xref: /csrg-svn/lib/libc/db/hash/hash.h (revision 46367)
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  * Margo Seltzer.
7  *
8  * %sccs.include.redist.c%
9  *
10  *	@(#)hash.h	5.1 (Berkeley) 02/12/91
11  */
12 
13 /* Operations */
14 typedef enum { HASH_GET, HASH_PUT, HASH_PUTNEW, HASH_DELETE,
15 		HASH_FIRST, HASH_NEXT } ACTION;
16 
17 /* Buffer Management structures */
18 typedef struct _bufhead BUFHEAD;
19 
20 struct _bufhead {
21     BUFHEAD	*prev;		/* LRU links */
22     BUFHEAD	*next;		/* LRU links */
23     BUFHEAD	*ovfl;		/* Overflow page buffer header */
24     int		addr;		/* Address of this page */
25     char	*page;		/* Actual page data */
26     char	flags;	/* Combination of BUF_MOD, BUF_DISK, BUF_BUCKET */
27 };
28 
29 /* Flag Values */
30 #define	BUF_MOD		0x0001
31 #define BUF_DISK	0x0002
32 #define	BUF_BUCKET	0x0004
33 
34 #define IS_BUCKET(X)	(X & BUF_BUCKET)
35 
36 typedef BUFHEAD	**SEGMENT;
37 
38 /* Hash Table Information */
39 typedef struct hashhdr {	/* Disk resident portion */
40 	int magic;	/* Magic NO for hash tables */
41 	int version;	/* Version ID */
42 	long lorder;	/* Byte Order */
43 	int bsize;	/* Bucket/Page Size */
44 	int bshift;	/* Bucket shift */
45 	int dsize;	/* Directory Size */
46 	int ssize;	/* Segment Size */
47 	int sshift;	/* Segment shift */
48 	int max_bucket;	/* ID of Maximum bucket in use */
49 	int high_mask;	/* Mask to modulo into entire table */
50 	int low_mask;	/* Mask to modulo into lower half of table */
51 	int ffactor;	/* Fill factor */
52 	int nkeys;	/* Number of keys in hash table */
53 	int hdrpages;	/* Size of table header */
54 	int h_charkey;	/* value of hash(CHARKEY) */
55 # define NCACHED		32	/* number of bit maps and spare points*/
56 	int spares[NCACHED];	/* spare pages for overflow */
57 	u_short bitmaps[NCACHED];	/* address of overflow page bitmaps */
58 } HASHHDR;
59 
60 typedef struct htab {	/* Memory resident data structure */
61 	HASHHDR hdr;	/* Header */
62 	int nsegs;	/* Number of allocated segments */
63 	int exsegs;	/* Number of extra allocated segments */
64 	int (*hash)();	/* Hash Function */
65 	int flags;	/* Flag values */
66 	int fp;		/* File pointer */
67 	char *tmp_buf;	/* Temporary Buffer for BIG data */
68 	char *tmp_key;	/* Temporary Buffer for BIG keys */
69 	BUFHEAD *cpage;	/* Current page */
70 	int cbucket;	/* Current bucket */
71 	int cndx;  	/* Index of next item on cpage */
72 	int errno;	/* Error Number -- for DBM compatability */
73 	int new_file;	/* Indicates whether fd is backing store or no */
74 	int save_file;	/* Indicates whether we need to flush file at exit */
75 	u_long *mapp[NCACHED];	/* Pointers to page maps */
76 	int nbufs;	/* Number of buffers left to allocate */
77 	BUFHEAD	bufhead; /* Header of buffer lru list */
78 	SEGMENT	 *dir;	/* Hash Bucket directory */
79 } HTAB;
80 
81 
82 /*
83  * Constants
84  */
85 #define	MAX_BSIZE		65536	/* 2^16 */
86 #define MIN_BUFFERS		6
87 #define MINHDRSIZE		512
88 #define DEF_BUFSIZE		65536	/* 64 K */
89 #define DEF_BUCKET_SIZE	256
90 #define DEF_BUCKET_SHIFT	8	/* log2(BUCKET) */
91 #define DEF_SEGSIZE		256
92 #define DEF_SEGSIZE_SHIFT		8      /* log2(SEGSIZE)	 */
93 #define DEF_DIRSIZE		256
94 #define DEF_FFACTOR		5
95 #define SPLTMAX		8
96 #define CHARKEY		"%$sniglet^&"
97 #define NUMKEY			1038583
98 #define VERSION_NO		3
99 #define BYTE_SHIFT		3
100 #define INT_TO_BYTE		2
101 #define INT_BYTE_SHIFT		5
102 #define ALL_SET		((unsigned)0xFFFFFFFF)
103 #define ALL_CLEAR		0
104 
105 
106 #define PTROF(X)	((BUFHEAD *)((unsigned)(X)&~0x3))
107 #define ISMOD(X)	((unsigned)(X)&0x1)
108 #define DOMOD(X)	(X = (char *)( (unsigned)X | 0x1))
109 #define ISDISK(X)	((unsigned)(X)&0x2)
110 #define DODISK(X)	(X = (char *)( (unsigned)X | 0x2))
111 
112 #define BITS_PER_MAP    32
113 
114 /* Given the address of the beginning of a big map, clear/set the nth bit */
115 
116 #define CLRBIT(A,N) ((A)[N/BITS_PER_MAP] &= ~(1<<(N%BITS_PER_MAP)))
117 #define SETBIT(A,N) ((A)[N/BITS_PER_MAP] |= (1<<(N%BITS_PER_MAP)))
118 #define ISSET(A,N) ((A)[N/BITS_PER_MAP] & (1<<(N%BITS_PER_MAP)))
119 
120 /* Overflow management */
121 /*
122 	Overflow page numbers are allocated per split point.
123 	At each doubling of the table, we can allocate extra
124 	pages.  So, an overflow page number has the top 5 bits
125 	indicate which split point and the lower 11 bits indicate
126 	which page at that split point is indicated (pages within
127 	split points are numberered starting with 1).
128 
129 
130 */
131 
132 #define SPLITSHIFT	11
133 #define SPLITMASK	0x7FF
134 #define SPLITNUM(N)	(((unsigned)N) >> SPLITSHIFT)
135 #define OPAGENUM(N)	(N & SPLITMASK)
136 #define	OADDR_OF(S,O)	((S << SPLITSHIFT) + O)
137 
138 #define BUCKET_TO_PAGE(B) \
139 	B + hashp->HDRPAGES + (B ? hashp->SPARES[__log2(B+1)-1] : 0)
140 #define OADDR_TO_PAGE(B) 	\
141 	BUCKET_TO_PAGE ( (1 << SPLITNUM(B)) -1 ) + OPAGENUM(B);
142 
143 /*
144     page.h contains a detailed description of the page format.
145 
146     Normally, keys and data are accessed from offset tables in the
147     top of each page which point to the beginning of the key and
148     data.  There are four flag values which may be stored in these
149     offset tables which indicate the following:
150 
151 	OVFLPAGE	Rather than a key data pair, this pair contains
152 			the address of an overflow page.  The format of
153 			the pair is:
154 			    OVERFLOW_PAGE_NUMBER OVFLPAGE
155 
156 	PARTIAL_KEY	This must be the first key/data pair on a page
157 			and implies that page contains only a partial key.
158 			That is, the key is too big to fit on a single page
159 			so it starts on this page and continues on the next.
160 			The format of the page is:
161 			    KEY_OFF PARTIAL_KEY OVFL_PAGENO OVFLPAGE
162 
163 			    KEY_OFF -- offset of the beginning of the key
164 			    PARTIAL_KEY -- 1
165 			    OVFL_PAGENO - page number of the next overflow page
166 			    OVFLPAGE -- 0
167 	FULL_KEY	This must be the first key/data pair on the page.  It
168 			is used in two cases.
169 
170 			Case 1:
171 			    There is a complete key on the page but no data
172 			    (because it wouldn't fit).  The next page contains
173 			    the data.
174 
175 			    Page format it:
176 			    KEY_OFF FULL_KEY OVFL_PAGENO OVFL_PAGE
177 
178 			    KEY_OFF -- offset of the beginning of the key
179 			    FULL_KEY -- 2
180 			    OVFL_PAGENO - page number of the next overflow page
181 			    OVFLPAGE -- 0
182 
183 			Case 2:
184 			    This page contains no key, but part of a large
185 			    data field, which is continued on the next page.
186 
187 			    Page format it:
188 			    DATA_OFF FULL_KEY OVFL_PAGENO OVFL_PAGE
189 
190 			    KEY_OFF -- offset of the beginning of the data on
191 					this page
192 			    FULL_KEY -- 2
193 			    OVFL_PAGENO - page number of the next overflow page
194 			    OVFLPAGE -- 0
195 
196 	FULL_KEY_DATA	This must be the first key/data pair on the page.
197 			There are two cases:
198 
199 			Case 1:
200 			    This page contains a key and the beginning of the
201 			    data field, but the data field is continued on the
202 			    next page.
203 
204 			    Page format is:
205 			    KEY_OFF FULL_KEY_DATA OVFL_PAGENO DATA_OFF
206 
207 			    KEY_OFF -- offset of the beginning of the key
208 			    FULL_KEY_DATA -- 3
209 			    OVFL_PAGENO - page number of the next overflow page
210 			    DATA_OFF -- offset of the beginning of the data
211 
212 			Case 2:
213 			    This page contains the last page of a big data pair.
214 			    There is no key, only the  tail end of the data
215 			    on this page.
216 
217 			    Page format is:
218 			    DATA_OFF FULL_KEY_DATA <OVFL_PAGENO> <OVFLPAGE>
219 
220 			    DATA_OFF -- offset of the beginning of the data on
221 					this page
222 			    FULL_KEY_DATA -- 3
223 			    OVFL_PAGENO - page number of the next overflow page
224 			    OVFLPAGE -- 0
225 
226 			    OVFL_PAGENO and OVFLPAGE are optional (they are
227 			    not present if there is no next page).
228 */
229 #define OVFLPAGE	0
230 #define PARTIAL_KEY	1
231 #define FULL_KEY	2
232 #define FULL_KEY_DATA	3
233 #define	REAL_KEY	4
234 
235 
236 /* Short hands for accessing structure */
237 #define BSIZE	hdr.bsize
238 #define BSHIFT	hdr.bshift
239 #define DSIZE	hdr.dsize
240 #define SGSIZE	hdr.ssize
241 #define SSHIFT	hdr.sshift
242 #define LORDER	hdr.lorder
243 #define MAX_BUCKET	hdr.max_bucket
244 #define FFACTOR		hdr.ffactor
245 #define HIGH_MASK	hdr.high_mask
246 #define LOW_MASK	hdr.low_mask
247 #define NKEYS		hdr.nkeys
248 #define HDRPAGES	hdr.hdrpages
249 #define SPARES		hdr.spares
250 #define BITMAPS		hdr.bitmaps
251 #define VERSION		hdr.version
252 #define MAGIC		hdr.magic
253 #define NEXT_FREE	hdr.next_free
254 #define H_CHARKEY	hdr.h_charkey
255