xref: /csrg-svn/lib/libc/db/btree/btree.h (revision 46139)
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  * %sccs.include.redist.c%
9  */
10 
11 /*
12  *  @(#)btree.h	5.1 (Berkeley) 01/23/91
13  */
14 
15 typedef char	*BTREE;		/* should really be (void *) */
16 
17 /* #define	DEBUG */
18 
19 #define RET_ERROR	-1
20 #define RET_SUCCESS	 0
21 #define RET_SPECIAL	 1
22 
23 #ifndef TRUE
24 #define TRUE	1
25 #define FALSE	0
26 #endif /* ndef TRUE */
27 
28 #ifndef NULL
29 #define NULL	0
30 #endif /* ndef NULL */
31 
32 /* libc */
33 extern char *malloc();
34 
35 /* these are defined in lrucache.c */
36 extern char	*lruinit();
37 extern char	*lruget();
38 extern char	*lrugetnew();
39 extern int	lrusync();
40 extern int	lruwrite();
41 extern int	lrurelease();
42 extern void	lrufree();
43 
44 /* these are defined here */
45 extern BTREE	bt_open();
46 extern int	bt_close();
47 extern int	bt_delete();
48 extern int	bt_get();
49 extern int	bt_put();
50 extern int	bt_seq();
51 extern int	bt_sync();
52 
53 /*
54  *  Private types.  What you choose for these depends on how big you
55  *  want to let files get, and how big you want to let pages get.
56  */
57 
58 typedef u_long	index_t;	/* so # bytes on a page fits in a long */
59 typedef u_long	pgno_t;		/* so # of pages in a btree fits in a long */
60 
61 /*
62  *  When we do searches, we push the parent page numbers onto a stack
63  *  as we descend the tree.  This is so that for insertions, we can
64  *  find our way back up to do internal page insertions and splits.
65  */
66 
67 typedef struct BTSTACK {
68 	pgno_t		bts_pgno;
69 	struct BTSTACK	*bts_next;
70 } BTSTACK;
71 
72 /*
73  *  Every btree page has a header that looks like this.  Flags are given
74  *  in the #define's for the F_ flags (see below).
75  */
76 
77 typedef struct BTHEADER {
78 	pgno_t h_pgno;		/* page number of this page */
79 	pgno_t h_prevpg;	/* left sibling */
80 	pgno_t h_nextpg;	/* right sibling */
81 
82 #define F_LEAF		0x01	/* leaf page, contains user data */
83 #define F_CONT		0x02	/* continuation page (large items) */
84 #define F_DIRTY		0x04	/* need to write to disk */
85 #define F_PRESERVE	0x08	/* never delete this chain of pages */
86 
87 	u_long h_flags;		/* page state */
88 	index_t h_lower;	/* lower bound of free space on page */
89 	index_t h_upper;	/* upper bound of free space on page */
90 	index_t h_linp[1];	/* VARIABLE LENGTH DATA AT END OF STRUCT */
91 } BTHEADER;
92 
93 /*
94  *  HTBUCKETs are hash table buckets for looking up pages of in-memory
95  *  btrees by page number.  We use this indirection, rather than direct
96  *  pointers, so that the code for manipulating in-memory trees is the
97  *  same as that for manipulating on-disk trees.
98  */
99 
100 typedef struct HTBUCKET {
101 	pgno_t		ht_pgno;
102 	BTHEADER	*ht_page;
103 	struct HTBUCKET	*ht_next;
104 } HTBUCKET;
105 
106 typedef HTBUCKET	**HTABLE;
107 
108 /* minimum size we'll let a page be */
109 #define MINPSIZE	512
110 
111 /* default cache size, in bytes */
112 #define DEFCACHE	(20 * 1024)
113 
114 /* hash table size for in-memory trees */
115 #define	HTSIZE		128
116 
117 /* generate a hash key from a page number */
118 #define HASHKEY(pgno)	((pgno - 1) % HTSIZE)
119 
120 /*
121  *  Disk btrees have a file descriptor, and may also have an lru buffer
122  *  cache, if the user asked for one.
123  */
124 
125 typedef struct BTDISK {
126 	int	d_fd;
127 	char	*d_cache;
128 } BTDISK;
129 
130 /*
131  *  Cursors keep track of the current location in a sequential scan of
132  *  the database.  Since btrees impose a total ordering on keys, we can
133  *  walk forward or backward through the database from any point.  Cursors
134  *  survive updates to the tree, and can be used to delete a particular
135  *  record.
136  */
137 
138 typedef struct CURSOR {
139 	pgno_t		c_pgno;		/* pgno of current item in scan */
140 	index_t		c_index;	/* index of current item in scan */
141 	char		*c_key;		/* current key, used for updates */
142 
143 #define CRSR_BEFORE	0x01
144 
145 	u_char		c_flags;	/* to handle updates properly */
146 } CURSOR;
147 
148 /*
149  *  The private btree data structure.  The user passes a pointer to one of
150  *  these when we are to manipulate a tree, but the BTREE type is opaque
151  *  to him.
152  */
153 
154 typedef struct BTREEDATA_P {
155 	char		*bt_fname;		/* NULL for in-memory trees */
156 	union {
157 		BTDISK	bt_d;			/* for on-disk btrees */
158 		HTABLE	bt_ht;			/* hash table for mem trees */
159 	} bt_s;
160 	size_t		bt_psize;		/* page size for btree pages */
161 	int		(*bt_compare)();	/* key comparison function */
162 	pgno_t		bt_npages;		/* number of pages in tree */
163 	BTHEADER	*bt_curpage;		/* current page contents */
164 	pgno_t		bt_free;		/* free pg list for big data */
165 	CURSOR		bt_cursor;		/* cursor for scans */
166 	BTSTACK		*bt_stack;		/* parent stack for inserts */
167 	u_long		bt_lorder;		/* byte order (endian.h) */
168 
169 #define BTF_METAOK	0x01	/* meta-data written to start of file */
170 #define BTF_SEQINIT	0x02	/* we have called bt_seq */
171 #define BTF_ISWRITE	0x04	/* tree was opened for write */
172 #define BTF_NODUPS	0x08	/* tree created for unique keys */
173 
174 	u_long		bt_flags;		/* btree state */
175 } BTREEDATA_P;
176 
177 typedef BTREEDATA_P	*BTREE_P;
178 
179 /*
180  *  The first thing in a btree file is a BTMETA structure.  The rest of
181  *  the first page is empty, so that all disk operations are page-aligned.
182  */
183 
184 typedef struct BTMETA {
185 	u_long	m_magic;
186 	u_long	m_version;
187 	size_t	m_psize;
188 	pgno_t	m_free;
189 	u_long	m_flags;
190 	u_long	m_lorder;
191 } BTMETA;
192 
193 #define P_NONE		0		/* invalid page number in tree */
194 #define P_ROOT		1		/* page number of root pg in btree */
195 
196 #define NORELEASE	0		/* don't release a page during write */
197 #define RELEASE		1		/* release a page during write */
198 
199 #define INSERT		0		/* doing an insert operation */
200 #define DELETE		1		/* doing a delete operation */
201 
202 /* get the next free index on a btree page */
203 #define NEXTINDEX(p)	((((int)(p)->h_lower) - ((int)((((char *)(&(p)->h_linp[0]))) - ((char *) (p)))))/(sizeof(index_t)))
204 
205 /* is a BTITEM actually on the btree page? */
206 #define VALIDITEM(t, i)	((i)->bti_index < NEXTINDEX((t)->bt_curpage))
207 
208 /* guarantee longword alignment so structure refs work */
209 #define LONGALIGN(p) (((long)(p) + 3) & ~ 0x03)
210 
211 /* get a particular datum (or idatum) off a page */
212 #define GETDATUM(h,i)	 (((char *) h) + h->h_linp[i])
213 
214 /* is a {key,datum} too big to put on a single page? */
215 #define TOOBIG(t, sz)	(sz >= t->bt_psize / 5)
216 
217 /* is this a disk tree or a memory tree? */
218 #define ISDISK(t)	(t->bt_fname != (char *) NULL)
219 
220 /* does the disk tree use a cache? */
221 #define ISCACHE(t)	(t->bt_s.bt_d.d_cache != (char *) NULL)
222 
223 /*
224  *  DATUMs are for user data -- one appears on leaf pages for every
225  *  tree entry.  The d_bytes[] array contains the key first, then the data.
226  *
227  *  If either the key or the datum is too big to store on a single page,
228  *  a bit is set in the flags entry, and the d_bytes[] array contains a
229  *  pgno pointing to the page at which the data is actually stored.
230  *
231  *  Note on alignment:  every DATUM is guaranteed to be longword aligned
232  *  on the disk page.  In order to force longword alignment of user key
233  *  and data values, we must guarantee that the d_bytes[] array starts
234  *  on a longword boundary.  This is the reason that d_flags is a u_long,
235  *  rather than a u_char (it really only needs to be two bits big).  This
236  *  is necessary because we call the user's comparison function with a
237  *  pointer to the start of the d_bytes array.  We don't need to force
238  *  longword alignment of the data following the key, since that is copied
239  *  to a longword-aligned buffer before being returned to the user.
240  */
241 
242 typedef struct DATUM {
243 	size_t d_ksize;		/* size of key */
244 	size_t d_dsize;		/* size of data */
245 
246 #define D_BIGDATA	0x01	/* indirect datum ptr flag */
247 #define D_BIGKEY	0x02	/* indirect key ptr flag */
248 
249 	u_long d_flags;		/* flags (indirect bit) */
250 	char d_bytes[1];	/* VARIABLE LENGTH DATA AT END OF STRUCT */
251 } DATUM;
252 
253 /* BTITEMs are used to return (page, index, datum) tuples from searches */
254 typedef struct BTITEM {
255 	pgno_t bti_pgno;
256 	index_t bti_index;
257 	DATUM *bti_datum;
258 } BTITEM;
259 
260 /*
261  *  IDATUMs are for data stored on internal pages.  This is the (key, pgno)
262  *  pair, such that key 'key' is the first entry on page 'pgno'.  If our
263  *  internal page contains keys (a) and (b) next to each other, then all
264  *  items >= to (a) and < (b) go on the same page as (a).  There are some
265  *  gotchas with duplicate keys, however.  See the split code for details.
266  *
267  *  If a key is too big to fit on a single page, then the i_bytes[] array
268  *  contains a pgno pointing to the start of a chain that actually stores
269  *  the bytes.  Since items on internal pages are never deleted from the
270  *  tree, these indirect chains are marked as special, so that they won't
271  *  be deleted if the corresponding leaf item is deleted.
272  *
273  *  As for DATUMs, IDATUMs have a u_long flag entry (rather than u_char)
274  *  in order to guarantee that user keys are longword aligned on the disk
275  *  page.
276  */
277 
278 typedef struct IDATUM {
279 	size_t i_size;
280 	pgno_t i_pgno;
281 	u_long i_flags;		/* see DATUM.d_flags, above */
282 	char i_bytes[1];	/* VARIABLE LENGTH DATA AT END OF STRUCT */
283 } IDATUM;
284 
285 /* all private interfaces have a leading _ in their names */
286 extern BTITEM	*_bt_search();
287 extern BTITEM	*_bt_searchr();
288 extern BTHEADER	*_bt_allocpg();
289 extern index_t	_bt_binsrch();
290 extern int	_bt_isonpage();
291 extern BTITEM	*_bt_first();
292 extern int	_bt_release();
293 extern int	_bt_wrtmeta();
294 extern int	_bt_delindir();
295 extern int	_bt_pgout();
296 extern int	_bt_pgin();
297 extern int	_bt_fixscan();
298 extern int	_bt_indirect();
299 extern int	_bt_crsrdel();
300 extern int	_bt_push();
301 extern pgno_t	_bt_pop();
302 extern int	strcmp();
303 
304