db/btree/FreeBSD/btree.h

   1 /*-
   2  * Copyright (c) 1991, 1993, 1994
   3  *      The Regents of the University of California.  All rights reserved.
   4  *
   5  * This code is derived from software contributed to Berkeley by
   6  * Mike Olson.
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions
  10  * are met:
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in the
  15  *    documentation and/or other materials provided with the distribution.
  16  * 3. All advertising materials mentioning features or use of this software
  17  *    must display the following acknowledgement:
  18  *      This product includes software developed by the University of
  19  *      California, Berkeley and its contributors.
  20  * 4. Neither the name of the University nor the names of its contributors
  21  *    may be used to endorse or promote products derived from this software
  22  *    without specific prior written permission.
  23  *
  24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  34  * SUCH DAMAGE.
  35  *
  36  *      @(#)btree.h     8.11 (Berkeley) 8/17/94
  37  * $FreeBSD: src/lib/libc/db/btree/btree.h,v 1.3 2002/03/22 23:41:40 obrien Exp $
  38  */
  39
  40 /* Macros to set/clear/test flags. */
  41 #define F_SET(p, f)     (p)->flags |= (f)
  42 #define F_CLR(p, f)     (p)->flags &= ~(f)
  43 #define F_ISSET(p, f)   ((p)->flags & (f))
  44
  45 #include <mpool.h>
  46
  47 #define DEFMINKEYPAGE   (2)             /* Minimum keys per page */
  48 #define MINCACHE        (5)             /* Minimum cached pages */
  49 #define MINPSIZE        (512)           /* Minimum page size */
  50
  51 /*
  52  * Page 0 of a btree file contains a copy of the meta-data.  This page is also
  53  * used as an out-of-band page, i.e. page pointers that point to nowhere point
  54  * to page 0.  Page 1 is the root of the btree.
  55  */
  56 #define P_INVALID        0              /* Invalid tree page number. */
  57 #define P_META           0              /* Tree metadata page number. */
  58 #define P_ROOT           1              /* Tree root page number. */
  59
  60 /*
  61  * There are five page layouts in the btree: btree internal pages (BINTERNAL),
  62  * btree leaf pages (BLEAF), recno internal pages (RINTERNAL), recno leaf pages
  63  * (RLEAF) and overflow pages.  All five page types have a page header (PAGE).
  64  * This implementation requires that values within structures NOT be padded.
  65  * (ANSI C permits random padding.)  If your compiler pads randomly you'll have
  66  * to do some work to get this package to run.
  67  */
  68 typedef struct _page {
  69         pgno_t  pgno;                   /* this page's page number */
  70         pgno_t  prevpg;                 /* left sibling */
  71         pgno_t  nextpg;                 /* right sibling */
  72
  73 #define P_BINTERNAL     0x01            /* btree internal page */
  74 #define P_BLEAF         0x02            /* leaf page */
  75 #define P_OVERFLOW      0x04            /* overflow page */
  76 #define P_RINTERNAL     0x08            /* recno internal page */
  77 #define P_RLEAF         0x10            /* leaf page */
  78 #define P_TYPE          0x1f            /* type mask */
  79 #define P_PRESERVE      0x20            /* never delete this chain of pages */
  80         u_int32_t flags;
  81
  82         indx_t  lower;                  /* lower bound of free space on page */
  83         indx_t  upper;                  /* upper bound of free space on page */
  84         indx_t  linp[1];                /* indx_t-aligned VAR. LENGTH DATA */
  85 } PAGE;
  86
  87 /* First and next index. */
  88 #define BTDATAOFF                                                       \
  89         (sizeof(pgno_t) + sizeof(pgno_t) + sizeof(pgno_t) +             \
  90             sizeof(u_int32_t) + sizeof(indx_t) + sizeof(indx_t))
  91 #define NEXTINDEX(p)    (((p)->lower - BTDATAOFF) / sizeof(indx_t))
  92
  93 /*
  94  * For pages other than overflow pages, there is an array of offsets into the
  95  * rest of the page immediately following the page header.  Each offset is to
  96  * an item which is unique to the type of page.  The h_lower offset is just
  97  * past the last filled-in index.  The h_upper offset is the first item on the
  98  * page.  Offsets are from the beginning of the page.
  99  *
 100  * If an item is too big to store on a single page, a flag is set and the item
 101  * is a { page, size } pair such that the page is the first page of an overflow
 102  * chain with size bytes of item.  Overflow pages are simply bytes without any
 103  * external structure.
 104  *
 105  * The page number and size fields in the items are pgno_t-aligned so they can
 106  * be manipulated without copying.  (This presumes that 32 bit items can be
 107  * manipulated on this system.)
 108  */
 109 #define LALIGN(n)       (((n) + sizeof(pgno_t) - 1) & ~(sizeof(pgno_t) - 1))
 110 #define NOVFLSIZE       (sizeof(pgno_t) + sizeof(u_int32_t))
 111
 112 /*
 113  * For the btree internal pages, the item is a key.  BINTERNALs are {key, pgno}
 114  * pairs, such that the key compares less than or equal to all of the records
 115  * on that page.  For a tree without duplicate keys, an internal page with two
 116  * consecutive keys, a and b, will have all records greater than or equal to a
 117  * and less than b stored on the page associated with a.  Duplicate keys are
 118  * somewhat special and can cause duplicate internal and leaf page records and
 119  * some minor modifications of the above rule.
 120  */
 121 typedef struct _binternal {
 122         u_int32_t ksize;                /* key size */
 123         pgno_t  pgno;                   /* page number stored on */
 124 #define P_BIGDATA       0x01            /* overflow data */
 125 #define P_BIGKEY        0x02            /* overflow key */
 126         u_char  flags;
 127         char    bytes[1];               /* data */
 128 } BINTERNAL;
 129
 130 /* Get the page's BINTERNAL structure at index indx. */
 131 #define GETBINTERNAL(pg, indx)                                          \
 132         ((BINTERNAL *)((char *)(pg) + (pg)->linp[indx]))
 133
 134 /* Get the number of bytes in the entry. */
 135 #define NBINTERNAL(len)                                                 \
 136         LALIGN(sizeof(u_int32_t) + sizeof(pgno_t) + sizeof(u_char) + (len))
 137
 138 /* Copy a BINTERNAL entry to the page. */
 139 #define WR_BINTERNAL(p, size, pgno, flags) {                            \
 140         *(u_int32_t *)p = size;                                         \
 141         p += sizeof(u_int32_t);                                         \
 142         *(pgno_t *)p = pgno;                                            \
 143         p += sizeof(pgno_t);                                            \
 144         *(u_char *)p = flags;                                           \
 145         p += sizeof(u_char);                                            \
 146 }
 147
 148 /*
 149  * For the recno internal pages, the item is a page number with the number of
 150  * keys found on that page and below.
 151  */
 152 typedef struct _rinternal {
 153         recno_t nrecs;                  /* number of records */
 154         pgno_t  pgno;                   /* page number stored below */
 155 } RINTERNAL;
 156
 157 /* Get the page's RINTERNAL structure at index indx. */
 158 #define GETRINTERNAL(pg, indx)                                          \
 159         ((RINTERNAL *)((char *)(pg) + (pg)->linp[indx]))
 160
 161 /* Get the number of bytes in the entry. */
 162 #define NRINTERNAL                                                      \
 163         LALIGN(sizeof(recno_t) + sizeof(pgno_t))
 164
 165 /* Copy a RINTERAL entry to the page. */
 166 #define WR_RINTERNAL(p, nrecs, pgno) {                                  \
 167         *(recno_t *)p = nrecs;                                          \
 168         p += sizeof(recno_t);                                           \
 169         *(pgno_t *)p = pgno;                                            \
 170 }
 171
 172 /* For the btree leaf pages, the item is a key and data pair. */
 173 typedef struct _bleaf {
 174         u_int32_t       ksize;          /* size of key */
 175         u_int32_t       dsize;          /* size of data */
 176         u_char  flags;                  /* P_BIGDATA, P_BIGKEY */
 177         char    bytes[1];               /* data */
 178 } BLEAF;
 179
 180 /* Get the page's BLEAF structure at index indx. */
 181 #define GETBLEAF(pg, indx)                                              \
 182         ((BLEAF *)((char *)(pg) + (pg)->linp[indx]))
 183
 184 /* Get the number of bytes in the entry. */
 185 #define NBLEAF(p)       NBLEAFDBT((p)->ksize, (p)->dsize)
 186
 187 /* Get the number of bytes in the user's key/data pair. */
 188 #define NBLEAFDBT(ksize, dsize)                                         \
 189         LALIGN(sizeof(u_int32_t) + sizeof(u_int32_t) + sizeof(u_char) + \
 190             (ksize) + (dsize))
 191
 192 /* Copy a BLEAF entry to the page. */
 193 #define WR_BLEAF(p, key, data, flags) {                                 \
 194         *(u_int32_t *)p = key->size;                                    \
 195         p += sizeof(u_int32_t);                                         \
 196         *(u_int32_t *)p = data->size;                                   \
 197         p += sizeof(u_int32_t);                                         \
 198         *(u_char *)p = flags;                                           \
 199         p += sizeof(u_char);                                            \
 200         memmove(p, key->data, key->size);                               \
 201         p += key->size;                                                 \
 202         memmove(p, data->data, data->size);                             \
 203 }
 204
 205 /* For the recno leaf pages, the item is a data entry. */
 206 typedef struct _rleaf {
 207         u_int32_t       dsize;          /* size of data */
 208         u_char  flags;                  /* P_BIGDATA */
 209         char    bytes[1];
 210 } RLEAF;
 211
 212 /* Get the page's RLEAF structure at index indx. */
 213 #define GETRLEAF(pg, indx)                                              \
 214         ((RLEAF *)((char *)(pg) + (pg)->linp[indx]))
 215
 216 /* Get the number of bytes in the entry. */
 217 #define NRLEAF(p)       NRLEAFDBT((p)->dsize)
 218
 219 /* Get the number of bytes from the user's data. */
 220 #define NRLEAFDBT(dsize)                                                \
 221         LALIGN(sizeof(u_int32_t) + sizeof(u_char) + (dsize))
 222
 223 /* Copy a RLEAF entry to the page. */
 224 #define WR_RLEAF(p, data, flags) {                                      \
 225         *(u_int32_t *)p = data->size;                                   \
 226         p += sizeof(u_int32_t);                                         \
 227         *(u_char *)p = flags;                                           \
 228         p += sizeof(u_char);                                            \
 229         memmove(p, data->data, data->size);                             \
 230 }
 231
 232 /*
 233  * A record in the tree is either a pointer to a page and an index in the page
 234  * or a page number and an index.  These structures are used as a cursor, stack
 235  * entry and search returns as well as to pass records to other routines.
 236  *
 237  * One comment about searches.  Internal page searches must find the largest
 238  * record less than key in the tree so that descents work.  Leaf page searches
 239  * must find the smallest record greater than key so that the returned index
 240  * is the record's correct position for insertion.
 241  */
 242 typedef struct _epgno {
 243         pgno_t  pgno;                   /* the page number */
 244         indx_t  index;                  /* the index on the page */
 245 } EPGNO;
 246
 247 typedef struct _epg {
 248         PAGE    *page;                  /* the (pinned) page */
 249         indx_t   index;                 /* the index on the page */
 250 } EPG;
 251
 252 /*
 253  * About cursors.  The cursor (and the page that contained the key/data pair
 254  * that it referenced) can be deleted, which makes things a bit tricky.  If
 255  * there are no duplicates of the cursor key in the tree (i.e. B_NODUPS is set
 256  * or there simply aren't any duplicates of the key) we copy the key that it
 257  * referenced when it's deleted, and reacquire a new cursor key if the cursor
 258  * is used again.  If there are duplicates keys, we move to the next/previous
 259  * key, and set a flag so that we know what happened.  NOTE: if duplicate (to
 260  * the cursor) keys are added to the tree during this process, it is undefined
 261  * if they will be returned or not in a cursor scan.
 262  *
 263  * The flags determine the possible states of the cursor:
 264  *
 265  * CURS_INIT    The cursor references *something*.
 266  * CURS_ACQUIRE The cursor was deleted, and a key has been saved so that
 267  *              we can reacquire the right position in the tree.
 268  * CURS_AFTER, CURS_BEFORE
 269  *              The cursor was deleted, and now references a key/data pair
 270  *              that has not yet been returned, either before or after the
 271  *              deleted key/data pair.
 272  * XXX
 273  * This structure is broken out so that we can eventually offer multiple
 274  * cursors as part of the DB interface.
 275  */
 276 typedef struct _cursor {
 277         EPGNO    pg;                    /* B: Saved tree reference. */
 278         DBT      key;                   /* B: Saved key, or key.data == NULL. */
 279         recno_t  rcursor;               /* R: recno cursor (1-based) */
 280
 281 #define CURS_ACQUIRE    0x01            /*  B: Cursor needs to be reacquired. */
 282 #define CURS_AFTER      0x02            /*  B: Unreturned cursor after key. */
 283 #define CURS_BEFORE     0x04            /*  B: Unreturned cursor before key. */
 284 #define CURS_INIT       0x08            /* RB: Cursor initialized. */
 285         u_int8_t flags;
 286 } CURSOR;
 287
 288 /*
 289  * The metadata of the tree.  The nrecs field is used only by the RECNO code.
 290  * This is because the btree doesn't really need it and it requires that every
 291  * put or delete call modify the metadata.
 292  */
 293 typedef struct _btmeta {
 294         u_int32_t       magic;          /* magic number */
 295         u_int32_t       version;        /* version */
 296         u_int32_t       psize;          /* page size */
 297         u_int32_t       free;           /* page number of first free page */
 298         u_int32_t       nrecs;          /* R: number of records */
 299
 300 #define SAVEMETA        (B_NODUPS | R_RECNO)
 301         u_int32_t       flags;          /* bt_flags & SAVEMETA */
 302 } BTMETA;
 303
 304 /* The in-memory btree/recno data structure. */
 305 typedef struct _btree {
 306         MPOOL    *bt_mp;                /* memory pool cookie */
 307
 308         DB       *bt_dbp;               /* pointer to enclosing DB */
 309
 310         EPG       bt_cur;               /* current (pinned) page */
 311         PAGE     *bt_pinned;            /* page pinned across calls */
 312
 313         CURSOR    bt_cursor;            /* cursor */
 314
 315 #define BT_PUSH(t, p, i) {                                              \
 316         t->bt_sp->pgno = p;                                             \
 317         t->bt_sp->index = i;                                            \
 318         ++t->bt_sp;                                                     \
 319 }
 320 #define BT_POP(t)       (t->bt_sp == t->bt_stack ? NULL : --t->bt_sp)
 321 #define BT_CLR(t)       (t->bt_sp = t->bt_stack)
 322         EPGNO     bt_stack[50];         /* stack of parent pages */
 323         EPGNO    *bt_sp;                /* current stack pointer */
 324
 325         DBT       bt_rkey;              /* returned key */
 326         DBT       bt_rdata;             /* returned data */
 327
 328         int       bt_fd;                /* tree file descriptor */
 329
 330         pgno_t    bt_free;              /* next free page */
 331         u_int32_t bt_psize;             /* page size */
 332         indx_t    bt_ovflsize;          /* cut-off for key/data overflow */
 333         int       bt_lorder;            /* byte order */
 334                                         /* sorted order */
 335         enum { NOT, BACK, FORWARD } bt_order;
 336         EPGNO     bt_last;              /* last insert */
 337
 338                                         /* B: key comparison function */
 339         int     (*bt_cmp)(const DBT *, const DBT *);
 340                                         /* B: prefix comparison function */
 341         size_t  (*bt_pfx)(const DBT *, const DBT *);
 342                                         /* R: recno input function */
 343         int     (*bt_irec)(struct _btree *, recno_t);
 344
 345         FILE     *bt_rfp;               /* R: record FILE pointer */
 346         int       bt_rfd;               /* R: record file descriptor */
 347
 348         caddr_t   bt_cmap;              /* R: current point in mapped space */
 349         caddr_t   bt_smap;              /* R: start of mapped space */
 350         caddr_t   bt_emap;              /* R: end of mapped space */
 351         size_t    bt_msize;             /* R: size of mapped region. */
 352
 353         recno_t   bt_nrecs;             /* R: number of records */
 354         size_t    bt_reclen;            /* R: fixed record length */
 355         u_char    bt_bval;              /* R: delimiting byte/pad character */
 356
 357 /*
 358  * NB:
 359  * B_NODUPS and R_RECNO are stored on disk, and may not be changed.
 360  */
 361 #define B_INMEM         0x00001         /* in-memory tree */
 362 #define B_METADIRTY     0x00002         /* need to write metadata */
 363 #define B_MODIFIED      0x00004         /* tree modified */
 364 #define B_NEEDSWAP      0x00008         /* if byte order requires swapping */
 365 #define B_RDONLY        0x00010         /* read-only tree */
 366
 367 #define B_NODUPS        0x00020         /* no duplicate keys permitted */
 368 #define R_RECNO         0x00080         /* record oriented tree */
 369
 370 #define R_CLOSEFP       0x00040         /* opened a file pointer */
 371 #define R_EOF           0x00100         /* end of input file reached. */
 372 #define R_FIXLEN        0x00200         /* fixed length records */
 373 #define R_MEMMAPPED     0x00400         /* memory mapped file. */
 374 #define R_INMEM         0x00800         /* in-memory file */
 375 #define R_MODIFIED      0x01000         /* modified file */
 376 #define R_RDONLY        0x02000         /* read-only file */
 377
 378 #define B_DB_LOCK       0x04000         /* DB_LOCK specified. */
 379 #define B_DB_SHMEM      0x08000         /* DB_SHMEM specified. */
 380 #define B_DB_TXN        0x10000         /* DB_TXN specified. */
 381         u_int32_t flags;
 382 } BTREE;
 383
 384 #include "extern.h"