder->ii->header.common->flags & GRN_OBJ_WITH_POSITION) { rc = grn_ii_builder_term_append(ctx, term, 0); } else { rc = grn_ii_builder_term_append(ctx, term, term->pos_or_freq); @@ -13457,7 +13551,7 @@ grn_ii_builder_append_token(grn_ctx *ctx, grn_ii_builder *builder, grn_id tid, uint32_t pos) { grn_rc rc; - uint32_t ii_flags = builder->ii->header->flags; + uint32_t ii_flags = builder->ii->header.common->flags; grn_ii_builder_term *term; rc = grn_ii_builder_get_term(ctx, builder, tid, &term); if (rc != GRN_SUCCESS) { @@ -13635,7 +13729,7 @@ grn_ii_builder_append_obj(grn_ctx *ctx, grn_ii_builder *builder, if (sec->length == 0) { continue; } - if ((builder->ii->header->flags & GRN_OBJ_WITH_SECTION) && + if ((builder->ii->header.common->flags & GRN_OBJ_WITH_SECTION) && builder->have_tokenizer) { sid = i + 1; } @@ -14113,7 +14207,7 @@ grn_ii_builder_flush_chunk(grn_ctx *ctx, grn_ii_builder *builder) cinfo->size = chunk->enc_offset; cinfo->dgap = chunk->rid_gap; - builder->buf.ii->header->total_chunk_size += chunk->enc_offset; + builder->buf.ii->header.common->total_chunk_size += chunk->enc_offset; grn_ii_builder_chunk_clear(ctx, chunk); return GRN_SUCCESS; } @@ -14126,7 +14220,7 @@ grn_ii_builder_read_to_chunk(grn_ctx *ctx, grn_ii_builder *builder, grn_rc rc; uint64_t value; uint32_t rid = GRN_ID_NIL, last_sid = 0; - uint32_t ii_flags = builder->ii->header->flags; + uint32_t ii_flags = builder->ii->header.common->flags; grn_ii_builder_chunk *chunk = &builder->chunk; for (;;) { Modified: lib/index_column.c (+12 -5) =================================================================== --- lib/index_column.c 2019-03-29 10:45:34 +0900 (51e05131d) +++ lib/index_column.c 2019-04-01 17:34:47 +0900 (7a47159b7) @@ -180,12 +180,18 @@ grn_index_column_build(grn_ctx *ctx, grn_obj *index_column) target = GRN_OBJ_TABLEP(src) ? src : grn_ctx_at(ctx, src->header.domain); if (target) { int i, ncol = DB_OBJ(index_column)->source_size / sizeof(grn_id); - grn_table_flags flags; + grn_table_flags lexicon_flags; grn_ii *ii = (grn_ii *)index_column; grn_bool use_grn_ii_build; grn_obj *tokenizer = NULL; - grn_table_get_info(ctx, ii->lexicon, &flags, NULL, &tokenizer, NULL, NULL); - switch (flags & GRN_OBJ_TABLE_TYPE_MASK) { + grn_table_get_info(ctx, + ii->lexicon, + &lexicon_flags, + NULL, + &tokenizer, + NULL, + NULL); + switch (lexicon_flags & GRN_OBJ_TABLE_TYPE_MASK) { case GRN_OBJ_TABLE_PAT_KEY : case GRN_OBJ_TABLE_DAT_KEY : use_grn_ii_build = GRN_TRUE; @@ -194,10 +200,11 @@ grn_index_column_build(grn_ctx *ctx, grn_obj *index_column) use_grn_ii_build = GRN_FALSE; break; } - if ((ii->header->flags & GRN_OBJ_WITH_WEIGHT)) { + const grn_column_flags flags = grn_ii_get_flags(ctx, ii); + if ((flags & GRN_OBJ_WITH_WEIGHT)) { use_grn_ii_build = GRN_FALSE; } - if ((ii->header->flags & GRN_OBJ_WITH_POSITION) && + if ((flags & GRN_OBJ_WITH_POSITION) && (!tokenizer && !GRN_TYPE_IS_TEXT_FAMILY(ii->lexicon->header.domain))) { /* TODO: Support offline index construction for WITH_POSITION Modified: lib/proc.c (+8 -4) =================================================================== --- lib/proc.c 2019-03-29 10:45:34 +0900 (ae85f4faf) +++ lib/proc.c 2019-04-01 17:34:47 +0900 (140e20a39) @@ -1,7 +1,7 @@ /* -*- c-basic-offset: 2 -*- */ /* Copyright(C) 2009-2018 Brazil - Copyright(C) 2018 Kouhei Sutou <kou****@clear*****> + Copyright(C) 2018-2019 Kouhei Sutou <kou****@clear*****> This library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public @@ -1074,7 +1074,7 @@ proc_check(grn_ctx *ctx, int nargs, grn_obj **args, grn_user_data *user_data) case GRN_COLUMN_INDEX : { grn_ii *ii = (grn_ii *)obj; - struct grn_ii_header *h = ii->header; + grn_ii_header_common *h = ii->header.common; char buf[8]; GRN_OUTPUT_ARRAY_OPEN("RESULT", 8); { @@ -1138,8 +1138,12 @@ proc_check(grn_ctx *ctx, int nargs, grn_obj **args, grn_user_data *user_data) } GRN_OUTPUT_ARRAY_CLOSE(); GRN_OUTPUT_MAP_CLOSE(); - for (i = 0; i < GRN_II_MAX_LSEG; i++) { - if (h->binfo[i] < 0x20000) { grn_ii_buffer_check(ctx, ii, i); } + const uint32_t n_logical_segments = grn_ii_n_logical_segments(ii); + for (uint32_t lseg = 0; lseg < n_logical_segments; lseg++) { + const uint32_t pseg = grn_ii_get_buffer_pseg(ii, lseg); + if (pseg != GRN_II_PSEG_NOT_ASSIGNED) { + grn_ii_buffer_check(ctx, ii, lseg); + } } } GRN_OUTPUT_ARRAY_CLOSE(); Modified: lib/proc/proc_column.c (+2 -1) =================================================================== --- lib/proc/proc_column.c 2019-03-29 10:45:34 +0900 (124d12ce1) +++ lib/proc/proc_column.c 2019-04-01 17:34:47 +0900 (767d0e216) @@ -1,7 +1,7 @@ /* -*- c-basic-offset: 2 -*- */ /* Copyright(C) 2009-2016 Brazil - Copyright(C) 2018 Kouhei Sutou <kou****@clear*****> + Copyright(C) 2018-2019 Kouhei Sutou <kou****@clear*****> This library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public @@ -61,6 +61,7 @@ grn_proc_column_parse_flags(grn_ctx *ctx, CHECK_FLAG(RING_BUFFER); CHECK_FLAG(INDEX_SMALL); CHECK_FLAG(INDEX_MEDIUM); + CHECK_FLAG(INDEX_LARGE); #undef CHECK_FLAG Modified: lib/proc/proc_object_inspect.c (+13 -13) =================================================================== --- lib/proc/proc_object_inspect.c 2019-03-29 10:45:34 +0900 (3b03ccb5e) +++ lib/proc/proc_object_inspect.c 2019-04-01 17:34:47 +0900 (d9ad7dfc5) @@ -1,6 +1,7 @@ /* -*- c-basic-offset: 2 -*- */ /* Copyright(C) 2016-2017 Brazil + Copyright(C) 2019 Kouhei Sutou <kou****@clear*****> This library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public @@ -243,7 +244,7 @@ command_object_inspect_column_index_value_statistics(grn_ctx *ctx, { grn_ctx_output_map_open(ctx, "statistics", 11); { - struct grn_ii_header *h = ii->header; + grn_ii_header_common *h = ii->header.common; grn_ctx_output_cstr(ctx, "max_section_id"); grn_ctx_output_uint64(ctx, grn_ii_max_section(ii)); @@ -268,16 +269,15 @@ command_object_inspect_column_index_value_statistics(grn_ctx *ctx, grn_ctx_output_uint64(ctx, n_garbage_segments); } + const uint32_t n_logical_segments = grn_ii_n_logical_segments(ii); grn_ctx_output_cstr(ctx, "max_array_segment_id"); grn_ctx_output_uint64(ctx, h->amax); grn_ctx_output_cstr(ctx, "n_array_segments"); { - uint32_t i; - - for (i = 0; i < GRN_II_MAX_LSEG; i++) { - uint32_t id = h->ainfo[i]; - if (id != GRN_II_PSEG_NOT_ASSIGNED) { - if (id > max_id) { max_id = id; } + for (uint32_t lseg = 0; lseg < n_logical_segments; lseg++) { + const uint32_t pseg = grn_ii_get_array_pseg(ii, lseg); + if (pseg != GRN_II_PSEG_NOT_ASSIGNED) { + if (pseg > max_id) { max_id = pseg; } n_array_segments++; } } @@ -288,12 +288,10 @@ command_object_inspect_column_index_value_statistics(grn_ctx *ctx, grn_ctx_output_uint64(ctx, h->bmax); grn_ctx_output_cstr(ctx, "n_buffer_segments"); { - uint32_t i; - - for (i = 0; i < GRN_II_MAX_LSEG; i++) { - uint32_t id = h->binfo[i]; - if (id != GRN_II_PSEG_NOT_ASSIGNED) { - if (id > max_id) { max_id = id; } + for (uint32_t lseg = 0; lseg < n_logical_segments; lseg++) { + const uint32_t pseg = grn_ii_get_buffer_pseg(ii, lseg); + if (pseg != GRN_II_PSEG_NOT_ASSIGNED) { + if (pseg > max_id) { max_id = pseg; } n_buffer_segments++; } } @@ -411,6 +409,8 @@ command_object_inspect_column_value(grn_ctx *ctx, grn_obj *column) grn_ctx_output_cstr(ctx, "small"); } else if ((column_flags & GRN_OBJ_INDEX_MEDIUM) != 0) { grn_ctx_output_cstr(ctx, "medium"); + } else if ((column_flags & GRN_OBJ_INDEX_LARGE) != 0) { + grn_ctx_output_cstr(ctx, "large"); } else { grn_ctx_output_cstr(ctx, "normal"); } Added: test/command/suite/dump/schema/column/index/index_large.expected (+15 -0) 100644 =================================================================== --- /dev/null +++ test/command/suite/dump/schema/column/index/index_large.expected 2019-04-01 17:34:47 +0900 (666b70f30) @@ -0,0 +1,15 @@ +table_create Users TABLE_HASH_KEY ShortText +[[0,0.0,0.0],true] +column_create Users name COLUMN_SCALAR ShortText +[[0,0.0,0.0],true] +table_create Names TABLE_HASH_KEY ShortText +[[0,0.0,0.0],true] +column_create Names users COLUMN_INDEX|INDEX_LARGE Users name +[[0,0.0,0.0],true] +dump +table_create Names TABLE_HASH_KEY ShortText + +table_create Users TABLE_HASH_KEY ShortText +column_create Users name COLUMN_SCALAR ShortText + +column_create Names users COLUMN_INDEX|INDEX_LARGE Users name Added: test/command/suite/dump/schema/column/index/index_large.test (+7 -0) 100644 =================================================================== --- /dev/null +++ test/command/suite/dump/schema/column/index/index_large.test 2019-04-01 17:34:47 +0900 (07cf2fe29) @@ -0,0 +1,7 @@ +table_create Users TABLE_HASH_KEY ShortText +column_create Users name COLUMN_SCALAR ShortText + +table_create Names TABLE_HASH_KEY ShortText +column_create Names users COLUMN_INDEX|INDEX_LARGE Users name + +dump Added: test/command/suite/schema/tables/columns/type/index_large.expected (+437 -0) 100644 =================================================================== --- /dev/null +++ test/command/suite/schema/tables/columns/type/index_large.expected 2019-04-01 17:34:47 +0900 (73498d9e6) @@ -0,0 +1,437 @@ +table_create Posts TABLE_HASH_KEY ShortText +[[0,0.0,0.0],true] +column_create Posts title COLUMN_SCALAR ShortText +[[0,0.0,0.0],true] +column_create Posts content COLUMN_SCALAR Text +[[0,0.0,0.0],true] +table_create Terms TABLE_PAT_KEY ShortText --default_tokenizer TokenBigram --normalizer NormalizerAuto +[[0,0.0,0.0],true] +column_create Terms index COLUMN_INDEX|WITH_SECTION|WITH_WEIGHT|WITH_POSITION|INDEX_LARGE Posts _key,title,content +[[0,0.0,0.0],true] +schema +[ + [ + 0, + 0.0, + 0.0 + ], + { + "plugins": { + }, + "types": { + "Bool": { + "id": 3, + "name": "Bool", + "size": 1, + "can_be_key_type": true, + "can_be_value_type": true + }, + "Float": { + "id": 12, + "name": "Float", + "size": 8, + "can_be_key_type": true, + "can_be_value_type": true + }, + "Int16": { + "id": 6, + "name": "Int16", + "size": 2, + "can_be_key_type": true, + "can_be_value_type": true + }, + "Int32": { + "id": 8, + "name": "Int32", + "size": 4, + "can_be_key_type": true, + "can_be_value_type": true + }, + "Int64": { + "id": 10, + "name": "Int64", + "size": 8, + "can_be_key_type": true, + "can_be_value_type": true + }, + "Int8": { + "id": 4, + "name": "Int8", + "size": 1, + "can_be_key_type": true, + "can_be_value_type": true + }, + "LongText": { + "id": 16, + "name": "LongText", + "size": 2147483648, + "can_be_key_type": false, + "can_be_value_type": false + }, + "Object": { + "id": 2, + "name": "Object", + "size": 8, + "can_be_key_type": true, + "can_be_value_type": true + }, + "ShortText": { + "id": 14, + "name": "ShortText", + "size": 4096, + "can_be_key_type": true, + "can_be_value_type": false + }, + "Text": { + "id": 15, + "name": "Text", + "size": 65536, + "can_be_key_type": false, + "can_be_value_type": false + }, + "Time": { + "id": 13, + "name": "Time", + "size": 8, + "can_be_key_type": true, + "can_be_value_type": true + }, + "TokyoGeoPoint": { + "id": 17, + "name": "TokyoGeoPoint", + "size": 8, + "can_be_key_type": true, + "can_be_value_type": true + }, + "UInt16": { + "id": 7, + "name": "UInt16", + "size": 2, + "can_be_key_type": true, + "can_be_value_type": true + }, + "UInt32": { + "id": 9, + "name": "UInt32", + "size": 4, + "can_be_key_type": true, + "can_be_value_type": true + }, + "UInt64": { + "id": 11, + "name": "UInt64", + "size": 8, + "can_be_key_type": true, + "can_be_value_type": true + }, + "UInt8": { + "id": 5, + "name": "UInt8", + "size": 1, + "can_be_key_type": true, + "can_be_value_type": true + }, + "WGS84GeoPoint": { + "id": 18, + "name": "WGS84GeoPoint", + "size": 8, + "can_be_key_type": true, + "can_be_value_type": true + } + }, + "tokenizers": { + "TokenBigram": { + "id": 67, + "name": "TokenBigram" + }, + "TokenBigramIgnoreBlank": { + "id": 72, + "name": "TokenBigramIgnoreBlank" + }, + "TokenBigramIgnoreBlankSplitSymbol": { + "id": 73, + "name": "TokenBigramIgnoreBlankSplitSymbol" + }, + "TokenBigramIgnoreBlankSplitSymbolAlpha": { + "id": 74, + "name": "TokenBigramIgnoreBlankSplitSymbolAlpha" + }, + "TokenBigramIgnoreBlankSplitSymbolAlphaDigit": { + "id": 75, + "name": "TokenBigramIgnoreBlankSplitSymbolAlphaDigit" + }, + "TokenBigramSplitSymbol": { + "id": 69, + "name": "TokenBigramSplitSymbol" + }, + "TokenBigramSplitSymbolAlpha": { + "id": 70, + "name": "TokenBigramSplitSymbolAlpha" + }, + "TokenBigramSplitSymbolAlphaDigit": { + "id": 71, + "name": "TokenBigramSplitSymbolAlphaDigit" + }, + "TokenDelimit": { + "id": 65, + "name": "TokenDelimit" + }, + "TokenDelimitNull": { + "id": 76, + "name": "TokenDelimitNull" + }, + "TokenMecab": { + "id": 64, + "name": "TokenMecab" + }, + "TokenNgram": { + "id": 78, + "name": "TokenNgram" + }, + "TokenPattern": { + "id": 79, + "name": "TokenPattern" + }, + "TokenRegexp": { + "id": 77, + "name": "TokenRegexp" + }, + "TokenTable": { + "id": 80, + "name": "TokenTable" + }, + "TokenTrigram": { + "id": 68, + "name": "TokenTrigram" + }, + "TokenUnigram": { + "id": 66, + "name": "TokenUnigram" + } + }, + "normalizers": { + "NormalizerAuto": { + "id": 81, + "name": "NormalizerAuto" + }, + "NormalizerNFKC100": { + "id": 83, + "name": "NormalizerNFKC100" + }, + "NormalizerNFKC51": { + "id": 82, + "name": "NormalizerNFKC51" + } + }, + "token_filters": { + "TokenFilterNFKC100": { + "id": 215, + "name": "TokenFilterNFKC100" + } + }, + "tables": { + "Posts": { + "id": 256, + "name": "Posts", + "type": "hash table", + "key_type": { + "id": 14, + "name": "ShortText", + "type": "type" + }, + "value_type": null, + "tokenizer": null, + "normalizer": null, + "token_filters": [ + + ], + "indexes": [ + { + "id": 260, + "full_name": "Terms.index", + "table": "Terms", + "name": "index", + "section": 1 + } + ], + "command": { + "name": "table_create", + "arguments": { + "name": "Posts", + "flags": "TABLE_HASH_KEY", + "key_type": "ShortText" + }, + "command_line": "table_create --name Posts --flags TABLE_HASH_KEY --key_type ShortText" + }, + "columns": { + "content": { + "id": 258, + "name": "content", + "table": "Posts", + "full_name": "Posts.content", + "type": "scalar", + "value_type": { + "id": 15, + "name": "Text", + "type": "type" + }, + "compress": null, + "section": false, + "weight": false, + "position": false, + "sources": [ + + ], + "indexes": [ + { + "id": 260, + "full_name": "Terms.index", + "table": "Terms", + "name": "index", + "section": 3 + } + ], + "command": { + "name": "column_create", + "arguments": { + "table": "Posts", + "name": "content", + "flags": "COLUMN_SCALAR", + "type": "Text" + }, + "command_line": "column_create --table Posts --name content --flags COLUMN_SCALAR --type Text" + } + }, + "title": { + "id": 257, + "name": "title", + "table": "Posts", + "full_name": "Posts.title", + "type": "scalar", + "value_type": { + "id": 14, + "name": "ShortText", + "type": "type" + }, + "compress": null, + "section": false, + "weight": false, + "position": false, + "sources": [ + + ], + "indexes": [ + { + "id": 260, + "full_name": "Terms.index", + "table": "Terms", + "name": "index", + "section": 2 + } + ], + "command": { + "name": "column_create", + "arguments": { + "table": "Posts", + "name": "title", + "flags": "COLUMN_SCALAR", + "type": "ShortText" + }, + "command_line": "column_create --table Posts --name title --flags COLUMN_SCALAR --type ShortText" + } + } + } + }, + "Terms": { + "id": 259, + "name": "Terms", + "type": "patricia trie", + "key_type": { + "id": 14, + "name": "ShortText", + "type": "type" + }, + "value_type": null, + "tokenizer": { + "id": 67, + "name": "TokenBigram", + "options": null + }, + "normalizer": { + "id": 81, + "name": "NormalizerAuto", + "options": null + }, + "token_filters": [ + + ], + "indexes": [ + + ], + "command": { + "name": "table_create", + "arguments": { + "name": "Terms", + "flags": "TABLE_PAT_KEY", + "key_type": "ShortText", + "default_tokenizer": "TokenBigram", + "normalizer": "NormalizerAuto" + }, + "command_line": "table_create --name Terms --flags TABLE_PAT_KEY --key_type ShortText --default_tokenizer TokenBigram --normalizer NormalizerAuto" + }, + "columns": { + "index": { + "id": 260, + "name": "index", + "table": "Terms", + "full_name": "Terms.index", + "type": "index", + "value_type": { + "id": 256, + "name": "Posts", + "type": "reference" + }, + "compress": null, + "section": true, + "weight": true, + "position": true, + "sources": [ + { + "id": null, + "name": "_key", + "table": "Posts", + "full_name": "Posts._key" + }, + { + "id": 257, + "name": "title", + "table": "Posts", + "full_name": "Posts.title" + }, + { + "id": 258, + "name": "content", + "table": "Posts", + "full_name": "Posts.content" + } + ], + "indexes": [ + + ], + "command": { + "name": "column_create", + "arguments": { + "table": "Terms", + "name": "index", + "flags": "COLUMN_INDEX|WITH_SECTION|WITH_WEIGHT|WITH_POSITION|INDEX_LARGE", + "type": "Posts", + "source": "_key,title,content" + }, + "command_line": "column_create --table Terms --name index --flags COLUMN_INDEX|WITH_SECTION|WITH_WEIGHT|WITH_POSITION|INDEX_LARGE --type Posts --source _key,title,content" + } + } + } + } + } + } +] Added: test/command/suite/schema/tables/columns/type/index_large.test (+12 -0) 100644 =================================================================== --- /dev/null +++ test/command/suite/schema/tables/columns/type/index_large.test 2019-04-01 17:34:47 +0900 (4883cf884) @@ -0,0 +1,12 @@ +table_create Posts TABLE_HASH_KEY ShortText +column_create Posts title COLUMN_SCALAR ShortText +column_create Posts content COLUMN_SCALAR Text + +table_create Terms TABLE_PAT_KEY ShortText \ + --default_tokenizer TokenBigram \ + --normalizer NormalizerAuto +column_create Terms index \ + COLUMN_INDEX|WITH_SECTION|WITH_WEIGHT|WITH_POSITION|INDEX_LARGE \ + Posts _key,title,content + +schema
