bt_put.c revision 1573
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)
38static char sccsid[] = "@(#)bt_put.c	8.3 (Berkeley) 9/16/93";
39#endif /* LIBC_SCCS and not lint */
40
41#include <sys/types.h>
42
43#include <errno.h>
44#include <stdio.h>
45#include <stdlib.h>
46#include <string.h>
47
48#include <db.h>
49#include "btree.h"
50
51static EPG *bt_fast __P((BTREE *, const DBT *, const DBT *, int *));
52
53/*
54 * __BT_PUT -- Add a btree item to the tree.
55 *
56 * Parameters:
57 *	dbp:	pointer to access method
58 *	key:	key
59 *	data:	data
60 *	flag:	R_NOOVERWRITE
61 *
62 * Returns:
63 *	RET_ERROR, RET_SUCCESS and RET_SPECIAL if the key is already in the
64 *	tree and R_NOOVERWRITE specified.
65 */
66int
67__bt_put(dbp, key, data, flags)
68	const DB *dbp;
69	DBT *key;
70	const DBT *data;
71	u_int flags;
72{
73	BTREE *t;
74	DBT tkey, tdata;
75	EPG *e;
76	PAGE *h;
77	indx_t index, nxtindex;
78	pgno_t pg;
79	size_t nbytes;
80	int dflags, exact, status;
81	char *dest, db[NOVFLSIZE], kb[NOVFLSIZE];
82
83	t = dbp->internal;
84
85	/* Toss any page pinned across calls. */
86	if (t->bt_pinned != NULL) {
87		mpool_put(t->bt_mp, t->bt_pinned, 0);
88		t->bt_pinned = NULL;
89	}
90
91	switch (flags) {
92	case R_CURSOR:
93		if (!ISSET(t, B_SEQINIT))
94			goto einval;
95		if (ISSET(t, B_DELCRSR))
96			goto einval;
97		break;
98	case 0:
99	case R_NOOVERWRITE:
100		break;
101	default:
102einval:		errno = EINVAL;
103		return (RET_ERROR);
104	}
105
106	if (ISSET(t, B_RDONLY)) {
107		errno = EPERM;
108		return (RET_ERROR);
109	}
110
111	/*
112	 * If the key/data won't fit on a page, store it on indirect pages.
113	 * Only store the key on the overflow page if it's too big after the
114	 * data is on an overflow page.
115	 *
116	 * XXX
117	 * If the insert fails later on, these pages aren't recovered.
118	 */
119	dflags = 0;
120	if (key->size + data->size > t->bt_ovflsize) {
121		if (key->size > t->bt_ovflsize) {
122storekey:		if (__ovfl_put(t, key, &pg) == RET_ERROR)
123				return (RET_ERROR);
124			tkey.data = kb;
125			tkey.size = NOVFLSIZE;
126			memmove(kb, &pg, sizeof(pgno_t));
127			memmove(kb + sizeof(pgno_t),
128			    &key->size, sizeof(size_t));
129			dflags |= P_BIGKEY;
130			key = &tkey;
131		}
132		if (key->size + data->size > t->bt_ovflsize) {
133			if (__ovfl_put(t, data, &pg) == RET_ERROR)
134				return (RET_ERROR);
135			tdata.data = db;
136			tdata.size = NOVFLSIZE;
137			memmove(db, &pg, sizeof(pgno_t));
138			memmove(db + sizeof(pgno_t),
139			    &data->size, sizeof(size_t));
140			dflags |= P_BIGDATA;
141			data = &tdata;
142		}
143		if (key->size + data->size > t->bt_ovflsize)
144			goto storekey;
145	}
146
147	/* Replace the cursor. */
148	if (flags == R_CURSOR) {
149		if ((h = mpool_get(t->bt_mp, t->bt_bcursor.pgno, 0)) == NULL)
150			return (RET_ERROR);
151		index = t->bt_bcursor.index;
152		goto delete;
153	}
154
155	/*
156	 * Find the key to delete, or, the location at which to insert.  Bt_fast
157	 * and __bt_search pin the returned page.
158	 */
159	if (t->bt_order == NOT || (e = bt_fast(t, key, data, &exact)) == NULL)
160		if ((e = __bt_search(t, key, &exact)) == NULL)
161			return (RET_ERROR);
162	h = e->page;
163	index = e->index;
164
165	/*
166	 * Add the specified key/data pair to the tree.  If an identical key
167	 * is already in the tree, and R_NOOVERWRITE is set, an error is
168	 * returned.  If R_NOOVERWRITE is not set, the key is either added (if
169	 * duplicates are permitted) or an error is returned.
170	 *
171	 * Pages are split as required.
172	 */
173	switch (flags) {
174	case R_NOOVERWRITE:
175		if (!exact)
176			break;
177		/*
178		 * One special case is if the cursor references the record and
179		 * it's been flagged for deletion.  Then, we delete the record,
180		 * leaving the cursor there -- this means that the inserted
181		 * record will not be seen in a cursor scan.
182		 */
183		if (ISSET(t, B_DELCRSR) && t->bt_bcursor.pgno == h->pgno &&
184		    t->bt_bcursor.index == index) {
185			CLR(t, B_DELCRSR);
186			goto delete;
187		}
188		mpool_put(t->bt_mp, h, 0);
189		return (RET_SPECIAL);
190	default:
191		if (!exact || !ISSET(t, B_NODUPS))
192			break;
193delete:		if (__bt_dleaf(t, h, index) == RET_ERROR) {
194			mpool_put(t->bt_mp, h, 0);
195			return (RET_ERROR);
196		}
197		break;
198	}
199
200	/*
201	 * If not enough room, or the user has put a ceiling on the number of
202	 * keys permitted in the page, split the page.  The split code will
203	 * insert the key and data and unpin the current page.  If inserting
204	 * into the offset array, shift the pointers up.
205	 */
206	nbytes = NBLEAFDBT(key->size, data->size);
207	if (h->upper - h->lower < nbytes + sizeof(indx_t)) {
208		if ((status = __bt_split(t, h, key,
209		    data, dflags, nbytes, index)) != RET_SUCCESS)
210			return (status);
211		goto success;
212	}
213
214	if (index < (nxtindex = NEXTINDEX(h)))
215		memmove(h->linp + index + 1, h->linp + index,
216		    (nxtindex - index) * sizeof(indx_t));
217	h->lower += sizeof(indx_t);
218
219	h->linp[index] = h->upper -= nbytes;
220	dest = (char *)h + h->upper;
221	WR_BLEAF(dest, key, data, dflags);
222
223	if (t->bt_order == NOT)
224		if (h->nextpg == P_INVALID) {
225			if (index == NEXTINDEX(h) - 1) {
226				t->bt_order = FORWARD;
227				t->bt_last.index = index;
228				t->bt_last.pgno = h->pgno;
229			}
230		} else if (h->prevpg == P_INVALID) {
231			if (index == 0) {
232				t->bt_order = BACK;
233				t->bt_last.index = 0;
234				t->bt_last.pgno = h->pgno;
235			}
236		}
237
238	mpool_put(t->bt_mp, h, MPOOL_DIRTY);
239
240success:
241	if (flags == R_SETCURSOR) {
242		t->bt_bcursor.pgno = e->page->pgno;
243		t->bt_bcursor.index = e->index;
244	}
245	SET(t, B_MODIFIED);
246	return (RET_SUCCESS);
247}
248
249#ifdef STATISTICS
250u_long bt_cache_hit, bt_cache_miss;
251#endif
252
253/*
254 * BT_FAST -- Do a quick check for sorted data.
255 *
256 * Parameters:
257 *	t:	tree
258 *	key:	key to insert
259 *
260 * Returns:
261 * 	EPG for new record or NULL if not found.
262 */
263static EPG *
264bt_fast(t, key, data, exactp)
265	BTREE *t;
266	const DBT *key, *data;
267	int *exactp;
268{
269	PAGE *h;
270	size_t nbytes;
271	int cmp;
272
273	if ((h = mpool_get(t->bt_mp, t->bt_last.pgno, 0)) == NULL) {
274		t->bt_order = NOT;
275		return (NULL);
276	}
277	t->bt_cur.page = h;
278	t->bt_cur.index = t->bt_last.index;
279
280	/*
281	 * If won't fit in this page or have too many keys in this page, have
282	 * to search to get split stack.
283	 */
284	nbytes = NBLEAFDBT(key->size, data->size);
285	if (h->upper - h->lower < nbytes + sizeof(indx_t))
286		goto miss;
287
288	if (t->bt_order == FORWARD) {
289		if (t->bt_cur.page->nextpg != P_INVALID)
290			goto miss;
291		if (t->bt_cur.index != NEXTINDEX(h) - 1)
292			goto miss;
293		if ((cmp = __bt_cmp(t, key, &t->bt_cur)) < 0)
294			goto miss;
295		t->bt_last.index = cmp ? ++t->bt_cur.index : t->bt_cur.index;
296	} else {
297		if (t->bt_cur.page->prevpg != P_INVALID)
298			goto miss;
299		if (t->bt_cur.index != 0)
300			goto miss;
301		if ((cmp = __bt_cmp(t, key, &t->bt_cur)) > 0)
302			goto miss;
303		t->bt_last.index = 0;
304	}
305	*exactp = cmp == 0;
306#ifdef STATISTICS
307	++bt_cache_hit;
308#endif
309	return (&t->bt_cur);
310
311miss:
312#ifdef STATISTICS
313	++bt_cache_miss;
314#endif
315	t->bt_order = NOT;
316	mpool_put(t->bt_mp, h, 0);
317	return (NULL);
318}
319