xref: /csrg-svn/lib/libc/db/btree/bt_put.c (revision 50989)
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 #if defined(LIBC_SCCS) && !defined(lint)
12 static char sccsid[] = "@(#)bt_put.c	5.4 (Berkeley) 09/04/91";
13 #endif /* LIBC_SCCS and not lint */
14 
15 #include <sys/types.h>
16 #include <errno.h>
17 #include <db.h>
18 #include <stdio.h>
19 #include <stdlib.h>
20 #include <string.h>
21 #include "btree.h"
22 
23 static EPG *bt_fast __P((BTREE *, const DBT *, const DBT *, int *));
24 
25 /*
26  * __BT_PUT -- Add a btree item to the tree.
27  *
28  * Parameters:
29  *	dbp:	pointer to access method
30  *	key:	key
31  *	data:	data
32  *	flag:	R_NOOVERWRITE
33  *
34  * Returns:
35  *	RET_ERROR, RET_SUCCESS and RET_SPECIAL if the key is already in the
36  *	tree and R_NOOVERWRITE specified.
37  */
38 int
39 __bt_put(dbp, key, data, flags)
40 	const DB *dbp;
41 	const DBT *key, *data;
42 	u_int flags;
43 {
44 	BTREE *t;
45 	DBT tkey, tdata;
46 	EPG *e;
47 	PAGE *h;
48 	index_t index, nxtindex;
49 	pgno_t pg;
50 	size_t nbytes;
51 	int dflags, exact;
52 	char *dest, db[NOVFLSIZE], kb[NOVFLSIZE];
53 
54 	if (flags && flags != R_NOOVERWRITE) {
55 		errno = EINVAL;
56 		return (RET_ERROR);
57 	}
58 	t = dbp->internal;
59 	if (ISSET(t, BTF_RDONLY)) {
60 		errno = EPERM;
61 		return (RET_ERROR);
62 	}
63 
64 	/*
65 	 * If the key/data won't fit on a page, store it on indirect pages.
66 	 *
67 	 * XXX
68 	 * If the insert fails later on, these pages aren't recovered.
69 	 */
70 	dflags = 0;
71 	if (key->size >= t->bt_minkeypage) {
72 		if (__ovfl_put(t, key, &pg) == RET_ERROR)
73 			return (RET_ERROR);
74 		tkey.data = kb;
75 		tkey.size = NOVFLSIZE;
76 		*(pgno_t *)kb = pg;
77 		*(size_t *)(kb + sizeof(pgno_t)) = key->size;
78 		dflags |= P_BIGKEY;
79 		key = &tkey;
80 	}
81 	if (data->size >= t->bt_minkeypage) {
82 		if (__ovfl_put(t, data, &pg) == RET_ERROR)
83 			return (RET_ERROR);
84 		tdata.data = db;
85 		tdata.size = NOVFLSIZE;
86 		*(pgno_t *)db = pg;
87 		*(size_t *)(db + sizeof(pgno_t)) = data->size;
88 		dflags |= P_BIGDATA;
89 		data = &tdata;
90 	}
91 
92 	/* bt_fast and __bt_search pin the returned page. */
93 	if (t->bt_order == NOT || (e = bt_fast(t, key, data, &exact)) == NULL)
94 		if ((e = __bt_search(t, key, &exact)) == NULL)
95 			return (RET_ERROR);
96 
97 	h = e->page;
98 	index = e->index;
99 
100 	/*
101 	 * Add the specified key/data pair to the tree.  If an identical key
102 	 * is already in the tree, and R_NOOVERWRITE is set, an error is
103 	 * returned.  If R_NOOVERWRITE is not set, the key is either added (if
104 	 * duplicates are permitted) or an error is returned.
105 	 *
106 	 * Pages are split as required.
107 	 */
108 	switch (flags) {
109 	case R_NOOVERWRITE:
110 		if (!exact)
111 			break;
112 		/*
113 		 * One special case is if the cursor references the record and
114 		 * it's been flagged for deletion.  If so, we delete it and
115 		 * pretend it was never there.  Since the cursor will move to
116 		 * the next record the inserted record won't be seen.
117 		 */
118 		if (ISSET(t, BTF_DELCRSR) && t->bt_bcursor.pgno == h->pgno &&
119 		    t->bt_bcursor.index == index) {
120 			UNSET(t, BTF_DELCRSR);
121 			goto delete;
122 		}
123 		BT_CLR(t);
124 		mpool_put(t->bt_mp, h, 0);
125 		return (RET_SPECIAL);
126 	default:
127 		if (!exact || NOTSET(t, BTF_NODUPS))
128 			break;
129 delete:		if (__bt_dleaf(t, h, index) == RET_ERROR) {
130 			BT_CLR(t);
131 			mpool_put(t->bt_mp, h, 0);
132 			return (RET_ERROR);
133 		}
134 		break;
135 	}
136 
137 	/*
138 	 * If not enough room, or the user has put a ceiling on the number of
139 	 * keys permitted in the page, split the page.  The split code will
140 	 * insert the key and data and unpin the current page.  If inserting
141 	 * into the offset array, shift the pointers up.
142 	 */
143 	nbytes = NBLEAFDBT(key->size, data->size);
144 	if (h->upper - h->lower < nbytes + sizeof(index_t) ||
145 	    t->bt_maxkeypage && t->bt_maxkeypage < NEXTINDEX(h))
146 		return (__bt_split(t, h, key, data, dflags, nbytes, index));
147 
148 	if (index < (nxtindex = NEXTINDEX(h)))
149 		bcopy(h->linp + index, h->linp + index + 1,
150 		    (nxtindex - index) * sizeof(index_t));
151 	h->lower += sizeof(index_t);
152 
153 	h->linp[index] = h->upper -= nbytes;
154 	dest = (char *)h + h->upper;
155 	WR_BLEAF(dest, key, data, dflags);
156 
157 	if (t->bt_order == NOT)
158 		if (h->nextpg == P_INVALID) {
159 			if (index == NEXTINDEX(h) - 1) {
160 				t->bt_order = FORWARD;
161 				t->bt_last.index = index;
162 				t->bt_last.pgno = h->pgno;
163 			}
164 		} else if (h->prevpg == P_INVALID) {
165 			if (index == 0) {
166 				t->bt_order = BACK;
167 				t->bt_last.index = 0;
168 				t->bt_last.pgno = h->pgno;
169 			}
170 		}
171 
172 	BT_CLR(t);
173 	mpool_put(t->bt_mp, h, MPOOL_DIRTY);
174 	SET(t, BTF_MODIFIED);
175 	return (RET_SUCCESS);
176 }
177 
178 #ifdef STATISTICS
179 u_long bt_cache_hit, bt_cache_miss;
180 #endif
181 
182 /*
183  * BT_FAST -- Do a quick check for sorted data.
184  *
185  * Parameters:
186  *	t:	tree
187  *	key:	key to insert
188  *
189  * Returns:
190  * 	EPG for new record or NULL if not found.
191  */
192 static EPG *
193 bt_fast(t, key, data, exactp)
194 	BTREE *t;
195 	const DBT *key, *data;
196 	int *exactp;
197 {
198 	EPG e;
199 	PAGE *h;
200 	size_t nbytes;
201 	int cmp;
202 
203 	if ((h = mpool_get(t->bt_mp, t->bt_last.pgno, 0)) == NULL) {
204 		t->bt_order = NOT;
205 		return (NULL);
206 	}
207 	e.page = h;
208 	e.index = t->bt_last.index;
209 
210 	/*
211 	 * If won't fit in this page or have too many keys in this page, have
212 	 * to search to get split stack.
213 	 */
214 	nbytes =
215 	    NBLEAFDBT(key->size >= t->bt_minkeypage ? NOVFLSIZE : key->size,
216 	    data->size >= t->bt_minkeypage ? NOVFLSIZE : data->size);
217 	if (h->upper - h->lower < nbytes + sizeof(index_t) ||
218 	    t->bt_maxkeypage && t->bt_maxkeypage < NEXTINDEX(h))
219 		goto miss;
220 
221 	if (t->bt_order == FORWARD) {
222 		if (e.page->nextpg != P_INVALID)
223 			goto miss;
224 		if (e.index != NEXTINDEX(h) - 1)
225 			goto miss;
226 		if ((cmp = __bt_cmp(t, key, &e)) < 0)
227 			goto miss;
228 		t->bt_last.index = ++e.index;
229 	} else {
230 		if (e.page->prevpg != P_INVALID)
231 			goto miss;
232 		if (e.index != 0)
233 			goto miss;
234 		if ((cmp = __bt_cmp(t, key, &e)) > 0)
235 			goto miss;
236 		t->bt_last.index = 0;
237 	}
238 	*exactp = cmp == 0;
239 #ifdef STATISTICS
240 	++bt_cache_hit;
241 #endif
242 	return (&e);
243 
244 miss:
245 #ifdef STATISTICS
246 	++bt_cache_miss;
247 #endif
248 	t->bt_order = NOT;
249 	mpool_put(t->bt_mp, h, 0);
250 	return (NULL);
251 }
252