Home | History | Annotate | Line # | Download | only in btree
btree.h revision 1.1
      1 /*-
      2  * Copyright (c) 1990 The Regents of the University of California.
      3  * 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 
     37 /*
     38  *  @(#)btree.h	5.2 (Berkeley) 2/22/91
     39  */
     40 
     41 typedef char	*BTREE;		/* should really be (void *) */
     42 
     43 /* #define	DEBUG */
     44 
     45 #define RET_ERROR	-1
     46 #define RET_SUCCESS	 0
     47 #define RET_SPECIAL	 1
     48 
     49 #ifndef TRUE
     50 #define TRUE	1
     51 #define FALSE	0
     52 #endif /* ndef TRUE */
     53 
     54 #ifndef NULL
     55 #define NULL	0
     56 #endif /* ndef NULL */
     57 
     58 /* these are defined in lrucache.c */
     59 extern char	*lruinit();
     60 extern char	*lruget();
     61 extern char	*lrugetnew();
     62 extern int	lrusync();
     63 extern int	lruwrite();
     64 extern int	lrurelease();
     65 extern void	lrufree();
     66 
     67 /* these are defined here */
     68 extern BTREE	bt_open();
     69 extern int	bt_close();
     70 extern int	bt_delete();
     71 extern int	bt_get();
     72 extern int	bt_put();
     73 extern int	bt_seq();
     74 extern int	bt_sync();
     75 
     76 /*
     77  *  Private types.  What you choose for these depends on how big you
     78  *  want to let files get, and how big you want to let pages get.
     79  */
     80 
     81 typedef u_long	index_t;	/* so # bytes on a page fits in a long */
     82 typedef u_long	pgno_t;		/* so # of pages in a btree fits in a long */
     83 
     84 /*
     85  *  When we do searches, we push the parent page numbers onto a stack
     86  *  as we descend the tree.  This is so that for insertions, we can
     87  *  find our way back up to do internal page insertions and splits.
     88  */
     89 
     90 typedef struct BTSTACK {
     91 	pgno_t		bts_pgno;
     92 	struct BTSTACK	*bts_next;
     93 } BTSTACK;
     94 
     95 /*
     96  *  Every btree page has a header that looks like this.  Flags are given
     97  *  in the #define's for the F_ flags (see below).
     98  */
     99 
    100 typedef struct BTHEADER {
    101 	pgno_t h_pgno;		/* page number of this page */
    102 	pgno_t h_prevpg;	/* left sibling */
    103 	pgno_t h_nextpg;	/* right sibling */
    104 
    105 #define F_LEAF		0x01	/* leaf page, contains user data */
    106 #define F_CONT		0x02	/* continuation page (large items) */
    107 #define F_DIRTY		0x04	/* need to write to disk */
    108 #define F_PRESERVE	0x08	/* never delete this chain of pages */
    109 
    110 	u_long h_flags;		/* page state */
    111 	index_t h_lower;	/* lower bound of free space on page */
    112 	index_t h_upper;	/* upper bound of free space on page */
    113 	index_t h_linp[1];	/* VARIABLE LENGTH DATA AT END OF STRUCT */
    114 } BTHEADER;
    115 
    116 /*
    117  *  HTBUCKETs are hash table buckets for looking up pages of in-memory
    118  *  btrees by page number.  We use this indirection, rather than direct
    119  *  pointers, so that the code for manipulating in-memory trees is the
    120  *  same as that for manipulating on-disk trees.
    121  */
    122 
    123 typedef struct HTBUCKET {
    124 	pgno_t		ht_pgno;
    125 	BTHEADER	*ht_page;
    126 	struct HTBUCKET	*ht_next;
    127 } HTBUCKET;
    128 
    129 typedef HTBUCKET	**HTABLE;
    130 
    131 /* minimum size we'll let a page be */
    132 #define MINPSIZE	512
    133 
    134 /* default cache size, in bytes */
    135 #define DEFCACHE	(20 * 1024)
    136 
    137 /* hash table size for in-memory trees */
    138 #define	HTSIZE		128
    139 
    140 /* generate a hash key from a page number */
    141 #define HASHKEY(pgno)	((pgno - 1) % HTSIZE)
    142 
    143 /*
    144  *  Disk btrees have a file descriptor, and may also have an lru buffer
    145  *  cache, if the user asked for one.
    146  */
    147 
    148 typedef struct BTDISK {
    149 	int	d_fd;
    150 	char	*d_cache;
    151 } BTDISK;
    152 
    153 /*
    154  *  Cursors keep track of the current location in a sequential scan of
    155  *  the database.  Since btrees impose a total ordering on keys, we can
    156  *  walk forward or backward through the database from any point.  Cursors
    157  *  survive updates to the tree, and can be used to delete a particular
    158  *  record.
    159  */
    160 
    161 typedef struct CURSOR {
    162 	pgno_t		c_pgno;		/* pgno of current item in scan */
    163 	index_t		c_index;	/* index of current item in scan */
    164 	char		*c_key;		/* current key, used for updates */
    165 
    166 #define CRSR_BEFORE	0x01
    167 
    168 	u_char		c_flags;	/* to handle updates properly */
    169 } CURSOR;
    170 
    171 /*
    172  *  The private btree data structure.  The user passes a pointer to one of
    173  *  these when we are to manipulate a tree, but the BTREE type is opaque
    174  *  to him.
    175  */
    176 
    177 typedef struct BTREEDATA_P {
    178 	char		*bt_fname;		/* NULL for in-memory trees */
    179 	union {
    180 		BTDISK	bt_d;			/* for on-disk btrees */
    181 		HTABLE	bt_ht;			/* hash table for mem trees */
    182 	} bt_s;
    183 	size_t		bt_psize;		/* page size for btree pages */
    184 	int		(*bt_compare)();	/* key comparison function */
    185 	pgno_t		bt_npages;		/* number of pages in tree */
    186 	BTHEADER	*bt_curpage;		/* current page contents */
    187 	pgno_t		bt_free;		/* free pg list for big data */
    188 	CURSOR		bt_cursor;		/* cursor for scans */
    189 	BTSTACK		*bt_stack;		/* parent stack for inserts */
    190 	u_long		bt_lorder;		/* byte order (endian.h) */
    191 
    192 #define BTF_METAOK	0x01	/* meta-data written to start of file */
    193 #define BTF_SEQINIT	0x02	/* we have called bt_seq */
    194 #define BTF_ISWRITE	0x04	/* tree was opened for write */
    195 #define BTF_NODUPS	0x08	/* tree created for unique keys */
    196 
    197 	u_long		bt_flags;		/* btree state */
    198 } BTREEDATA_P;
    199 
    200 typedef BTREEDATA_P	*BTREE_P;
    201 
    202 /*
    203  *  The first thing in a btree file is a BTMETA structure.  The rest of
    204  *  the first page is empty, so that all disk operations are page-aligned.
    205  */
    206 
    207 typedef struct BTMETA {
    208 	u_long	m_magic;
    209 	u_long	m_version;
    210 	size_t	m_psize;
    211 	pgno_t	m_free;
    212 	u_long	m_flags;
    213 	u_long	m_lorder;
    214 } BTMETA;
    215 
    216 #define P_NONE		0		/* invalid page number in tree */
    217 #define P_ROOT		1		/* page number of root pg in btree */
    218 
    219 #define NORELEASE	0		/* don't release a page during write */
    220 #define RELEASE		1		/* release a page during write */
    221 
    222 #define INSERT		0		/* doing an insert operation */
    223 #define DELETE		1		/* doing a delete operation */
    224 
    225 /* get the next free index on a btree page */
    226 #define NEXTINDEX(p)	((((int)(p)->h_lower) - ((int)((((char *)(&(p)->h_linp[0]))) - ((char *) (p)))))/(sizeof(index_t)))
    227 
    228 /* is a BTITEM actually on the btree page? */
    229 #define VALIDITEM(t, i)	((i)->bti_index < NEXTINDEX((t)->bt_curpage))
    230 
    231 /* guarantee longword alignment so structure refs work */
    232 #define LONGALIGN(p) (((long)(p) + 3) & ~ 0x03)
    233 
    234 /* get a particular datum (or idatum) off a page */
    235 #define GETDATUM(h,i)	 (((char *) h) + h->h_linp[i])
    236 
    237 /* is a {key,datum} too big to put on a single page? */
    238 #define TOOBIG(t, sz)	(sz >= t->bt_psize / 5)
    239 
    240 /* is this a disk tree or a memory tree? */
    241 #define ISDISK(t)	(t->bt_fname != (char *) NULL)
    242 
    243 /* does the disk tree use a cache? */
    244 #define ISCACHE(t)	(t->bt_s.bt_d.d_cache != (char *) NULL)
    245 
    246 /*
    247  *  DATUMs are for user data -- one appears on leaf pages for every
    248  *  tree entry.  The d_bytes[] array contains the key first, then the data.
    249  *
    250  *  If either the key or the datum is too big to store on a single page,
    251  *  a bit is set in the flags entry, and the d_bytes[] array contains a
    252  *  pgno pointing to the page at which the data is actually stored.
    253  *
    254  *  Note on alignment:  every DATUM is guaranteed to be longword aligned
    255  *  on the disk page.  In order to force longword alignment of user key
    256  *  and data values, we must guarantee that the d_bytes[] array starts
    257  *  on a longword boundary.  This is the reason that d_flags is a u_long,
    258  *  rather than a u_char (it really only needs to be two bits big).  This
    259  *  is necessary because we call the user's comparison function with a
    260  *  pointer to the start of the d_bytes array.  We don't need to force
    261  *  longword alignment of the data following the key, since that is copied
    262  *  to a longword-aligned buffer before being returned to the user.
    263  */
    264 
    265 typedef struct DATUM {
    266 	size_t d_ksize;		/* size of key */
    267 	size_t d_dsize;		/* size of data */
    268 
    269 #define D_BIGDATA	0x01	/* indirect datum ptr flag */
    270 #define D_BIGKEY	0x02	/* indirect key ptr flag */
    271 
    272 	u_long d_flags;		/* flags (indirect bit) */
    273 	char d_bytes[1];	/* VARIABLE LENGTH DATA AT END OF STRUCT */
    274 } DATUM;
    275 
    276 /* BTITEMs are used to return (page, index, datum) tuples from searches */
    277 typedef struct BTITEM {
    278 	pgno_t bti_pgno;
    279 	index_t bti_index;
    280 	DATUM *bti_datum;
    281 } BTITEM;
    282 
    283 /*
    284  *  IDATUMs are for data stored on internal pages.  This is the (key, pgno)
    285  *  pair, such that key 'key' is the first entry on page 'pgno'.  If our
    286  *  internal page contains keys (a) and (b) next to each other, then all
    287  *  items >= to (a) and < (b) go on the same page as (a).  There are some
    288  *  gotchas with duplicate keys, however.  See the split code for details.
    289  *
    290  *  If a key is too big to fit on a single page, then the i_bytes[] array
    291  *  contains a pgno pointing to the start of a chain that actually stores
    292  *  the bytes.  Since items on internal pages are never deleted from the
    293  *  tree, these indirect chains are marked as special, so that they won't
    294  *  be deleted if the corresponding leaf item is deleted.
    295  *
    296  *  As for DATUMs, IDATUMs have a u_long flag entry (rather than u_char)
    297  *  in order to guarantee that user keys are longword aligned on the disk
    298  *  page.
    299  */
    300 
    301 typedef struct IDATUM {
    302 	size_t i_size;
    303 	pgno_t i_pgno;
    304 	u_long i_flags;		/* see DATUM.d_flags, above */
    305 	char i_bytes[1];	/* VARIABLE LENGTH DATA AT END OF STRUCT */
    306 } IDATUM;
    307 
    308 /* all private interfaces have a leading _ in their names */
    309 extern BTITEM	*_bt_search();
    310 extern BTITEM	*_bt_searchr();
    311 extern BTHEADER	*_bt_allocpg();
    312 extern index_t	_bt_binsrch();
    313 extern int	_bt_isonpage();
    314 extern BTITEM	*_bt_first();
    315 extern int	_bt_release();
    316 extern int	_bt_wrtmeta();
    317 extern int	_bt_delindir();
    318 extern int	_bt_pgout();
    319 extern int	_bt_pgin();
    320 extern int	_bt_fixscan();
    321 extern int	_bt_indirect();
    322 extern int	_bt_crsrdel();
    323 extern int	_bt_push();
    324 extern pgno_t	_bt_pop();
    325 extern int	strcmp();
    326 
    327