summaryrefslogtreecommitdiff
path: root/db/btree/btree.h
diff options
context:
space:
mode:
Diffstat (limited to 'db/btree/btree.h')
-rw-r--r--db/btree/btree.h383
1 files changed, 383 insertions, 0 deletions
diff --git a/db/btree/btree.h b/db/btree/btree.h
new file mode 100644
index 0000000000..36d35c998b
--- /dev/null
+++ b/db/btree/btree.h
@@ -0,0 +1,383 @@
+/*-
+ * Copyright (c) 1991, 1993, 1994
+ * The Regents of the University of California. All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Mike Olson.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in the
+ * documentation and/or other materials provided with the distribution.
+ * 3. All advertising materials mentioning features or use of this software
+ * must display the following acknowledgement:
+ * This product includes software developed by the University of
+ * California, Berkeley and its contributors.
+ * 4. Neither the name of the University nor the names of its contributors
+ * may be used to endorse or promote products derived from this software
+ * without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ *
+ * @(#)btree.h 8.11 (Berkeley) 8/17/94
+ */
+
+/* Macros to set/clear/test flags. */
+#define F_SET(p, f) (p)->flags |= (f)
+#define F_CLR(p, f) (p)->flags &= ~(f)
+#define F_ISSET(p, f) ((p)->flags & (f))
+
+#include <mpool.h>
+
+#define DEFMINKEYPAGE (2) /* Minimum keys per page */
+#define MINCACHE (5) /* Minimum cached pages */
+#define MINPSIZE (512) /* Minimum page size */
+
+/*
+ * Page 0 of a btree file contains a copy of the meta-data. This page is also
+ * used as an out-of-band page, i.e. page pointers that point to nowhere point
+ * to page 0. Page 1 is the root of the btree.
+ */
+#define P_INVALID 0 /* Invalid tree page number. */
+#define P_META 0 /* Tree metadata page number. */
+#define P_ROOT 1 /* Tree root page number. */
+
+/*
+ * There are five page layouts in the btree: btree internal pages (BINTERNAL),
+ * btree leaf pages (BLEAF), recno internal pages (RINTERNAL), recno leaf pages
+ * (RLEAF) and overflow pages. All five page types have a page header (PAGE).
+ * This implementation requires that values within structures NOT be padded.
+ * (ANSI C permits random padding.) If your compiler pads randomly you'll have
+ * to do some work to get this package to run.
+ */
+typedef struct _page {
+ pgno_t pgno; /* this page's page number */
+ pgno_t prevpg; /* left sibling */
+ pgno_t nextpg; /* right sibling */
+
+#define P_BINTERNAL 0x01 /* btree internal page */
+#define P_BLEAF 0x02 /* leaf page */
+#define P_OVERFLOW 0x04 /* overflow page */
+#define P_RINTERNAL 0x08 /* recno internal page */
+#define P_RLEAF 0x10 /* leaf page */
+#define P_TYPE 0x1f /* type mask */
+#define P_PRESERVE 0x20 /* never delete this chain of pages */
+ u_int32_t flags;
+
+ indx_t lower; /* lower bound of free space on page */
+ indx_t upper; /* upper bound of free space on page */
+ indx_t linp[1]; /* indx_t-aligned VAR. LENGTH DATA */
+} PAGE;
+
+/* First and next index. */
+#define BTDATAOFF \
+ (sizeof(pgno_t) + sizeof(pgno_t) + sizeof(pgno_t) + \
+ sizeof(u_int32_t) + sizeof(indx_t) + sizeof(indx_t))
+#define NEXTINDEX(p) (((p)->lower - BTDATAOFF) / sizeof(indx_t))
+
+/*
+ * For pages other than overflow pages, there is an array of offsets into the
+ * rest of the page immediately following the page header. Each offset is to
+ * an item which is unique to the type of page. The h_lower offset is just
+ * past the last filled-in index. The h_upper offset is the first item on the
+ * page. Offsets are from the beginning of the page.
+ *
+ * If an item is too big to store on a single page, a flag is set and the item
+ * is a { page, size } pair such that the page is the first page of an overflow
+ * chain with size bytes of item. Overflow pages are simply bytes without any
+ * external structure.
+ *
+ * The page number and size fields in the items are pgno_t-aligned so they can
+ * be manipulated without copying. (This presumes that 32 bit items can be
+ * manipulated on this system.)
+ */
+#define LALIGN(n) (((n) + sizeof(pgno_t) - 1) & ~(sizeof(pgno_t) - 1))
+#define NOVFLSIZE (sizeof(pgno_t) + sizeof(u_int32_t))
+
+/*
+ * For the btree internal pages, the item is a key. BINTERNALs are {key, pgno}
+ * pairs, such that the key compares less than or equal to all of the records
+ * on that page. For a tree without duplicate keys, an internal page with two
+ * consecutive keys, a and b, will have all records greater than or equal to a
+ * and less than b stored on the page associated with a. Duplicate keys are
+ * somewhat special and can cause duplicate internal and leaf page records and
+ * some minor modifications of the above rule.
+ */
+typedef struct _binternal {
+ u_int32_t ksize; /* key size */
+ pgno_t pgno; /* page number stored on */
+#define P_BIGDATA 0x01 /* overflow data */
+#define P_BIGKEY 0x02 /* overflow key */
+ u_char flags;
+ char bytes[1]; /* data */
+} BINTERNAL;
+
+/* Get the page's BINTERNAL structure at index indx. */
+#define GETBINTERNAL(pg, indx) \
+ ((BINTERNAL *)((char *)(pg) + (pg)->linp[indx]))
+
+/* Get the number of bytes in the entry. */
+#define NBINTERNAL(len) \
+ LALIGN(sizeof(u_int32_t) + sizeof(pgno_t) + sizeof(u_char) + (len))
+
+/* Copy a BINTERNAL entry to the page. */
+#define WR_BINTERNAL(p, size, pgno, flags) { \
+ *(u_int32_t *)p = size; \
+ p += sizeof(u_int32_t); \
+ *(pgno_t *)p = pgno; \
+ p += sizeof(pgno_t); \
+ *(u_char *)p = flags; \
+ p += sizeof(u_char); \
+}
+
+/*
+ * For the recno internal pages, the item is a page number with the number of
+ * keys found on that page and below.
+ */
+typedef struct _rinternal {
+ recno_t nrecs; /* number of records */
+ pgno_t pgno; /* page number stored below */
+} RINTERNAL;
+
+/* Get the page's RINTERNAL structure at index indx. */
+#define GETRINTERNAL(pg, indx) \
+ ((RINTERNAL *)((char *)(pg) + (pg)->linp[indx]))
+
+/* Get the number of bytes in the entry. */
+#define NRINTERNAL \
+ LALIGN(sizeof(recno_t) + sizeof(pgno_t))
+
+/* Copy a RINTERAL entry to the page. */
+#define WR_RINTERNAL(p, nrecs, pgno) { \
+ *(recno_t *)p = nrecs; \
+ p += sizeof(recno_t); \
+ *(pgno_t *)p = pgno; \
+}
+
+/* For the btree leaf pages, the item is a key and data pair. */
+typedef struct _bleaf {
+ u_int32_t ksize; /* size of key */
+ u_int32_t dsize; /* size of data */
+ u_char flags; /* P_BIGDATA, P_BIGKEY */
+ char bytes[1]; /* data */
+} BLEAF;
+
+/* Get the page's BLEAF structure at index indx. */
+#define GETBLEAF(pg, indx) \
+ ((BLEAF *)((char *)(pg) + (pg)->linp[indx]))
+
+/* Get the number of bytes in the entry. */
+#define NBLEAF(p) NBLEAFDBT((p)->ksize, (p)->dsize)
+
+/* Get the number of bytes in the user's key/data pair. */
+#define NBLEAFDBT(ksize, dsize) \
+ LALIGN(sizeof(u_int32_t) + sizeof(u_int32_t) + sizeof(u_char) + \
+ (ksize) + (dsize))
+
+/* Copy a BLEAF entry to the page. */
+#define WR_BLEAF(p, key, data, flags) { \
+ *(u_int32_t *)p = key->size; \
+ p += sizeof(u_int32_t); \
+ *(u_int32_t *)p = data->size; \
+ p += sizeof(u_int32_t); \
+ *(u_char *)p = flags; \
+ p += sizeof(u_char); \
+ memmove(p, key->data, key->size); \
+ p += key->size; \
+ memmove(p, data->data, data->size); \
+}
+
+/* For the recno leaf pages, the item is a data entry. */
+typedef struct _rleaf {
+ u_int32_t dsize; /* size of data */
+ u_char flags; /* P_BIGDATA */
+ char bytes[1];
+} RLEAF;
+
+/* Get the page's RLEAF structure at index indx. */
+#define GETRLEAF(pg, indx) \
+ ((RLEAF *)((char *)(pg) + (pg)->linp[indx]))
+
+/* Get the number of bytes in the entry. */
+#define NRLEAF(p) NRLEAFDBT((p)->dsize)
+
+/* Get the number of bytes from the user's data. */
+#define NRLEAFDBT(dsize) \
+ LALIGN(sizeof(u_int32_t) + sizeof(u_char) + (dsize))
+
+/* Copy a RLEAF entry to the page. */
+#define WR_RLEAF(p, data, flags) { \
+ *(u_int32_t *)p = data->size; \
+ p += sizeof(u_int32_t); \
+ *(u_char *)p = flags; \
+ p += sizeof(u_char); \
+ memmove(p, data->data, data->size); \
+}
+
+/*
+ * A record in the tree is either a pointer to a page and an index in the page
+ * or a page number and an index. These structures are used as a cursor, stack
+ * entry and search returns as well as to pass records to other routines.
+ *
+ * One comment about searches. Internal page searches must find the largest
+ * record less than key in the tree so that descents work. Leaf page searches
+ * must find the smallest record greater than key so that the returned index
+ * is the record's correct position for insertion.
+ */
+typedef struct _epgno {
+ pgno_t pgno; /* the page number */
+ indx_t index; /* the index on the page */
+} EPGNO;
+
+typedef struct _epg {
+ PAGE *page; /* the (pinned) page */
+ indx_t index; /* the index on the page */
+} EPG;
+
+/*
+ * About cursors. The cursor (and the page that contained the key/data pair
+ * that it referenced) can be deleted, which makes things a bit tricky. If
+ * there are no duplicates of the cursor key in the tree (i.e. B_NODUPS is set
+ * or there simply aren't any duplicates of the key) we copy the key that it
+ * referenced when it's deleted, and reacquire a new cursor key if the cursor
+ * is used again. If there are duplicates keys, we move to the next/previous
+ * key, and set a flag so that we know what happened. NOTE: if duplicate (to
+ * the cursor) keys are added to the tree during this process, it is undefined
+ * if they will be returned or not in a cursor scan.
+ *
+ * The flags determine the possible states of the cursor:
+ *
+ * CURS_INIT The cursor references *something*.
+ * CURS_ACQUIRE The cursor was deleted, and a key has been saved so that
+ * we can reacquire the right position in the tree.
+ * CURS_AFTER, CURS_BEFORE
+ * The cursor was deleted, and now references a key/data pair
+ * that has not yet been returned, either before or after the
+ * deleted key/data pair.
+ * XXX
+ * This structure is broken out so that we can eventually offer multiple
+ * cursors as part of the DB interface.
+ */
+typedef struct _cursor {
+ EPGNO pg; /* B: Saved tree reference. */
+ DBT key; /* B: Saved key, or key.data == NULL. */
+ recno_t rcursor; /* R: recno cursor (1-based) */
+
+#define CURS_ACQUIRE 0x01 /* B: Cursor needs to be reacquired. */
+#define CURS_AFTER 0x02 /* B: Unreturned cursor after key. */
+#define CURS_BEFORE 0x04 /* B: Unreturned cursor before key. */
+#define CURS_INIT 0x08 /* RB: Cursor initialized. */
+ u_int8_t flags;
+} CURSOR;
+
+/*
+ * The metadata of the tree. The nrecs field is used only by the RECNO code.
+ * This is because the btree doesn't really need it and it requires that every
+ * put or delete call modify the metadata.
+ */
+typedef struct _btmeta {
+ u_int32_t magic; /* magic number */
+ u_int32_t version; /* version */
+ u_int32_t psize; /* page size */
+ u_int32_t free; /* page number of first free page */
+ u_int32_t nrecs; /* R: number of records */
+
+#define SAVEMETA (B_NODUPS | R_RECNO)
+ u_int32_t flags; /* bt_flags & SAVEMETA */
+} BTMETA;
+
+/* The in-memory btree/recno data structure. */
+typedef struct _btree {
+ MPOOL *bt_mp; /* memory pool cookie */
+
+ DB *bt_dbp; /* pointer to enclosing DB */
+
+ EPG bt_cur; /* current (pinned) page */
+ PAGE *bt_pinned; /* page pinned across calls */
+
+ CURSOR bt_cursor; /* cursor */
+
+#define BT_PUSH(t, p, i) { \
+ t->bt_sp->pgno = p; \
+ t->bt_sp->index = i; \
+ ++t->bt_sp; \
+}
+#define BT_POP(t) (t->bt_sp == t->bt_stack ? NULL : --t->bt_sp)
+#define BT_CLR(t) (t->bt_sp = t->bt_stack)
+ EPGNO bt_stack[50]; /* stack of parent pages */
+ EPGNO *bt_sp; /* current stack pointer */
+
+ DBT bt_rkey; /* returned key */
+ DBT bt_rdata; /* returned data */
+
+ int bt_fd; /* tree file descriptor */
+
+ pgno_t bt_free; /* next free page */
+ u_int32_t bt_psize; /* page size */
+ indx_t bt_ovflsize; /* cut-off for key/data overflow */
+ int bt_lorder; /* byte order */
+ /* sorted order */
+ enum { NOT, BACK, FORWARD } bt_order;
+ EPGNO bt_last; /* last insert */
+
+ /* B: key comparison function */
+ int (*bt_cmp) __P((const DBT *, const DBT *));
+ /* B: prefix comparison function */
+ size_t (*bt_pfx) __P((const DBT *, const DBT *));
+ /* R: recno input function */
+ int (*bt_irec) __P((struct _btree *, recno_t));
+
+ FILE *bt_rfp; /* R: record FILE pointer */
+ int bt_rfd; /* R: record file descriptor */
+
+ caddr_t bt_cmap; /* R: current point in mapped space */
+ caddr_t bt_smap; /* R: start of mapped space */
+ caddr_t bt_emap; /* R: end of mapped space */
+ size_t bt_msize; /* R: size of mapped region. */
+
+ recno_t bt_nrecs; /* R: number of records */
+ size_t bt_reclen; /* R: fixed record length */
+ u_char bt_bval; /* R: delimiting byte/pad character */
+
+/*
+ * NB:
+ * B_NODUPS and R_RECNO are stored on disk, and may not be changed.
+ */
+#define B_INMEM 0x00001 /* in-memory tree */
+#define B_METADIRTY 0x00002 /* need to write metadata */
+#define B_MODIFIED 0x00004 /* tree modified */
+#define B_NEEDSWAP 0x00008 /* if byte order requires swapping */
+#define B_RDONLY 0x00010 /* read-only tree */
+
+#define B_NODUPS 0x00020 /* no duplicate keys permitted */
+#define R_RECNO 0x00080 /* record oriented tree */
+
+#define R_CLOSEFP 0x00040 /* opened a file pointer */
+#define R_EOF 0x00100 /* end of input file reached. */
+#define R_FIXLEN 0x00200 /* fixed length records */
+#define R_MEMMAPPED 0x00400 /* memory mapped file. */
+#define R_INMEM 0x00800 /* in-memory file */
+#define R_MODIFIED 0x01000 /* modified file */
+#define R_RDONLY 0x02000 /* read-only file */
+
+#define B_DB_LOCK 0x04000 /* DB_LOCK specified. */
+#define B_DB_SHMEM 0x08000 /* DB_SHMEM specified. */
+#define B_DB_TXN 0x10000 /* DB_TXN specified. */
+ u_int32_t flags;
+} BTREE;
+
+#include "extern.h"