bt_put.c revision 1.7 1 /* $NetBSD: bt_put.c,v 1.7 1995/02/27 13:20:40 cgd Exp $ */
2
3 /*-
4 * Copyright (c) 1990, 1993, 1994
5 * The Regents of the University of California. All rights reserved.
6 *
7 * This code is derived from software contributed to Berkeley by
8 * Mike Olson.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the University of
21 * California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * SUCH DAMAGE.
37 */
38
39 #if defined(LIBC_SCCS) && !defined(lint)
40 #if 0
41 static char sccsid[] = "@(#)bt_put.c 8.4 (Berkeley) 5/31/94";
42 #else
43 static char rcsid[] = "$NetBSD: bt_put.c,v 1.7 1995/02/27 13:20:40 cgd Exp $";
44 #endif
45 #endif /* LIBC_SCCS and not lint */
46
47 #include <sys/types.h>
48
49 #include <errno.h>
50 #include <stdio.h>
51 #include <stdlib.h>
52 #include <string.h>
53
54 #include <db.h>
55 #include "btree.h"
56
57 static EPG *bt_fast __P((BTREE *, const DBT *, const DBT *, int *));
58
59 /*
60 * __BT_PUT -- Add a btree item to the tree.
61 *
62 * Parameters:
63 * dbp: pointer to access method
64 * key: key
65 * data: data
66 * flag: R_NOOVERWRITE
67 *
68 * Returns:
69 * RET_ERROR, RET_SUCCESS and RET_SPECIAL if the key is already in the
70 * tree and R_NOOVERWRITE specified.
71 */
72 int
73 __bt_put(dbp, key, data, flags)
74 const DB *dbp;
75 DBT *key;
76 const DBT *data;
77 u_int flags;
78 {
79 BTREE *t;
80 DBT tkey, tdata;
81 EPG *e;
82 PAGE *h;
83 indx_t index, nxtindex;
84 pgno_t pg;
85 u_int32_t nbytes;
86 int dflags, exact, status;
87 char *dest, db[NOVFLSIZE], kb[NOVFLSIZE];
88
89 t = dbp->internal;
90
91 /* Toss any page pinned across calls. */
92 if (t->bt_pinned != NULL) {
93 mpool_put(t->bt_mp, t->bt_pinned, 0);
94 t->bt_pinned = NULL;
95 }
96
97 switch (flags) {
98 case R_CURSOR:
99 if (!ISSET(t, B_SEQINIT))
100 goto einval;
101 if (ISSET(t, B_DELCRSR))
102 goto einval;
103 break;
104 case 0:
105 case R_NOOVERWRITE:
106 break;
107 default:
108 einval: errno = EINVAL;
109 return (RET_ERROR);
110 }
111
112 if (ISSET(t, B_RDONLY)) {
113 errno = EPERM;
114 return (RET_ERROR);
115 }
116
117 /*
118 * If the key/data won't fit on a page, store it on indirect pages.
119 * Only store the key on the overflow page if it's too big after the
120 * data is on an overflow page.
121 *
122 * XXX
123 * If the insert fails later on, these pages aren't recovered.
124 */
125 dflags = 0;
126 if (key->size + data->size > t->bt_ovflsize) {
127 if (key->size > t->bt_ovflsize) {
128 storekey: if (__ovfl_put(t, key, &pg) == RET_ERROR)
129 return (RET_ERROR);
130 tkey.data = kb;
131 tkey.size = NOVFLSIZE;
132 memmove(kb, &pg, sizeof(pgno_t));
133 memmove(kb + sizeof(pgno_t),
134 &key->size, sizeof(u_int32_t));
135 dflags |= P_BIGKEY;
136 key = &tkey;
137 }
138 if (key->size + data->size > t->bt_ovflsize) {
139 if (__ovfl_put(t, data, &pg) == RET_ERROR)
140 return (RET_ERROR);
141 tdata.data = db;
142 tdata.size = NOVFLSIZE;
143 memmove(db, &pg, sizeof(pgno_t));
144 memmove(db + sizeof(pgno_t),
145 &data->size, sizeof(u_int32_t));
146 dflags |= P_BIGDATA;
147 data = &tdata;
148 }
149 if (key->size + data->size > t->bt_ovflsize)
150 goto storekey;
151 }
152
153 /* Replace the cursor. */
154 if (flags == R_CURSOR) {
155 if ((h = mpool_get(t->bt_mp, t->bt_bcursor.pgno, 0)) == NULL)
156 return (RET_ERROR);
157 index = t->bt_bcursor.index;
158 goto delete;
159 }
160
161 /*
162 * Find the key to delete, or, the location at which to insert. Bt_fast
163 * and __bt_search pin the returned page.
164 */
165 if (t->bt_order == NOT || (e = bt_fast(t, key, data, &exact)) == NULL)
166 if ((e = __bt_search(t, key, &exact)) == NULL)
167 return (RET_ERROR);
168 h = e->page;
169 index = e->index;
170
171 /*
172 * Add the specified key/data pair to the tree. If an identical key
173 * is already in the tree, and R_NOOVERWRITE is set, an error is
174 * returned. If R_NOOVERWRITE is not set, the key is either added (if
175 * duplicates are permitted) or an error is returned.
176 *
177 * Pages are split as required.
178 */
179 switch (flags) {
180 case R_NOOVERWRITE:
181 if (!exact)
182 break;
183 /*
184 * One special case is if the cursor references the record and
185 * it's been flagged for deletion. Then, we delete the record,
186 * leaving the cursor there -- this means that the inserted
187 * record will not be seen in a cursor scan.
188 */
189 if (ISSET(t, B_DELCRSR) && t->bt_bcursor.pgno == h->pgno &&
190 t->bt_bcursor.index == index) {
191 CLR(t, B_DELCRSR);
192 goto delete;
193 }
194 mpool_put(t->bt_mp, h, 0);
195 return (RET_SPECIAL);
196 default:
197 if (!exact || !ISSET(t, B_NODUPS))
198 break;
199 delete: if (__bt_dleaf(t, h, index) == RET_ERROR) {
200 mpool_put(t->bt_mp, h, 0);
201 return (RET_ERROR);
202 }
203 break;
204 }
205
206 /*
207 * If not enough room, or the user has put a ceiling on the number of
208 * keys permitted in the page, split the page. The split code will
209 * insert the key and data and unpin the current page. If inserting
210 * into the offset array, shift the pointers up.
211 */
212 nbytes = NBLEAFDBT(key->size, data->size);
213 if (h->upper - h->lower < nbytes + sizeof(indx_t)) {
214 if ((status = __bt_split(t, h, key,
215 data, dflags, nbytes, index)) != RET_SUCCESS)
216 return (status);
217 goto success;
218 }
219
220 if (index < (nxtindex = NEXTINDEX(h)))
221 memmove(h->linp + index + 1, h->linp + index,
222 (nxtindex - index) * sizeof(indx_t));
223 h->lower += sizeof(indx_t);
224
225 h->linp[index] = h->upper -= nbytes;
226 dest = (char *)h + h->upper;
227 WR_BLEAF(dest, key, data, dflags);
228
229 if (t->bt_order == NOT)
230 if (h->nextpg == P_INVALID) {
231 if (index == NEXTINDEX(h) - 1) {
232 t->bt_order = FORWARD;
233 t->bt_last.index = index;
234 t->bt_last.pgno = h->pgno;
235 }
236 } else if (h->prevpg == P_INVALID) {
237 if (index == 0) {
238 t->bt_order = BACK;
239 t->bt_last.index = 0;
240 t->bt_last.pgno = h->pgno;
241 }
242 }
243
244 mpool_put(t->bt_mp, h, MPOOL_DIRTY);
245
246 success:
247 if (flags == R_SETCURSOR) {
248 t->bt_bcursor.pgno = e->page->pgno;
249 t->bt_bcursor.index = e->index;
250 }
251 SET(t, B_MODIFIED);
252 return (RET_SUCCESS);
253 }
254
255 #ifdef STATISTICS
256 u_long bt_cache_hit, bt_cache_miss;
257 #endif
258
259 /*
260 * BT_FAST -- Do a quick check for sorted data.
261 *
262 * Parameters:
263 * t: tree
264 * key: key to insert
265 *
266 * Returns:
267 * EPG for new record or NULL if not found.
268 */
269 static EPG *
270 bt_fast(t, key, data, exactp)
271 BTREE *t;
272 const DBT *key, *data;
273 int *exactp;
274 {
275 PAGE *h;
276 u_int32_t nbytes;
277 int cmp;
278
279 if ((h = mpool_get(t->bt_mp, t->bt_last.pgno, 0)) == NULL) {
280 t->bt_order = NOT;
281 return (NULL);
282 }
283 t->bt_cur.page = h;
284 t->bt_cur.index = t->bt_last.index;
285
286 /*
287 * If won't fit in this page or have too many keys in this page, have
288 * to search to get split stack.
289 */
290 nbytes = NBLEAFDBT(key->size, data->size);
291 if (h->upper - h->lower < nbytes + sizeof(indx_t))
292 goto miss;
293
294 if (t->bt_order == FORWARD) {
295 if (t->bt_cur.page->nextpg != P_INVALID)
296 goto miss;
297 if (t->bt_cur.index != NEXTINDEX(h) - 1)
298 goto miss;
299 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) < 0)
300 goto miss;
301 t->bt_last.index = cmp ? ++t->bt_cur.index : t->bt_cur.index;
302 } else {
303 if (t->bt_cur.page->prevpg != P_INVALID)
304 goto miss;
305 if (t->bt_cur.index != 0)
306 goto miss;
307 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) > 0)
308 goto miss;
309 t->bt_last.index = 0;
310 }
311 *exactp = cmp == 0;
312 #ifdef STATISTICS
313 ++bt_cache_hit;
314 #endif
315 return (&t->bt_cur);
316
317 miss:
318 #ifdef STATISTICS
319 ++bt_cache_miss;
320 #endif
321 t->bt_order = NOT;
322 mpool_put(t->bt_mp, h, 0);
323 return (NULL);
324 }
325