X-Git-Url: http://git.indexdata.com/?p=idzebra-moved-to-github.git;a=blobdiff_plain;f=isamb%2Fisamb.c;h=cf96bb515ebc857b1ad53acdaaa9db5210efb20a;hp=7cd0ea936070e9dc030770ad3932f36e0a579c1b;hb=250de4ed23a44f5eb3552db317eef0d0fbe3265c;hpb=161aa7805930f6b91ccea027e4afa4ccce41c379 diff --git a/isamb/isamb.c b/isamb/isamb.c index 7cd0ea9..cf96bb5 100644 --- a/isamb/isamb.c +++ b/isamb/isamb.c @@ -1,8 +1,5 @@ -/* $Id: isamb.c,v 1.89 2006-12-18 23:40:08 adam Exp $ - Copyright (C) 1995-2006 - Index Data ApS - -This file is part of the Zebra server. +/* This file is part of the Zebra server. + Copyright (C) 2004-2013 Index Data Zebra is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free @@ -20,6 +17,9 @@ Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA */ +#if HAVE_CONFIG_H +#include +#endif #include #include #include @@ -50,14 +50,17 @@ struct ISAMB_head { #define INT_ENCODE 1 /* maximum size of encoded buffer */ -#define DST_ITEM_MAX 256 +#define DST_ITEM_MAX 5000 + +/* max page size for _any_ isamb use */ +#define ISAMB_MAX_PAGE 32768 #define ISAMB_MAX_LEVEL 10 /* approx 2*max page + max size of item */ -#define DST_BUF_SIZE (2*4096+300) +#define DST_BUF_SIZE (2*ISAMB_MAX_PAGE+DST_ITEM_MAX+100) /* should be maximum block size of multiple thereof */ -#define ISAMB_CACHE_ENTRY_SIZE 4096 +#define ISAMB_CACHE_ENTRY_SIZE ISAMB_MAX_PAGE /* CAT_MAX: _must_ be power of 2 */ #define CAT_MAX 4 @@ -98,7 +101,7 @@ struct ISAMB_s { int log_io; /* log level for bf_read/bf_write calls */ int log_freelist; /* log level for freelist handling */ zint skipped_numbers; /* on a leaf node */ - zint returned_numbers; + zint returned_numbers; zint skipped_nodes[ISAMB_MAX_LEVEL]; /* [0]=skipped leaves, 1 = higher etc */ zint accessed_nodes[ISAMB_MAX_LEVEL]; /* nodes we did not skip */ zint number_of_int_splits; @@ -133,7 +136,7 @@ struct ISAMB_PP_s { zint total_size; zint no_blocks; zint skipped_numbers; /* on a leaf node */ - zint returned_numbers; + zint returned_numbers; zint skipped_nodes[ISAMB_MAX_LEVEL]; /* [0]=skipped leaves, 1 = higher etc */ zint accessed_nodes[ISAMB_MAX_LEVEL]; /* nodes we did not skip */ struct ISAMB_block **block; @@ -149,8 +152,8 @@ static void encode_ptr(char **dst, zint pos) while (pos > 127) { - *bp++ = (unsigned char) (128 | (pos & 127)); - pos = pos >> 7; + *bp++ = (unsigned char) (128 | (pos & 127)); + pos = pos >> 7; } *bp++ = (unsigned char) pos; *dst = (char *) bp; @@ -182,8 +185,8 @@ static void decode_ptr(const char **src, zint *pos) #else static void decode_ptr(const char **src, zint *pos) { - memcpy(pos, *src, sizeof(*pos)); - (*src) += sizeof(*pos); + memcpy(pos, *src, sizeof(*pos)); + (*src) += sizeof(*pos); } #endif @@ -255,7 +258,7 @@ ISAMB isamb_open2(BFiles bfs, const char *name, int writeflag, ISAMC_M *method, sprintf(fname, "%s%c", name, i+'A'); if (cache) isamb->file[i].bf = bf_open(bfs, fname, ISAMB_CACHE_ENTRY_SIZE, - writeflag); + writeflag); else isamb->file[i].bf = bf_open(bfs, fname, sizes[i], writeflag); @@ -291,17 +294,20 @@ ISAMB isamb_open2(BFiles bfs, const char *name, int writeflag, ISAMC_M *method, if (memcmp(hbuf, "isamb", 5)) { yaz_log(YLOG_WARN, "bad isamb header for file %s", fname); + isamb_close(isamb); return 0; } if (sscanf(hbuf+5, "%d %d %d", &major, &minor, &len) != 3) { yaz_log(YLOG_WARN, "bad isamb header for file %s", fname); + isamb_close(isamb); return 0; } if (major != ISAMB_MAJOR_VERSION) { yaz_log(YLOG_WARN, "bad major version for file %s %d, must be %d", - fname, major, ISAMB_MAJOR_VERSION); + fname, major, ISAMB_MAJOR_VERSION); + isamb_close(isamb); return 0; } for (left = len - sizes[i]; left > 0; left = left - sizes[i]) @@ -309,9 +315,10 @@ ISAMB isamb_open2(BFiles bfs, const char *name, int writeflag, ISAMC_M *method, pos++; if (!bf_read(isamb->file[i].bf, pos, 0, 0, hbuf + pos*sizes[i])) { - yaz_log(YLOG_WARN, "truncated isamb header for " - "file=%s len=%d pos=%d", - fname, len, pos); + yaz_log(YLOG_WARN, "truncated isamb header for " + "file=%s len=%d pos=%d", + fname, len, pos); + isamb_close(isamb); return 0; } } @@ -326,8 +333,12 @@ ISAMB isamb_open2(BFiles bfs, const char *name, int writeflag, ISAMC_M *method, if (isamb->minor_version >= ISAMB_MINOR_VERSION_WITH_ROOT) decode_ptr(&src, &isamb->root_ptr); } - assert (isamb->file[i].head.block_size >= isamb->file[i].head.block_offset); - isamb->file[i].head_dirty = 0; + assert(isamb->file[i].head.block_size >= isamb->file[i].head.block_offset); + /* must rewrite the header if root ptr is in use (bug #1017) */ + if (use_root_ptr && writeflag) + isamb->file[i].head_dirty = 1; + else + isamb->file[i].head_dirty = 0; assert(isamb->file[i].head.block_size == sizes[i]); } #if ISAMB_DEBUG @@ -341,7 +352,7 @@ ISAMB isamb_open(BFiles bfs, const char *name, int writeflag, ISAMC_M *method, { int sizes[CAT_NO]; int i, b_size = ISAMB_MIN_SIZE; - + for (i = 0; ifile[cat].cache_entries) { @@ -368,12 +379,12 @@ static void flush_blocks (ISAMB b, int cat) } } -static int cache_block (ISAMB b, ISAM_P pos, unsigned char *userbuf, int wr) +static int cache_block(ISAMB b, ISAM_P pos, unsigned char *userbuf, int wr) { int cat = (int) (pos&CAT_MASK); - int off = (int) (((pos/CAT_MAX) & - (ISAMB_CACHE_ENTRY_SIZE / b->file[cat].head.block_size - 1)) - * b->file[cat].head.block_size); + int off = (int) (((pos/CAT_MAX) & + (ISAMB_CACHE_ENTRY_SIZE / b->file[cat].head.block_size - 1)) + * b->file[cat].head.block_size); zint norm = pos / (CAT_MASK*ISAMB_CACHE_ENTRY_SIZE / b->file[cat].head.block_size); int no = 0; struct ISAMB_cache_entry **ce, *ce_this = 0, **ce_last = 0; @@ -381,7 +392,7 @@ static int cache_block (ISAMB b, ISAM_P pos, unsigned char *userbuf, int wr) if (!b->cache) return 0; - assert (ISAMB_CACHE_ENTRY_SIZE >= b->file[cat].head.block_size); + assert(ISAMB_CACHE_ENTRY_SIZE >= b->file[cat].head.block_size); for (ce = &b->file[cat].cache_entries; *ce; ce = &(*ce)->next, no++) { ce_last = ce; @@ -389,25 +400,25 @@ static int cache_block (ISAMB b, ISAM_P pos, unsigned char *userbuf, int wr) { ce_this = *ce; *ce = (*ce)->next; /* remove from list */ - + ce_this->next = b->file[cat].cache_entries; /* move to front */ b->file[cat].cache_entries = ce_this; - + if (wr) { - memcpy (ce_this->buf + off, userbuf, - b->file[cat].head.block_size); + memcpy(ce_this->buf + off, userbuf, + b->file[cat].head.block_size); ce_this->dirty = 1; } else - memcpy (userbuf, ce_this->buf + off, - b->file[cat].head.block_size); + memcpy(userbuf, ce_this->buf + off, + b->file[cat].head.block_size); return 1; } } if (no >= b->cache_size) { - assert (ce_last && *ce_last); + assert(ce_last && *ce_last); ce_this = *ce_last; *ce_last = 0; /* remove the last entry from list */ if (ce_this->dirty) @@ -425,34 +436,35 @@ static int cache_block (ISAMB b, ISAM_P pos, unsigned char *userbuf, int wr) ce_this->pos = norm; yaz_log(b->log_io, "bf_read: cache_block"); if (!bf_read(b->file[cat].bf, norm, 0, 0, ce_this->buf)) - memset (ce_this->buf, 0, ISAMB_CACHE_ENTRY_SIZE); + memset(ce_this->buf, 0, ISAMB_CACHE_ENTRY_SIZE); if (wr) { - memcpy (ce_this->buf + off, userbuf, b->file[cat].head.block_size); + memcpy(ce_this->buf + off, userbuf, b->file[cat].head.block_size); ce_this->dirty = 1; } else { ce_this->dirty = 0; - memcpy (userbuf, ce_this->buf + off, b->file[cat].head.block_size); + memcpy(userbuf, ce_this->buf + off, b->file[cat].head.block_size); } return 1; } -void isamb_close (ISAMB isamb) +void isamb_close(ISAMB isamb) { int i; for (i = 0; isamb->accessed_nodes[i]; i++) yaz_log(YLOG_DEBUG, "isamb_close level leaf-%d: "ZINT_FORMAT" read, " - ZINT_FORMAT" skipped", - i, isamb->accessed_nodes[i], isamb->skipped_nodes[i]); + ZINT_FORMAT" skipped", + i, isamb->accessed_nodes[i], isamb->skipped_nodes[i]); yaz_log(YLOG_DEBUG, "isamb_close returned "ZINT_FORMAT" values, " - "skipped "ZINT_FORMAT, - isamb->skipped_numbers, isamb->returned_numbers); + "skipped "ZINT_FORMAT, + isamb->skipped_numbers, isamb->returned_numbers); + for (i = 0; ino_cat; i++) { - flush_blocks (isamb, i); + flush_blocks(isamb, i); if (isamb->file[i].head_dirty) { char hbuf[DST_BUF_SIZE]; @@ -501,9 +513,9 @@ void isamb_close (ISAMB isamb) 0: leader byte, != 0 leaf, == 0, non-leaf 1-2: used size of block 3-7*: number of items and all children - + * Reserve 5 bytes for large block sizes. 1 for small ones .. Number - of items. We can thus have at most 2^40 nodes. + of items. We can thus have at most 2^40 nodes. */ static struct ISAMB_block *open_block(ISAMB b, ISAM_P pos) { @@ -525,7 +537,7 @@ static struct ISAMB_block *open_block(ISAMB b, ISAM_P pos) if (bf_read(b->file[cat].bf, pos/CAT_MAX, 0, 0, p->buf) != 1) { yaz_log(YLOG_FATAL, "isamb: read fail for pos=%ld block=%ld", - (long) pos, (long) pos/CAT_MAX); + (long) pos, (long) pos/CAT_MAX); zebra_exit("isamb:open_block"); } } @@ -535,9 +547,9 @@ static struct ISAMB_block *open_block(ISAMB b, ISAM_P pos) if (p->size < 0) { yaz_log(YLOG_FATAL, "Bad block size %d in pos=" ZINT_FORMAT "\n", - p->size, pos); + p->size, pos); } - assert (p->size >= 0); + assert(p->size >= 0); src = (char*) p->buf + 3; decode_ptr(&src, &p->no_items); @@ -548,7 +560,7 @@ static struct ISAMB_block *open_block(ISAMB b, ISAM_P pos) return p; } -struct ISAMB_block *new_block (ISAMB b, int leaf, int cat) +struct ISAMB_block *new_block(ISAMB b, int leaf, int cat) { struct ISAMB_block *p; @@ -560,28 +572,34 @@ struct ISAMB_block *new_block (ISAMB b, int leaf, int cat) zint block_no; block_no = b->file[cat].head.last_block++; p->pos = block_no * CAT_MAX + cat; + if (b->log_freelist) + yaz_log(b->log_freelist, "got block " + ZINT_FORMAT " from last %d:" ZINT_FORMAT, p->pos, + cat, p->pos/CAT_MAX); } else { p->pos = b->file[cat].head.free_list; assert((p->pos & CAT_MASK) == cat); - if (!cache_block (b, p->pos, p->buf, 0)) + if (!cache_block(b, p->pos, p->buf, 0)) { yaz_log(b->log_io, "bf_read: new_block"); if (!bf_read(b->file[cat].bf, p->pos/CAT_MAX, 0, 0, p->buf)) { yaz_log(YLOG_FATAL, "isamb: read fail for pos=%ld block=%ld", - (long) p->pos/CAT_MAX, (long) p->pos/CAT_MAX); + (long) p->pos/CAT_MAX, (long) p->pos/CAT_MAX); zebra_exit("isamb:new_block"); } } - yaz_log(b->log_freelist, "got block " ZINT_FORMAT " from freelist %d:" ZINT_FORMAT, p->pos, - cat, p->pos/CAT_MAX); - memcpy (&b->file[cat].head.free_list, p->buf, sizeof(zint)); + if (b->log_freelist) + yaz_log(b->log_freelist, "got block " + ZINT_FORMAT " from freelist %d:" ZINT_FORMAT, p->pos, + cat, p->pos/CAT_MAX); + memcpy(&b->file[cat].head.free_list, p->buf, sizeof(zint)); } p->cat = cat; b->file[cat].head_dirty = 1; - memset (p->buf, 0, b->file[cat].head.block_size); + memset(p->buf, 0, b->file[cat].head.block_size); p->bytes = (char*)p->buf + b->file[cat].head.block_offset; p->leaf = leaf; p->size = 0; @@ -593,18 +611,18 @@ struct ISAMB_block *new_block (ISAMB b, int leaf, int cat) return p; } -struct ISAMB_block *new_leaf (ISAMB b, int cat) +struct ISAMB_block *new_leaf(ISAMB b, int cat) { - return new_block (b, 1, cat); + return new_block(b, 1, cat); } -struct ISAMB_block *new_int (ISAMB b, int cat) +struct ISAMB_block *new_int(ISAMB b, int cat) { - return new_block (b, 0, cat); + return new_block(b, 0, cat); } -static void check_block (ISAMB b, struct ISAMB_block *p) +static void check_block(ISAMB b, struct ISAMB_block *p) { assert(b); /* mostly to make the compiler shut up about unused b */ if (p->leaf) @@ -619,9 +637,9 @@ static void check_block (ISAMB b, struct ISAMB_block *p) char *endp = p->bytes + p->size; ISAM_P pos; void *c1 = (*b->method->codec.start)(); - + decode_ptr(&src, &pos); - assert ((pos&CAT_MASK) == p->cat); + assert((pos&CAT_MASK) == p->cat); while (src != endp) { #if INT_ENCODE @@ -632,13 +650,13 @@ static void check_block (ISAMB b, struct ISAMB_block *p) #else zint item_len; decode_item_len(&src, &item_len); - assert (item_len > 0 && item_len < 80); + assert(item_len > 0 && item_len < 128); src += item_len; #endif decode_ptr(&src, &pos); if ((pos&CAT_MASK) != p->cat) { - assert ((pos&CAT_MASK) == p->cat); + assert((pos&CAT_MASK) == p->cat); } } (*b->method->codec.stop)(c1); @@ -652,10 +670,11 @@ void close_block(ISAMB b, struct ISAMB_block *p) if (p->deleted) { yaz_log(b->log_freelist, "release block " ZINT_FORMAT " from freelist %d:" ZINT_FORMAT, - p->pos, p->cat, p->pos/CAT_MAX); - memcpy (p->buf, &b->file[p->cat].head.free_list, sizeof(zint)); + p->pos, p->cat, p->pos/CAT_MAX); + memcpy(p->buf, &b->file[p->cat].head.free_list, sizeof(zint)); b->file[p->cat].head.free_list = p->pos; - if (!cache_block (b, p->pos, p->buf, 1)) + b->file[p->cat].head_dirty = 1; + if (!cache_block(b, p->pos, p->buf, 1)) { yaz_log(b->log_io, "bf_write: close_block (deleted)"); bf_write(b->file[p->cat].bf, p->pos/CAT_MAX, 0, 0, p->buf); @@ -666,8 +685,8 @@ void close_block(ISAMB b, struct ISAMB_block *p) int offset = b->file[p->cat].head.block_offset; int size = p->size + offset; char *dst = (char*)p->buf + 3; - assert (p->size >= 0); - + assert(p->size >= 0); + /* memset becuase encode_ptr usually does not write all bytes */ memset(p->buf, 0, b->file[p->cat].head.block_offset); p->buf[0] = p->leaf; @@ -675,7 +694,7 @@ void close_block(ISAMB b, struct ISAMB_block *p) p->buf[2] = size >> 8; encode_ptr(&dst, p->no_items); check_block(b, p); - if (!cache_block (b, p->pos, p->buf, 1)) + if (!cache_block(b, p->pos, p->buf, 1)) { yaz_log(b->log_io, "bf_write: close_block"); bf_write(b->file[p->cat].bf, p->pos/CAT_MAX, 0, 0, p->buf); @@ -686,17 +705,17 @@ void close_block(ISAMB b, struct ISAMB_block *p) xfree(p); } -int insert_sub (ISAMB b, struct ISAMB_block **p, - void *new_item, int *mode, - ISAMC_I *stream, - struct ISAMB_block **sp, - void *sub_item, int *sub_size, - const void *max_item); - -int insert_int (ISAMB b, struct ISAMB_block *p, void *lookahead_item, - int *mode, - ISAMC_I *stream, struct ISAMB_block **sp, - void *split_item, int *split_size, const void *last_max_item) +int insert_sub(ISAMB b, struct ISAMB_block **p, + void *new_item, int *mode, + ISAMC_I *stream, + struct ISAMB_block **sp, + void *sub_item, int *sub_size, + const void *max_item); + +int insert_int(ISAMB b, struct ISAMB_block *p, void *lookahead_item, + int *mode, + ISAMC_I *stream, struct ISAMB_block **sp, + void *split_item, int *split_size, const void *last_max_item) { char *startp = p->bytes; const char *src = startp; @@ -726,11 +745,11 @@ int insert_int (ISAMB b, struct ISAMB_block *p, void *lookahead_item, if (d > 0) { sub_p1 = open_block(b, pos); - assert (sub_p1); + assert(sub_p1); diff_terms -= sub_p1->no_items; - more = insert_sub (b, &sub_p1, lookahead_item, mode, - stream, &sub_p2, - sub_item, &sub_size, file_item_buf); + more = insert_sub(b, &sub_p1, lookahead_item, mode, + stream, &sub_p2, + sub_item, &sub_size, file_item_buf); diff_terms += sub_p1->no_items; src = src0; break; @@ -742,11 +761,11 @@ int insert_int (ISAMB b, struct ISAMB_block *p, void *lookahead_item, if (d > 0) { sub_p1 = open_block(b, pos); - assert (sub_p1); + assert(sub_p1); diff_terms -= sub_p1->no_items; - more = insert_sub (b, &sub_p1, lookahead_item, mode, - stream, &sub_p2, - sub_item, &sub_size, src); + more = insert_sub(b, &sub_p1, lookahead_item, mode, + stream, &sub_p2, + sub_item, &sub_size, src); diff_terms += sub_p1->no_items; src = src0; break; @@ -759,10 +778,10 @@ int insert_int (ISAMB b, struct ISAMB_block *p, void *lookahead_item, { /* we reached the end. So lookahead > last item */ sub_p1 = open_block(b, pos); - assert (sub_p1); + assert(sub_p1); diff_terms -= sub_p1->no_items; - more = insert_sub (b, &sub_p1, lookahead_item, mode, stream, &sub_p2, - sub_item, &sub_size, last_max_item); + more = insert_sub(b, &sub_p1, lookahead_item, mode, stream, &sub_p2, + sub_item, &sub_size, last_max_item); diff_terms += sub_p1->no_items; } if (sub_p2) @@ -780,18 +799,18 @@ int insert_int (ISAMB b, struct ISAMB_block *p, void *lookahead_item, #if INT_ENCODE const char *sub_item_ptr = sub_item; #endif - assert (sub_size < 80 && sub_size > 1); + assert(sub_size < DST_ITEM_MAX && sub_size > 1); + + memcpy(dst, startp, src - startp); - memcpy (dst, startp, src - startp); - dst += src - startp; #if INT_ENCODE (*b->method->codec.reset)(c1); (*b->method->codec.encode)(c1, &dst, &sub_item_ptr); #else - encode_item_len (&dst, sub_size); /* sub length and item */ - memcpy (dst, sub_item, sub_size); + encode_item_len(&dst, sub_size); /* sub length and item */ + memcpy(dst, sub_item, sub_size); dst += sub_size; #endif @@ -799,16 +818,16 @@ int insert_int (ISAMB b, struct ISAMB_block *p, void *lookahead_item, if (endp - src) /* remaining data */ { - memcpy (dst, src, endp - src); + memcpy(dst, src, endp - src); dst += endp - src; } p->size = dst - dst_buf; - assert (p->size >= 0); + assert(p->size >= 0); if (p->size <= b->file[p->cat].head.block_max) { /* it fits OK in this block */ - memcpy (startp, dst_buf, dst - dst_buf); + memcpy(startp, dst_buf, dst - dst_buf); close_block(b, sub_p2); } @@ -827,7 +846,7 @@ int insert_int (ISAMB b, struct ISAMB_block *p, void *lookahead_item, const char *half; src = dst_buf; endp = dst; - + b->number_of_int_splits++; p->dirty = 1; @@ -867,7 +886,7 @@ int insert_int (ISAMB b, struct ISAMB_block *p, void *lookahead_item, } /* p is first half */ p_new_size = src - dst_buf; - memcpy (p->bytes, dst_buf, p_new_size); + memcpy(p->bytes, dst_buf, p_new_size); #if INT_ENCODE file_item = file_item_buf; @@ -878,13 +897,13 @@ int insert_int (ISAMB b, struct ISAMB_block *p, void *lookahead_item, #else decode_item_len(&src, &split_size_tmp); *split_size = (int) split_size_tmp; - memcpy (split_item, src, *split_size); + memcpy(split_item, src, *split_size); src += *split_size; #endif /* *sp is second half */ - *sp = new_int (b, p->cat); + *sp = new_int(b, p->cat); (*sp)->size = endp - src; - memcpy ((*sp)->bytes, src, (*sp)->size); + memcpy((*sp)->bytes, src, (*sp)->size); p->size = p_new_size; @@ -899,11 +918,11 @@ int insert_int (ISAMB b, struct ISAMB_block *p, void *lookahead_item, return more; } -int insert_leaf (ISAMB b, struct ISAMB_block **sp1, void *lookahead_item, - int *lookahead_mode, ISAMC_I *stream, - struct ISAMB_block **sp2, - void *sub_item, int *sub_size, - const void *max_item) +int insert_leaf(ISAMB b, struct ISAMB_block **sp1, void *lookahead_item, + int *lookahead_mode, ISAMC_I *stream, + struct ISAMB_block **sp2, + void *sub_item, int *sub_size, + const void *max_item) { struct ISAMB_block *p = *sp1; char *endp = 0; @@ -929,7 +948,7 @@ int insert_leaf (ISAMB b, struct ISAMB_block **sp1, void *lookahead_item, { char file_item_buf[DST_ITEM_MAX]; char *file_item = file_item_buf; - + src = p->bytes; endp = p->bytes + p->size; (*b->method->codec.decode)(c1, &file_item, &src); @@ -939,12 +958,12 @@ int insert_leaf (ISAMB b, struct ISAMB_block **sp1, void *lookahead_item, char *lookahead_next; char *dst_0 = dst; int d = -1; - + if (lookahead_item) d = (*b->method->compare_item)(file_item_buf, lookahead_item); - + /* d now holds comparison between existing file item and - lookahead item + lookahead item d = 0: equal d > 0: lookahead before file d < 0: lookahead after file @@ -960,14 +979,22 @@ int insert_leaf (ISAMB b, struct ISAMB_block **sp1, void *lookahead_item, assert(*lookahead_mode); } } + else if (d == 0 && *lookahead_mode == 2) + { + /* For mode == 2, we insert the new key anyway - even + though the comparison is 0. */ + dst_item = lookahead_item; + p->dirty = 1; + } else dst_item = file_item_buf; - if (!*lookahead_mode && d == 0) + if (d == 0 && !*lookahead_mode) { - /* it's a deletion and they match so there is nothing to be - inserted anyway .. But mark the thing bad (file item - was part of input.. The item will not be part of output */ + /* it's a deletion and they match so there is nothing + to be inserted anyway .. But mark the thing dirty + (file item was part of input.. The item will not be + part of output */ p->dirty = 1; } else if (!half1 && dst > mid_cut) @@ -978,11 +1005,11 @@ int insert_leaf (ISAMB b, struct ISAMB_block **sp1, void *lookahead_item, /* encode the resulting item */ (*b->method->codec.encode)(c2, &dst, &dst_item); - + cut_item_size = dst_item - dst_item_0; assert(cut_item_size > 0); - memcpy (cut_item_buf, dst_item_0, cut_item_size); - + memcpy(cut_item_buf, dst_item_0, cut_item_size); + half2 = dst; no_items_1 = no_items; no_items++; @@ -995,7 +1022,7 @@ int insert_leaf (ISAMB b, struct ISAMB_block **sp1, void *lookahead_item, } /* now move "pointers" .. result has been encoded .. */ - if (d > 0) + if (d > 0) { /* we must move the lookahead pointer */ @@ -1022,7 +1049,7 @@ int insert_leaf (ISAMB b, struct ISAMB_block **sp1, void *lookahead_item, leaf. Mark it as "gone" */ lookahead_item = 0; } - + p->dirty = 1; } } @@ -1063,7 +1090,7 @@ int insert_leaf (ISAMB b, struct ISAMB_block **sp1, void *lookahead_item, char *dst_item; const char *src = lookahead_item; char *dst_0 = dst; - + /* if we have a lookahead item, we stop if we exceed the value of it */ if (max_item && (*b->method->compare_item)(max_item, lookahead_item) <= 0) @@ -1081,13 +1108,13 @@ int insert_leaf (ISAMB b, struct ISAMB_block **sp1, void *lookahead_item, { const char *src_0 = src; half1 = dst; /* candidate for splitting */ - + (*b->method->codec.encode)(c2, &dst, &src); - + cut_item_size = src - src_0; assert(cut_item_size > 0); - memcpy (cut_item_buf, src_0, cut_item_size); - + memcpy(cut_item_buf, src_0, cut_item_size); + no_items_1 = no_items; half2 = dst; } @@ -1111,7 +1138,7 @@ int insert_leaf (ISAMB b, struct ISAMB_block **sp1, void *lookahead_item, } } new_size = dst - dst_buf; - if (p && p->cat != b->no_cat-1 && + if (p && p->cat != b->no_cat-1 && new_size > b->file[p->cat].head.block_max) { /* non-btree block will be removed */ @@ -1128,26 +1155,26 @@ int insert_leaf (ISAMB b, struct ISAMB_block **sp1, void *lookahead_item, break; if (i == b->no_cat) i = b->no_cat - 1; - p = new_leaf (b, i); + p = new_leaf(b, i); } if (new_size > b->file[p->cat].head.block_max) { char *first_dst; const char *cut_item = cut_item_buf; - assert (half1); - assert (half2); + assert(half1); + assert(half2); assert(cut_item_size > 0); - + /* first half */ p->size = half1 - dst_buf; assert(p->size <= b->file[p->cat].head.block_max); - memcpy (p->bytes, dst_buf, half1 - dst_buf); + memcpy(p->bytes, dst_buf, half1 - dst_buf); p->no_items = no_items_1; /* second half */ - *sp2 = new_leaf (b, p->cat); + *sp2 = new_leaf(b, p->cat); (*b->method->codec.reset)(c2); @@ -1157,19 +1184,19 @@ int insert_leaf (ISAMB b, struct ISAMB_block **sp1, void *lookahead_item, (*b->method->codec.encode)(c2, &first_dst, &cut_item); - memcpy (first_dst, half2, dst - half2); + memcpy(first_dst, half2, dst - half2); (*sp2)->size = (first_dst - (*sp2)->bytes) + (dst - half2); assert((*sp2)->size <= b->file[p->cat].head.block_max); (*sp2)->no_items = no_items - no_items_1; (*sp2)->dirty = 1; p->dirty = 1; - memcpy (sub_item, cut_item_buf, cut_item_size); + memcpy(sub_item, cut_item_buf, cut_item_size); *sub_size = cut_item_size; } else { - memcpy (p->bytes, dst_buf, dst - dst_buf); + memcpy(p->bytes, dst_buf, dst - dst_buf); p->size = new_size; p->no_items = no_items; } @@ -1179,22 +1206,22 @@ int insert_leaf (ISAMB b, struct ISAMB_block **sp1, void *lookahead_item, return more; } -int insert_sub (ISAMB b, struct ISAMB_block **p, void *new_item, - int *mode, - ISAMC_I *stream, - struct ISAMB_block **sp, - void *sub_item, int *sub_size, - const void *max_item) +int insert_sub(ISAMB b, struct ISAMB_block **p, void *new_item, + int *mode, + ISAMC_I *stream, + struct ISAMB_block **sp, + void *sub_item, int *sub_size, + const void *max_item) { if (!*p || (*p)->leaf) - return insert_leaf (b, p, new_item, mode, stream, sp, sub_item, - sub_size, max_item); - else - return insert_int (b, *p, new_item, mode, stream, sp, sub_item, + return insert_leaf(b, p, new_item, mode, stream, sp, sub_item, sub_size, max_item); + else + return insert_int(b, *p, new_item, mode, stream, sp, sub_item, + sub_size, max_item); } -int isamb_unlink (ISAMB b, ISAM_P pos) +int isamb_unlink(ISAMB b, ISAM_P pos) { struct ISAMB_block *p1; @@ -1211,7 +1238,7 @@ int isamb_unlink (ISAMB b, ISAM_P pos) #endif decode_ptr(&src, &sub_p); isamb_unlink(b, sub_p); - + while (src != p1->bytes + p1->size) { #if INT_ENCODE @@ -1262,14 +1289,14 @@ void isamb_merge(ISAMB b, ISAM_P *pos, ISAMC_I *stream) struct ISAMB_block *p = 0, *sp = 0; char sub_item[DST_ITEM_MAX]; int sub_size; - + if (*pos) p = open_block(b, *pos); - more = insert_sub (b, &p, item_buf, &i_mode, stream, &sp, - sub_item, &sub_size, 0); + more = insert_sub(b, &p, item_buf, &i_mode, stream, &sp, + sub_item, &sub_size, 0); if (sp) { /* increase level of tree by one */ - struct ISAMB_block *p2 = new_int (b, p->cat); + struct ISAMB_block *p2 = new_int(b, p->cat); char *dst = p2->bytes + p2->size; #if INT_ENCODE void *c1 = (*b->method->codec.start)(); @@ -1277,17 +1304,17 @@ void isamb_merge(ISAMB b, ISAM_P *pos, ISAMC_I *stream) #endif encode_ptr(&dst, p->pos); - assert (sub_size < 80 && sub_size > 1); + assert(sub_size < DST_ITEM_MAX && sub_size > 1); #if INT_ENCODE (*b->method->codec.reset)(c1); (*b->method->codec.encode)(c1, &dst, &sub_item_ptr); #else - encode_item_len (&dst, sub_size); - memcpy (dst, sub_item, sub_size); + encode_item_len(&dst, sub_size); + memcpy(dst, sub_item, sub_size); dst += sub_size; #endif encode_ptr(&dst, sp->pos); - + p2->size = dst - p2->bytes; p2->no_items = p->no_items + sp->no_items; *pos = p2->pos; /* return new super page */ @@ -1347,7 +1374,7 @@ ISAMB_PP isamb_pp_open_x(ISAMB isamb, ISAM_P pos, int *level, int scope) decode_ptr(&src, &pos); p->offset = src - p->bytes; pp->level++; - pp->accessed_nodes[pp->level]++; + pp->accessed_nodes[pp->level]++; } pp->block[pp->level+1] = 0; pp->maxlevel = pp->level; @@ -1356,7 +1383,7 @@ ISAMB_PP isamb_pp_open_x(ISAMB isamb, ISAM_P pos, int *level, int scope) return pp; } -ISAMB_PP isamb_pp_open (ISAMB isamb, ISAM_P pos, int scope) +ISAMB_PP isamb_pp_open(ISAMB isamb, ISAM_P pos, int scope) { return isamb_pp_open_x(isamb, pos, 0, scope); } @@ -1366,14 +1393,14 @@ void isamb_pp_close_x(ISAMB_PP pp, zint *size, zint *blocks) int i; if (!pp) return; - yaz_log(YLOG_DEBUG, "isamb_pp_close lev=%d returned "ZINT_FORMAT" values, " - "skipped "ZINT_FORMAT, - pp->maxlevel, pp->skipped_numbers, pp->returned_numbers); + yaz_log(YLOG_DEBUG, "isamb_pp_close lev=%d returned "ZINT_FORMAT" values, " + "skipped "ZINT_FORMAT, + pp->maxlevel, pp->skipped_numbers, pp->returned_numbers); for (i = pp->maxlevel; i>=0; i--) if (pp->skipped_nodes[i] || pp->accessed_nodes[i]) yaz_log(YLOG_DEBUG, "isamb_pp_close level leaf-%d: " - ZINT_FORMAT" read, "ZINT_FORMAT" skipped", i, - pp->accessed_nodes[i], pp->skipped_nodes[i]); + ZINT_FORMAT" read, "ZINT_FORMAT" skipped", i, + pp->accessed_nodes[i], pp->skipped_nodes[i]); pp->isamb->skipped_numbers += pp->skipped_numbers; pp->isamb->returned_numbers += pp->returned_numbers; for (i = pp->maxlevel; i>=0; i--) @@ -1391,21 +1418,21 @@ void isamb_pp_close_x(ISAMB_PP pp, zint *size, zint *blocks) xfree(pp); } -int isamb_block_info (ISAMB isamb, int cat) +int isamb_block_info(ISAMB isamb, int cat) { if (cat >= 0 && cat < isamb->no_cat) return isamb->file[cat].head.block_size; return -1; } -void isamb_pp_close (ISAMB_PP pp) +void isamb_pp_close(ISAMB_PP pp) { isamb_pp_close_x(pp, 0, 0); } /* simple recursive dumper .. */ -static void isamb_dump_r (ISAMB b, ISAM_P pos, void (*pr)(const char *str), - int level) +static void isamb_dump_r(ISAMB b, ISAM_P pos, void (*pr)(const char *str), + int level) { char buf[1024]; char prefix_str[1024]; @@ -1439,7 +1466,7 @@ static void isamb_dump_r (ISAMB b, ISAM_P pos, void (*pr)(const char *str), p->offset = src - (char*) p->bytes; isamb_dump_r(b, sub, pr, level+1); - + while (p->offset < p->size) { #if INT_ENCODE @@ -1456,11 +1483,11 @@ static void isamb_dump_r (ISAMB b, ISAM_P pos, void (*pr)(const char *str), src += item_len; #endif decode_ptr(&src, &sub); - + p->offset = src - (char*) p->bytes; - + isamb_dump_r(b, sub, pr, level+1); - } + } } close_block(b, p); } @@ -1477,428 +1504,23 @@ int isamb_pp_read(ISAMB_PP pp, void *buf) } -static int isamb_pp_on_right_node(ISAMB_PP pp, int level, const void *untilbuf) -{ /* looks one node higher to see if we should be on this node at all */ - /* useful in backing off quickly, and in avoiding tail descends */ - /* call with pp->level to begin with */ - struct ISAMB_block *p; - int cmp; - const char *src; - ISAMB b = pp->isamb; - - assert(level >= 0); - if (level == 0) - { -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isamb_pp_on_right returning true for root"); -#endif - return 1; /* we can never skip the root node */ - } - level--; - p = pp->block[level]; - assert(p->offset <= p->size); - if (p->offset < p->size) - { -#if INT_ENCODE - char file_item_buf[DST_ITEM_MAX]; - char *file_item = file_item_buf; - void *c1 = (*b->method->codec.start)(); - assert(p->offset > 0); - src = p->bytes + p->offset; - (*b->method->codec.decode)(c1, &file_item, &src); - (*b->method->codec.stop)(c1); - cmp = (*b->method->compare_item)(untilbuf, file_item_buf); -#else - zint item_len; - assert(p->offset > 0); - src = p->bytes + p->offset; - decode_item_len(&src, &item_len); -#if ISAMB_DEBUG - (*b->method->codec.log_item)(YLOG_DEBUG, untilbuf, "on_leaf: until"); - (*b->method->codec.log_item)(YLOG_DEBUG, src, "on_leaf: value"); -#endif - cmp = (*b->method->compare_item)(untilbuf, src); -#endif - if (cmp < pp->scope) - { /* cmp<2 */ -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isamb_pp_on_right returning true " - "cmp=%d lev=%d ofs=%d", cmp, level, p->offset); -#endif - return 1; - } - else - { -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isamb_pp_on_right returning false " - "cmp=%d lev=%d ofs=%d", cmp, level, p->offset); -#endif - return 0; - } - } - else { -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isamb_pp_on_right at tail, looking higher " - "lev=%d", level); -#endif - return isamb_pp_on_right_node(pp, level, untilbuf); - } -} /* isamb_pp_on_right_node */ - -static int isamb_pp_read_on_leaf(ISAMB_PP pp, void *buf) -{ - /* reads the next item on the current leaf, returns 0 if end of leaf*/ - struct ISAMB_block *p = pp->block[pp->level]; - char *dst; - const char *src; - assert(pp); - assert(buf); - if (p->offset == p->size) - { -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isamb_pp_read_on_leaf returning 0 on " - "node %d", p->pos); -#endif - return 0; /* at end of leaf */ - } - src = p->bytes + p->offset; - dst = buf; - (*pp->isamb->method->codec.decode)(p->decodeClientData, &dst, &src); - p->offset = src - (char*) p->bytes; -#if ISAMB_DEBUG - (*pp->isamb->method->codec.log_item)(YLOG_DEBUG, buf, - "read_on_leaf returning 1"); -#endif - pp->returned_numbers++; - return 1; -} /* read_on_leaf */ - -static int isamb_pp_forward_on_leaf(ISAMB_PP pp, void *buf, const void *untilbuf) -{ /* forwards on the current leaf, returns 0 if not found */ - int cmp; - int skips = 0; - while (1) - { - if (!isamb_pp_read_on_leaf(pp, buf)) - return 0; - /* FIXME - this is an extra function call, inline the read? */ - cmp=(*pp->isamb->method->compare_item)(untilbuf, buf); - if (cmp scope) - { /* cmp<2 found a good one */ -#if ISAMB_DEBUG - if (skips) - yaz_log(YLOG_DEBUG, "isam_pp_fwd_on_leaf skipped %d items", skips); -#endif - pp->returned_numbers++; - return 1; - } - if (!skips) - if (!isamb_pp_on_right_node(pp, pp->level, untilbuf)) - return 0; /* never mind the rest of this leaf */ - pp->skipped_numbers++; - skips++; - } -} /* forward_on_leaf */ - -static int isamb_pp_climb_level(ISAMB_PP pp, ISAM_P *pos) -{ /* climbs higher in the tree, until finds a level with data left */ - /* returns the node to (consider to) descend to in *pos) */ - struct ISAMB_block *p = pp->block[pp->level]; - const char *src; -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isamb_pp_climb_level starting " - "at level %d node %d ofs=%d sz=%d", - pp->level, p->pos, p->offset, p->size); -#endif - assert(pp->level >= 0); - assert(p->offset <= p->size); - if (pp->level==0) - { -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isamb_pp_climb_level returning 0 at root"); -#endif - return 0; - } - assert(pp->level>0); - close_block(pp->isamb, pp->block[pp->level]); - pp->block[pp->level] = 0; - (pp->level)--; - p = pp->block[pp->level]; -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isamb_pp_climb_level climbed to level %d node %d ofs=%d", - pp->level, p->pos, p->offset); -#endif - assert(!p->leaf); - assert(p->offset <= p->size); - if (p->offset == p->size) - { - /* we came from the last pointer, climb on */ - if (!isamb_pp_climb_level(pp, pos)) - return 0; - p = pp->block[pp->level]; - } - else - { -#if INT_ENCODE - char file_item_buf[DST_ITEM_MAX]; - char *file_item = file_item_buf; - ISAMB b = pp->isamb; - void *c1 = (*b->method->codec.start)(); -#else - zint item_len; -#endif - /* skip the child we just came from */ -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isam_pp_climb_level: skipping lev=%d ofs=%d sz=%d", - pp->level, p->offset, p->size); -#endif - assert (p->offset < p->size); - src = p->bytes + p->offset; -#if INT_ENCODE - (*b->method->codec.decode)(c1, &file_item, &src); - (*b->method->codec.stop)(c1); -#else - decode_item_len(&src, &item_len); - src += item_len; -#endif - decode_ptr(&src, pos); - p->offset = src - (char *)p->bytes; - - } - return 1; -} /* climb_level */ - - -static zint isamb_pp_forward_unode(ISAMB_PP pp, zint pos, const void *untilbuf) -{ /* scans a upper node until it finds a child <= untilbuf */ - /* pp points to the key value, as always. pos is the child read from */ - /* the buffer */ - /* if all values are too small, returns the last child in the node */ - /* FIXME - this can be detected, and avoided by looking at the */ - /* parent node, but that gets messy. Presumably the cost is */ - /* pretty low anyway */ - ISAMB b = pp->isamb; - struct ISAMB_block *p = pp->block[pp->level]; - const char *src = p->bytes + p->offset; - int cmp; - zint nxtpos; -#if ISAMB_DEBUG - int skips = 0; - yaz_log(YLOG_DEBUG, "isamb_pp_forward_unode starting " - "at level %d node %d ofs=%di sz=%d", - pp->level, p->pos, p->offset, p->size); -#endif - assert(!p->leaf); - assert(p->offset <= p->size); - if (p->offset == p->size) - { -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isamb_pp_forward_unode returning at end " - "at level %d node %d ofs=%di sz=%d", - pp->level, p->pos, p->offset, p->size); -#endif - return pos; /* already at the end of it */ - } - while(p->offset < p->size) - { -#if INT_ENCODE - char file_item_buf[DST_ITEM_MAX]; - char *file_item = file_item_buf; - void *c1 = (*b->method->codec.start)(); - (*b->method->codec.decode)(c1, &file_item, &src); - (*b->method->codec.stop)(c1); - cmp = (*b->method->compare_item)(untilbuf, file_item_buf); -#else - zint item_len; - decode_item_len(&src, &item_len); - cmp = (*b->method->compare_item)(untilbuf, src); - src += item_len; -#endif - decode_ptr(&src, &nxtpos); - if (cmpscope) /* cmp<2 */ - { -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isamb_pp_forward_unode returning a hit " - "at level %d node %d ofs=%d sz=%d", - pp->level, p->pos, p->offset, p->size); -#endif - return pos; - } /* found one */ - pos = nxtpos; - p->offset = src-(char*)p->bytes; - (pp->skipped_nodes[pp->maxlevel - pp->level -1])++; -#if ISAMB_DEBUG - skips++; -#endif - } -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isamb_pp_forward_unode returning at tail " - "at level %d node %d ofs=%d sz=%d skips=%d", - pp->level, p->pos, p->offset, p->size, skips); -#endif - return pos; /* that's the last one in the line */ - -} /* forward_unode */ - -static void isamb_pp_descend_to_leaf(ISAMB_PP pp, ISAM_P pos, - const void *untilbuf) -{ /* climbs down the tree, from pos, to the leftmost leaf */ - struct ISAMB_block *p = pp->block[pp->level]; - const char *src; - assert(!p->leaf); -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isamb_pp_descend_to_leaf " - "starting at lev %d node %d ofs=%d lf=%d u=%p", - pp->level, p->pos, p->offset, p->leaf, untilbuf); -#endif - if (untilbuf) - pos = isamb_pp_forward_unode(pp, pos, untilbuf); - ++(pp->level); - assert(pos); - p = open_block(pp->isamb, pos); - pp->block[pp->level] = p; - ++(pp->accessed_nodes[pp->maxlevel-pp->level]); - ++(pp->no_blocks); -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isamb_pp_descend_to_leaf " - "got lev %d node %d lf=%d", - pp->level, p->pos, p->leaf); -#endif - if (p->leaf) - return; - assert (p->offset==0); - src = p->bytes + p->offset; - decode_ptr(&src, &pos); - p->offset = src-(char*)p->bytes; - isamb_pp_descend_to_leaf(pp, pos, untilbuf); -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isamb_pp_descend_to_leaf " - "returning at lev %d node %d ofs=%d lf=%d", - pp->level, p->pos, p->offset, p->leaf); -#endif -} /* descend_to_leaf */ - -static int isamb_pp_find_next_leaf(ISAMB_PP pp) -{ /* finds the next leaf by climbing up and down */ - ISAM_P pos; - if (!isamb_pp_climb_level(pp, &pos)) - return 0; - isamb_pp_descend_to_leaf(pp, pos, 0); - return 1; -} - -static int isamb_pp_climb_desc(ISAMB_PP pp, const void *untilbuf) -{ /* climbs up and descends to a leaf where values >= *untilbuf are found */ - ISAM_P pos; -#if ISAMB_DEBUG - struct ISAMB_block *p = pp->block[pp->level]; - yaz_log(YLOG_DEBUG, "isamb_pp_climb_desc starting " - "at level %d node %d ofs=%d sz=%d", - pp->level, p->pos, p->offset, p->size); -#endif - if (!isamb_pp_climb_level(pp, &pos)) - return 0; - /* see if it would pay to climb one higher */ - if (!isamb_pp_on_right_node(pp, pp->level, untilbuf)) - if (!isamb_pp_climb_level(pp, &pos)) - return 0; - isamb_pp_descend_to_leaf(pp, pos, untilbuf); -#if ISAMB_DEBUG - p = pp->block[pp->level]; - yaz_log(YLOG_DEBUG, "isamb_pp_climb_desc done " - "at level %d node %d ofs=%d sz=%d", - pp->level, p->pos, p->offset, p->size); -#endif - return 1; -} /* climb_desc */ - -int isamb_pp_forward (ISAMB_PP pp, void *buf, const void *untilbuf) -{ -#if ISAMB_DEBUG - struct ISAMB_block *p = pp->block[pp->level]; - assert(p->leaf); - yaz_log(YLOG_DEBUG, "isamb_pp_forward starting " - "at level %d node %d ofs=%d sz=%d u=%p sc=%d", - pp->level, p->pos, p->offset, p->size, untilbuf, scope); -#endif - if (untilbuf) - { - if (isamb_pp_forward_on_leaf(pp, buf, untilbuf)) - { -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isamb_pp_forward (f) returning (A) " - "at level %d node %d ofs=%d sz=%d", - pp->level, p->pos, p->offset, p->size); -#endif - return 1; - } - if (! isamb_pp_climb_desc(pp, untilbuf)) - { -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isamb_pp_forward (f) returning notfound (B) " - "at level %d node %d ofs=%d sz=%d", - pp->level, p->pos, p->offset, p->size); -#endif - return 0; /* could not find a leaf */ - } - do { - if (isamb_pp_forward_on_leaf(pp, buf, untilbuf)) - { -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isamb_pp_forward (f) returning (c) " - "at level %d node %d ofs=%d sz=%d", - pp->level, p->pos, p->offset, p->size); -#endif - return 1; - } - } while (isamb_pp_find_next_leaf(pp)); - return 0; /* could not find at all */ - } - else { /* no untilbuf, a straight read */ - /* FIXME - this should be moved - * directly into the pp_read */ - /* keeping here now, to keep same - * interface as the old fwd */ - if (isamb_pp_read_on_leaf(pp, buf)) - { -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isamb_pp_forward (read) returning (D) " - "at level %d node %d ofs=%d sz=%d", - pp->level, p->pos, p->offset, p->size); -#endif - return 1; - } - if (isamb_pp_find_next_leaf(pp)) - { -#if ISAMB_DEBUG - yaz_log(YLOG_DEBUG, "isamb_pp_forward (read) returning (E) " - "at level %d node %d ofs=%d sz=%d", - pp->level, p->pos, p->offset, p->size); -#endif - return isamb_pp_read_on_leaf(pp, buf); - } - else - return 0; - } -} /* isam_pp_forward (new version) */ - void isamb_pp_pos(ISAMB_PP pp, double *current, double *total) { /* return an estimate of the current position and of the total number of */ /* occureences in the isam tree, based on the current leaf */ assert(total); assert(current); - + /* if end-of-stream PP may not be leaf */ *total = (double) (pp->block[0]->no_items); *current = (double) pp->returned_numbers; #if ISAMB_DEBUG yaz_log(YLOG_LOG, "isamb_pp_pos returning: cur= %0.1f tot=%0.1f rn=" - ZINT_FORMAT, *current, *total, pp->returned_numbers); + ZINT_FORMAT, *current, *total, pp->returned_numbers); #endif } -int isamb_pp_forward2(ISAMB_PP pp, void *buf, const void *untilb) +int isamb_pp_forward(ISAMB_PP pp, void *buf, const void *untilb) { char *dst = buf; const char *src; @@ -1906,7 +1528,7 @@ int isamb_pp_forward2(ISAMB_PP pp, void *buf, const void *untilb) ISAMB b = pp->isamb; if (!p) return 0; -again: + again: while (p->offset == p->size) { ISAM_P pos; @@ -1922,24 +1544,24 @@ again: { if (pp->level == 0) return 0; - close_block (pp->isamb, pp->block[pp->level]); + close_block(pp->isamb, pp->block[pp->level]); pp->block[pp->level] = 0; (pp->level)--; p = pp->block[pp->level]; - assert (!p->leaf); + assert(!p->leaf); } assert(!p->leaf); src = p->bytes + p->offset; - + #if INT_ENCODE c1 = (*b->method->codec.start)(); (*b->method->codec.decode)(c1, &file_item, &src); #else - decode_ptr (&src, &item_len); + decode_ptr(&src, &item_len); src += item_len; -#endif - decode_ptr (&src, &pos); +#endif + decode_ptr(&src, &pos); p->offset = src - (char*) p->bytes; src = p->bytes + p->offset; @@ -1954,18 +1576,18 @@ again: file_item = file_item_buf; (*b->method->codec.reset)(c1); (*b->method->codec.decode)(c1, &file_item, &src); - if ((*b->method->compare_item)(untilb, file_item_buf) <= 1) + if ((*b->method->compare_item)(untilb, file_item_buf) < pp->scope) { src = src_0; break; } #else decode_item_len(&src, &item_len); - if ((*b->method->compare_item)(untilb, src) <= 1) + if ((*b->method->compare_item)(untilb, src) < pp->scope) break; src += item_len; #endif - decode_ptr (&src, &pos); + decode_ptr(&src, &pos); p->offset = src - (char*) p->bytes; } @@ -1973,22 +1595,22 @@ again: while (1) { - pp->block[pp->level] = p = open_block (pp->isamb, pos); + pp->block[pp->level] = p = open_block(pp->isamb, pos); pp->total_size += p->size; pp->no_blocks++; - - if (p->leaf) + + if (p->leaf) { break; } - + src = p->bytes + p->offset; while(1) { - decode_ptr (&src, &pos); + decode_ptr(&src, &pos); p->offset = src - (char*) p->bytes; - + if (!untilb || p->offset == p->size) break; assert(p->offset < p->size); @@ -1997,14 +1619,14 @@ again: file_item = file_item_buf; (*b->method->codec.reset)(c1); (*b->method->codec.decode)(c1, &file_item, &src); - if ((*b->method->compare_item)(untilb, file_item_buf) <= 1) + if ((*b->method->compare_item)(untilb, file_item_buf) < pp->scope) { src = src_0; break; } #else - decode_ptr (&src, &item_len); - if ((*b->method->compare_item)(untilb, src) <= 1) + decode_ptr(&src, &item_len); + if ((*b->method->compare_item)(untilb, src) <= pp->scope) break; src += item_len; #endif @@ -2015,19 +1637,20 @@ again: (*b->method->codec.stop)(c1); #endif } - assert (p->offset < p->size); - assert (p->leaf); + assert(p->offset < p->size); + assert(p->leaf); while(1) { char *dst0 = dst; src = p->bytes + p->offset; (*pp->isamb->method->codec.decode)(p->decodeClientData, &dst, &src); p->offset = src - (char*) p->bytes; - if (!untilb || (*pp->isamb->method->compare_item)(untilb, dst0) <= 1) + if (!untilb || (*pp->isamb->method->compare_item)(untilb, dst0) < pp->scope) break; dst = dst0; if (p->offset == p->size) goto again; } + pp->returned_numbers++; return 1; } @@ -2055,6 +1678,7 @@ void isamb_set_root_ptr(ISAMB b, zint root_ptr) /* * Local variables: * c-basic-offset: 4 + * c-file-style: "Stroustrup" * indent-tabs-mode: nil * End: * vim: shiftwidth=4 tabstop=8 expandtab