bt_put.c revision 1.2 1 /*-
2 * Copyright (c) 1990, 1993
3 * The Regents of the University of California. 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 #if defined(LIBC_SCCS) && !defined(lint)
38 /*static char sccsid[] = "from: @(#)bt_put.c 8.1 (Berkeley) 6/4/93";*/
39 static char rcsid[] = "$Id: bt_put.c,v 1.2 1993/08/01 18:43:52 mycroft Exp $";
40 #endif /* LIBC_SCCS and not lint */
41
42 #include <sys/types.h>
43
44 #include <errno.h>
45 #include <stdio.h>
46 #include <stdlib.h>
47 #include <string.h>
48
49 #include <db.h>
50 #include "btree.h"
51
52 static EPG *bt_fast __P((BTREE *, const DBT *, const DBT *, int *));
53
54 /*
55 * __BT_PUT -- Add a btree item to the tree.
56 *
57 * Parameters:
58 * dbp: pointer to access method
59 * key: key
60 * data: data
61 * flag: R_NOOVERWRITE
62 *
63 * Returns:
64 * RET_ERROR, RET_SUCCESS and RET_SPECIAL if the key is already in the
65 * tree and R_NOOVERWRITE specified.
66 */
67 int
68 __bt_put(dbp, key, data, flags)
69 const DB *dbp;
70 DBT *key;
71 const DBT *data;
72 u_int flags;
73 {
74 BTREE *t;
75 DBT tkey, tdata;
76 EPG *e;
77 PAGE *h;
78 indx_t index, nxtindex;
79 pgno_t pg;
80 size_t nbytes;
81 int dflags, exact, status;
82 char *dest, db[NOVFLSIZE], kb[NOVFLSIZE];
83
84 t = dbp->internal;
85
86 switch (flags) {
87 case R_CURSOR:
88 if (!ISSET(t, B_SEQINIT))
89 goto einval;
90 if (ISSET(t, B_DELCRSR))
91 goto einval;
92 break;
93 case 0:
94 case R_NOOVERWRITE:
95 break;
96 default:
97 einval: errno = EINVAL;
98 return (RET_ERROR);
99 }
100
101 if (ISSET(t, B_RDONLY)) {
102 errno = EPERM;
103 return (RET_ERROR);
104 }
105
106 /*
107 * If the key/data won't fit on a page, store it on indirect pages.
108 * Only store the key on the overflow page if it's too big after the
109 * data is on an overflow page.
110 *
111 * XXX
112 * If the insert fails later on, these pages aren't recovered.
113 */
114 dflags = 0;
115 if (key->size + data->size > t->bt_ovflsize) {
116 if (key->size > t->bt_ovflsize) {
117 storekey: if (__ovfl_put(t, key, &pg) == RET_ERROR)
118 return (RET_ERROR);
119 tkey.data = kb;
120 tkey.size = NOVFLSIZE;
121 memmove(kb, &pg, sizeof(pgno_t));
122 memmove(kb + sizeof(pgno_t),
123 &key->size, sizeof(size_t));
124 dflags |= P_BIGKEY;
125 key = &tkey;
126 }
127 if (key->size + data->size > t->bt_ovflsize) {
128 if (__ovfl_put(t, data, &pg) == RET_ERROR)
129 return (RET_ERROR);
130 tdata.data = db;
131 tdata.size = NOVFLSIZE;
132 memmove(db, &pg, sizeof(pgno_t));
133 memmove(db + sizeof(pgno_t),
134 &data->size, sizeof(size_t));
135 dflags |= P_BIGDATA;
136 data = &tdata;
137 }
138 if (key->size + data->size > t->bt_ovflsize)
139 goto storekey;
140 }
141
142 /* Replace the cursor. */
143 if (flags == R_CURSOR) {
144 if ((h = mpool_get(t->bt_mp, t->bt_bcursor.pgno, 0)) == NULL)
145 return (RET_ERROR);
146 index = t->bt_bcursor.index;
147 goto delete;
148 }
149
150 /*
151 * Find the key to delete, or, the location at which to insert. Bt_fast
152 * and __bt_search pin the returned page.
153 */
154 if (t->bt_order == NOT || (e = bt_fast(t, key, data, &exact)) == NULL)
155 if ((e = __bt_search(t, key, &exact)) == NULL)
156 return (RET_ERROR);
157 h = e->page;
158 index = e->index;
159
160 /*
161 * Add the specified key/data pair to the tree. If an identical key
162 * is already in the tree, and R_NOOVERWRITE is set, an error is
163 * returned. If R_NOOVERWRITE is not set, the key is either added (if
164 * duplicates are permitted) or an error is returned.
165 *
166 * Pages are split as required.
167 */
168 switch (flags) {
169 case R_NOOVERWRITE:
170 if (!exact)
171 break;
172 /*
173 * One special case is if the cursor references the record and
174 * it's been flagged for deletion. Then, we delete the record,
175 * leaving the cursor there -- this means that the inserted
176 * record will not be seen in a cursor scan.
177 */
178 if (ISSET(t, B_DELCRSR) && t->bt_bcursor.pgno == h->pgno &&
179 t->bt_bcursor.index == index) {
180 CLR(t, B_DELCRSR);
181 goto delete;
182 }
183 mpool_put(t->bt_mp, h, 0);
184 return (RET_SPECIAL);
185 default:
186 if (!exact || !ISSET(t, B_NODUPS))
187 break;
188 delete: if (__bt_dleaf(t, h, index) == RET_ERROR) {
189 mpool_put(t->bt_mp, h, 0);
190 return (RET_ERROR);
191 }
192 break;
193 }
194
195 /*
196 * If not enough room, or the user has put a ceiling on the number of
197 * keys permitted in the page, split the page. The split code will
198 * insert the key and data and unpin the current page. If inserting
199 * into the offset array, shift the pointers up.
200 */
201 nbytes = NBLEAFDBT(key->size, data->size);
202 if (h->upper - h->lower < nbytes + sizeof(indx_t)) {
203 if ((status = __bt_split(t, h, key,
204 data, dflags, nbytes, index)) != RET_SUCCESS)
205 return (status);
206 goto success;
207 }
208
209 if (index < (nxtindex = NEXTINDEX(h)))
210 memmove(h->linp + index + 1, h->linp + index,
211 (nxtindex - index) * sizeof(indx_t));
212 h->lower += sizeof(indx_t);
213
214 h->linp[index] = h->upper -= nbytes;
215 dest = (char *)h + h->upper;
216 WR_BLEAF(dest, key, data, dflags);
217
218 if (t->bt_order == NOT)
219 if (h->nextpg == P_INVALID) {
220 if (index == NEXTINDEX(h) - 1) {
221 t->bt_order = FORWARD;
222 t->bt_last.index = index;
223 t->bt_last.pgno = h->pgno;
224 }
225 } else if (h->prevpg == P_INVALID) {
226 if (index == 0) {
227 t->bt_order = BACK;
228 t->bt_last.index = 0;
229 t->bt_last.pgno = h->pgno;
230 }
231 }
232
233 mpool_put(t->bt_mp, h, MPOOL_DIRTY);
234
235 success:
236 if (flags == R_SETCURSOR) {
237 t->bt_bcursor.pgno = e->page->pgno;
238 t->bt_bcursor.index = e->index;
239 }
240 SET(t, B_MODIFIED);
241 return (RET_SUCCESS);
242 }
243
244 #ifdef STATISTICS
245 u_long bt_cache_hit, bt_cache_miss;
246 #endif
247
248 /*
249 * BT_FAST -- Do a quick check for sorted data.
250 *
251 * Parameters:
252 * t: tree
253 * key: key to insert
254 *
255 * Returns:
256 * EPG for new record or NULL if not found.
257 */
258 static EPG *
259 bt_fast(t, key, data, exactp)
260 BTREE *t;
261 const DBT *key, *data;
262 int *exactp;
263 {
264 EPG e;
265 PAGE *h;
266 size_t nbytes;
267 int cmp;
268
269 if ((h = mpool_get(t->bt_mp, t->bt_last.pgno, 0)) == NULL) {
270 t->bt_order = NOT;
271 return (NULL);
272 }
273 e.page = h;
274 e.index = t->bt_last.index;
275
276 /*
277 * If won't fit in this page or have too many keys in this page, have
278 * to search to get split stack.
279 */
280 nbytes = NBLEAFDBT(key->size, data->size);
281 if (h->upper - h->lower < nbytes + sizeof(indx_t))
282 goto miss;
283
284 if (t->bt_order == FORWARD) {
285 if (e.page->nextpg != P_INVALID)
286 goto miss;
287 if (e.index != NEXTINDEX(h) - 1)
288 goto miss;
289 if ((cmp = __bt_cmp(t, key, &e)) < 0)
290 goto miss;
291 t->bt_last.index = cmp ? ++e.index : e.index;
292 } else {
293 if (e.page->prevpg != P_INVALID)
294 goto miss;
295 if (e.index != 0)
296 goto miss;
297 if ((cmp = __bt_cmp(t, key, &e)) > 0)
298 goto miss;
299 t->bt_last.index = 0;
300 }
301 *exactp = cmp == 0;
302 #ifdef STATISTICS
303 ++bt_cache_hit;
304 #endif
305 return (&e);
306
307 miss:
308 #ifdef STATISTICS
309 ++bt_cache_miss;
310 #endif
311 t->bt_order = NOT;
312 mpool_put(t->bt_mp, h, 0);
313 return (NULL);
314 }
315