]>
git.saurik.com Git - apple/libc.git/blob - db.subproj/hash.subproj/hash_bigkey.c
2 * Copyright (c) 1999 Apple Computer, Inc. All rights reserved.
4 * @APPLE_LICENSE_HEADER_START@
6 * The contents of this file constitute Original Code as defined in and
7 * are subject to the Apple Public Source License Version 1.1 (the
8 * "License"). You may not use this file except in compliance with the
9 * License. Please obtain a copy of the License at
10 * http://www.apple.com/publicsource and read it before using this file.
12 * This Original Code and all software distributed under the License are
13 * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
14 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
15 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the
17 * License for the specific language governing rights and limitations
20 * @APPLE_LICENSE_HEADER_END@
23 * Copyright (c) 1990, 1993
24 * The Regents of the University of California. All rights reserved.
26 * This code is derived from software contributed to Berkeley by
29 * Redistribution and use in source and binary forms, with or without
30 * modification, are permitted provided that the following conditions
32 * 1. Redistributions of source code must retain the above copyright
33 * notice, this list of conditions and the following disclaimer.
34 * 2. Redistributions in binary form must reproduce the above copyright
35 * notice, this list of conditions and the following disclaimer in the
36 * documentation and/or other materials provided with the distribution.
37 * 3. All advertising materials mentioning features or use of this software
38 * must display the following acknowledgement:
39 * This product includes software developed by the University of
40 * California, Berkeley and its contributors.
41 * 4. Neither the name of the University nor the names of its contributors
42 * may be used to endorse or promote products derived from this software
43 * without specific prior written permission.
45 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
46 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
47 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
48 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
49 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
50 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
51 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
52 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
53 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
54 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62 * Big key/data handling for the hashing package.
77 #include <sys/param.h>
93 static int collect_key
__P((HTAB
*, BUFHEAD
*, int, DBT
*, int));
94 static int collect_data
__P((HTAB
*, BUFHEAD
*, int, int));
99 * You need to do an insert and the key/data pair is too big
106 __big_insert(hashp
, bufp
, key
, val
)
109 const DBT
*key
, *val
;
112 int key_size
, n
, val_size
;
113 u_short space
, move_bytes
, off
;
114 char *cp
, *key_data
, *val_data
;
116 cp
= bufp
->page
; /* Character pointer of p. */
119 key_data
= (char *)key
->data
;
120 key_size
= key
->size
;
121 val_data
= (char *)val
->data
;
122 val_size
= val
->size
;
124 /* First move the Key */
125 for (space
= FREESPACE(p
) - BIGOVERHEAD
; key_size
;
126 space
= FREESPACE(p
) - BIGOVERHEAD
) {
127 move_bytes
= MIN(space
, key_size
);
128 off
= OFFSET(p
) - move_bytes
;
129 memmove(cp
+ off
, key_data
, move_bytes
);
130 key_size
-= move_bytes
;
131 key_data
+= move_bytes
;
135 FREESPACE(p
) = off
- PAGE_META(n
);
138 bufp
= __add_ovflpage(hashp
, bufp
);
144 move_bytes
= MIN(FREESPACE(p
), val_size
);
145 off
= OFFSET(p
) - move_bytes
;
147 memmove(cp
+ off
, val_data
, move_bytes
);
148 val_data
+= move_bytes
;
149 val_size
-= move_bytes
;
150 p
[n
- 2] = FULL_KEY_DATA
;
151 FREESPACE(p
) = FREESPACE(p
) - move_bytes
;
155 p
= (u_short
*)bufp
->page
;
157 bufp
->flags
|= BUF_MOD
;
160 /* Now move the data */
161 for (space
= FREESPACE(p
) - BIGOVERHEAD
; val_size
;
162 space
= FREESPACE(p
) - BIGOVERHEAD
) {
163 move_bytes
= MIN(space
, val_size
);
165 * Here's the hack to make sure that if the data ends on the
166 * same page as the key ends, FREESPACE is at least one.
168 if (space
== val_size
&& val_size
== val
->size
)
170 off
= OFFSET(p
) - move_bytes
;
171 memmove(cp
+ off
, val_data
, move_bytes
);
172 val_size
-= move_bytes
;
173 val_data
+= move_bytes
;
177 FREESPACE(p
) = off
- PAGE_META(n
);
181 bufp
= __add_ovflpage(hashp
, bufp
);
187 p
[n
] = FULL_KEY_DATA
;
188 bufp
->flags
|= BUF_MOD
;
194 * Called when bufp's page contains a partial key (index should be 1)
196 * All pages in the big key/data pair except bufp are freed. We cannot
197 * free bufp because the page pointing to it is lost and we can't get rid
205 __big_delete(hashp
, bufp
)
209 register BUFHEAD
*last_bfp
, *rbufp
;
215 bp
= (u_short
*)bufp
->page
;
219 while (!key_done
|| (bp
[2] != FULL_KEY_DATA
)) {
220 if (bp
[2] == FULL_KEY
|| bp
[2] == FULL_KEY_DATA
)
224 * If there is freespace left on a FULL_KEY_DATA page, then
225 * the data is short and fits entirely on this page, and this
228 if (bp
[2] == FULL_KEY_DATA
&& FREESPACE(bp
))
230 pageno
= bp
[bp
[0] - 1];
231 rbufp
->flags
|= BUF_MOD
;
232 rbufp
= __get_buf(hashp
, pageno
, rbufp
, 0);
234 __free_ovflpage(hashp
, last_bfp
);
237 return (-1); /* Error. */
238 bp
= (u_short
*)rbufp
->page
;
242 * If we get here then rbufp points to the last page of the big
243 * key/data pair. Bufp points to the first one -- it should now be
244 * empty pointing to the next page after this pair. Can't free it
245 * because we don't have the page pointing to it.
248 /* This is information from the last page of the pair. */
252 /* Now, bp is the first page of the pair. */
253 bp
= (u_short
*)bufp
->page
;
255 /* There is an overflow page. */
258 bufp
->ovfl
= rbufp
->ovfl
;
260 /* This is the last page. */
264 FREESPACE(bp
) = hashp
->BSIZE
- PAGE_META(n
);
265 OFFSET(bp
) = hashp
->BSIZE
- 1;
267 bufp
->flags
|= BUF_MOD
;
269 __free_ovflpage(hashp
, rbufp
);
270 if (last_bfp
!= rbufp
)
271 __free_ovflpage(hashp
, last_bfp
);
279 * -1 = get next overflow page
280 * -2 means key not found and this is big key/data
284 __find_bigpair(hashp
, bufp
, ndx
, key
, size
)
291 register u_short
*bp
;
297 bp
= (u_short
*)bufp
->page
;
302 for (bytes
= hashp
->BSIZE
- bp
[ndx
];
303 bytes
<= size
&& bp
[ndx
+ 1] == PARTIAL_KEY
;
304 bytes
= hashp
->BSIZE
- bp
[ndx
]) {
305 if (memcmp(p
+ bp
[ndx
], kkey
, bytes
))
309 bufp
= __get_buf(hashp
, bp
[ndx
+ 2], bufp
, 0);
317 if (bytes
!= ksize
|| memcmp(p
+ bp
[ndx
], kkey
, bytes
)) {
318 #ifdef HASH_STATISTICS
327 * Given the buffer pointer of the first overflow page of a big pair,
328 * find the end of the big pair
330 * This will set bpp to the buffer header of the last page of the big pair.
331 * It will return the pageno of the overflow page following the last page
332 * of the pair; 0 if there isn't any (i.e. big pair is the last key in the
336 __find_last_page(hashp
, bpp
)
345 bp
= (u_short
*)bufp
->page
;
350 * This is the last page if: the tag is FULL_KEY_DATA and
351 * either only 2 entries OVFLPAGE marker is explicit there
352 * is freespace on the page.
354 if (bp
[2] == FULL_KEY_DATA
&&
355 ((n
== 2) || (bp
[n
] == OVFLPAGE
) || (FREESPACE(bp
))))
359 bufp
= __get_buf(hashp
, pageno
, bufp
, 0);
361 return (0); /* Need to indicate an error! */
362 bp
= (u_short
*)bufp
->page
;
373 * Return the data for the key/data pair that begins on this page at this
374 * index (index should always be 1).
377 __big_return(hashp
, bufp
, ndx
, val
, set_current
)
385 u_short
*bp
, len
, off
, save_addr
;
388 bp
= (u_short
*)bufp
->page
;
389 while (bp
[ndx
+ 1] == PARTIAL_KEY
) {
390 bufp
= __get_buf(hashp
, bp
[bp
[0] - 1], bufp
, 0);
393 bp
= (u_short
*)bufp
->page
;
397 if (bp
[ndx
+ 1] == FULL_KEY
) {
398 bufp
= __get_buf(hashp
, bp
[bp
[0] - 1], bufp
, 0);
401 bp
= (u_short
*)bufp
->page
;
403 save_addr
= save_p
->addr
;
407 if (!FREESPACE(bp
)) {
409 * This is a hack. We can't distinguish between
410 * FULL_KEY_DATA that contains complete data or
411 * incomplete data, so we require that if the data
412 * is complete, there is at least 1 byte of free
418 save_addr
= bufp
->addr
;
419 bufp
= __get_buf(hashp
, bp
[bp
[0] - 1], bufp
, 0);
422 bp
= (u_short
*)bufp
->page
;
424 /* The data is all on one page. */
427 val
->data
= (u_char
*)tp
+ off
;
428 val
->size
= bp
[1] - off
;
430 if (bp
[0] == 2) { /* No more buckets in
436 hashp
->cpage
= __get_buf(hashp
,
437 bp
[bp
[0] - 1], bufp
, 0);
442 hashp
->cpage
->page
)[0]) {
451 val
->size
= collect_data(hashp
, bufp
, (int)len
, set_current
);
454 if (save_p
->addr
!= save_addr
) {
455 /* We are pretty short on buffers. */
456 errno
= EINVAL
; /* OUT OF BUFFERS */
459 memmove(hashp
->tmp_buf
, (save_p
->page
) + off
, len
);
460 val
->data
= (u_char
*)hashp
->tmp_buf
;
464 * Count how big the total datasize is by recursing through the pages. Then
465 * allocate a buffer and copy the data as you recurse up.
468 collect_data(hashp
, bufp
, len
, set
)
473 register u_short
*bp
;
481 mylen
= hashp
->BSIZE
- bp
[1];
482 save_addr
= bufp
->addr
;
484 if (bp
[2] == FULL_KEY_DATA
) { /* End of Data */
485 totlen
= len
+ mylen
;
487 free(hashp
->tmp_buf
);
488 if ((hashp
->tmp_buf
= (char *)malloc(totlen
)) == NULL
)
492 if (bp
[0] == 2) { /* No more buckets in chain */
497 __get_buf(hashp
, bp
[bp
[0] - 1], bufp
, 0);
500 else if (!((u_short
*)hashp
->cpage
->page
)[0]) {
507 xbp
= __get_buf(hashp
, bp
[bp
[0] - 1], bufp
, 0);
508 if (!xbp
|| ((totlen
=
509 collect_data(hashp
, xbp
, len
+ mylen
, set
)) < 1))
512 if (bufp
->addr
!= save_addr
) {
513 errno
= EINVAL
; /* Out of buffers. */
516 memmove(&hashp
->tmp_buf
[len
], (bufp
->page
) + bp
[1], mylen
);
521 * Fill in the key and data for this big pair.
524 __big_keydata(hashp
, bufp
, key
, val
, set
)
530 key
->size
= collect_key(hashp
, bufp
, 0, val
, set
);
533 key
->data
= (u_char
*)hashp
->tmp_key
;
538 * Count how big the total key size is by recursing through the pages. Then
539 * collect the data, allocate a buffer and copy the key as you recurse up.
542 collect_key(hashp
, bufp
, len
, val
, set
)
552 u_short
*bp
, save_addr
;
556 mylen
= hashp
->BSIZE
- bp
[1];
558 save_addr
= bufp
->addr
;
559 totlen
= len
+ mylen
;
560 if (bp
[2] == FULL_KEY
|| bp
[2] == FULL_KEY_DATA
) { /* End of Key. */
561 if (hashp
->tmp_key
!= NULL
)
562 free(hashp
->tmp_key
);
563 if ((hashp
->tmp_key
= (char *)malloc(totlen
)) == NULL
)
565 if (__big_return(hashp
, bufp
, 1, val
, set
))
568 xbp
= __get_buf(hashp
, bp
[bp
[0] - 1], bufp
, 0);
569 if (!xbp
|| ((totlen
=
570 collect_key(hashp
, xbp
, totlen
, val
, set
)) < 1))
573 if (bufp
->addr
!= save_addr
) {
574 errno
= EINVAL
; /* MIS -- OUT OF BUFFERS */
577 memmove(&hashp
->tmp_key
[len
], (bufp
->page
) + bp
[1], mylen
);
587 __big_split(hashp
, op
, np
, big_keyp
, addr
, obucket
, ret
)
589 BUFHEAD
*op
; /* Pointer to where to put keys that go in old bucket */
590 BUFHEAD
*np
; /* Pointer to new bucket page */
591 /* Pointer to first page containing the big key/data */
593 int addr
; /* Address of big_keyp */
594 u_int obucket
;/* Old Bucket */
597 register BUFHEAD
*tmpp
;
598 register u_short
*tp
;
602 u_short free_space
, n
, off
;
606 /* Now figure out where the big key/data goes */
607 if (__big_keydata(hashp
, big_keyp
, &key
, &val
, 0))
609 change
= (__call_hash(hashp
, key
.data
, key
.size
) != obucket
);
611 if (ret
->next_addr
= __find_last_page(hashp
, &big_keyp
)) {
613 __get_buf(hashp
, ret
->next_addr
, big_keyp
, 0)))
618 /* Now make one of np/op point to the big key/data pair */
620 assert(np
->ovfl
== NULL
);
627 tmpp
->flags
|= BUF_MOD
;
629 (void)fprintf(stderr
,
630 "BIG_SPLIT: %d->ovfl was %d is now %d\n", tmpp
->addr
,
631 (tmpp
->ovfl
? tmpp
->ovfl
->addr
: 0), (bp
? bp
->addr
: 0));
633 tmpp
->ovfl
= bp
; /* one of op/np point to big_keyp */
634 tp
= (u_short
*)tmpp
->page
;
636 assert(FREESPACE(tp
) >= OVFLSIZE
);
640 free_space
= FREESPACE(tp
);
641 tp
[++n
] = (u_short
)addr
;
645 FREESPACE(tp
) = free_space
- OVFLSIZE
;
648 * Finally, set the new and old return values. BIG_KEYP contains a
649 * pointer to the last page of the big key_data pair. Make sure that
650 * big_keyp has no following page (2 elements) or create an empty
657 tp
= (u_short
*)big_keyp
->page
;
658 big_keyp
->flags
|= BUF_MOD
;
661 * There may be either one or two offsets on this page. If
662 * there is one, then the overflow page is linked on normally
663 * and tp[4] is OVFLPAGE. If there are two, tp[4] contains
664 * the second offset and needs to get stuffed in after the
665 * next overflow page is added.
668 free_space
= FREESPACE(tp
);
671 FREESPACE(tp
) = free_space
+ OVFLSIZE
;
673 tmpp
= __add_ovflpage(hashp
, big_keyp
);