1 | /* bucket.c - The routines for playing with hash buckets. */
|
---|
2 |
|
---|
3 | /* This file is part of GDBM, the GNU data base manager, by Philip A. Nelson.
|
---|
4 | Copyright (C) 1990, 1991, 1993 Free Software Foundation, Inc.
|
---|
5 |
|
---|
6 | GDBM is free software; you can redistribute it and/or modify
|
---|
7 | it under the terms of the GNU General Public License as published by
|
---|
8 | the Free Software Foundation; either version 2, or (at your option)
|
---|
9 | any later version.
|
---|
10 |
|
---|
11 | GDBM is distributed in the hope that it will be useful,
|
---|
12 | but WITHOUT ANY WARRANTY; without even the implied warranty of
|
---|
13 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
---|
14 | GNU General Public License for more details.
|
---|
15 |
|
---|
16 | You should have received a copy of the GNU General Public License
|
---|
17 | along with GDBM; see the file COPYING. If not, write to
|
---|
18 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
|
---|
19 |
|
---|
20 | You may contact the author by:
|
---|
21 | e-mail: [email protected]
|
---|
22 | us-mail: Philip A. Nelson
|
---|
23 | Computer Science Department
|
---|
24 | Western Washington University
|
---|
25 | Bellingham, WA 98226
|
---|
26 |
|
---|
27 | *************************************************************************/
|
---|
28 |
|
---|
29 |
|
---|
30 | /* include system configuration before all else. */
|
---|
31 | #include "autoconf.h"
|
---|
32 |
|
---|
33 | #include "gdbmdefs.h"
|
---|
34 |
|
---|
35 | #include "gsdlmods.h"
|
---|
36 |
|
---|
37 |
|
---|
38 | /* Initializing a new hash buckets sets all bucket entries to -1 hash value. */
|
---|
39 | void
|
---|
40 | _gdbm_new_bucket (dbf, bucket, bits)
|
---|
41 | gdbm_file_info *dbf;
|
---|
42 | hash_bucket *bucket;
|
---|
43 | int bits;
|
---|
44 | {
|
---|
45 | int index;
|
---|
46 |
|
---|
47 | /* Initialize the avail block. */
|
---|
48 | bucket->av_count = 0;
|
---|
49 |
|
---|
50 | /* Set the information fields first. */
|
---|
51 | bucket->bucket_bits = bits;
|
---|
52 | bucket->count = 0;
|
---|
53 |
|
---|
54 | /* Initialize all bucket elements. */
|
---|
55 | for (index = 0; index < dbf->header->bucket_elems; index++)
|
---|
56 | bucket->h_table[index].hash_value = -1;
|
---|
57 | }
|
---|
58 |
|
---|
59 |
|
---|
60 |
|
---|
61 | /* Find a bucket for DBF that is pointed to by the bucket directory from
|
---|
62 | location DIR_INDEX. The bucket cache is first checked to see if it
|
---|
63 | is already in memory. If not, a bucket may be tossed to read the new
|
---|
64 | bucket. In any case, the requested bucket is make the "current" bucket
|
---|
65 | and dbf->bucket points to the correct bucket. */
|
---|
66 |
|
---|
67 | void
|
---|
68 | _gdbm_get_bucket (dbf, dir_index)
|
---|
69 | gdbm_file_info *dbf;
|
---|
70 | int dir_index;
|
---|
71 | {
|
---|
72 | off_t bucket_adr; /* The address of the correct hash bucket. */
|
---|
73 | int num_bytes; /* The number of bytes read. */
|
---|
74 | off_t file_pos; /* The return address for lseek. */
|
---|
75 | register int index; /* Loop index. */
|
---|
76 | int b;
|
---|
77 |
|
---|
78 | /* Initial set up. */
|
---|
79 | dbf->bucket_dir = dir_index;
|
---|
80 | bucket_adr = dbf->dir [dir_index];
|
---|
81 |
|
---|
82 | if (dbf->bucket_cache == NULL)
|
---|
83 | {
|
---|
84 | if(_gdbm_init_cache(dbf, DEFAULT_CACHESIZE) == -1)
|
---|
85 | _gdbm_fatal(dbf, "couldn't init cache");
|
---|
86 | }
|
---|
87 |
|
---|
88 | /* Is that one is not already current, we must find it. */
|
---|
89 | if (dbf->cache_entry->ca_adr != bucket_adr)
|
---|
90 | {
|
---|
91 | /* Look in the cache. */
|
---|
92 | for (index = 0; index < dbf->cache_size; index++)
|
---|
93 | {
|
---|
94 | if (dbf->bucket_cache[index].ca_adr == bucket_adr)
|
---|
95 | {
|
---|
96 | dbf->bucket = dbf->bucket_cache[index].ca_bucket;
|
---|
97 | dbf->cache_entry = &dbf->bucket_cache[index];
|
---|
98 | return;
|
---|
99 | }
|
---|
100 | }
|
---|
101 |
|
---|
102 | /* It is not in the cache, read it from the disk. */
|
---|
103 | dbf->last_read = (dbf->last_read + 1) % dbf->cache_size;
|
---|
104 | if (dbf->bucket_cache[dbf->last_read].ca_changed)
|
---|
105 | _gdbm_write_bucket (dbf, &dbf->bucket_cache[dbf->last_read]);
|
---|
106 | dbf->bucket_cache[dbf->last_read].ca_adr = bucket_adr;
|
---|
107 | dbf->bucket = dbf->bucket_cache[dbf->last_read].ca_bucket;
|
---|
108 | dbf->cache_entry = &dbf->bucket_cache[dbf->last_read];
|
---|
109 | dbf->cache_entry->ca_data.elem_loc = -1;
|
---|
110 | dbf->cache_entry->ca_changed = FALSE;
|
---|
111 |
|
---|
112 | /* Read the bucket. */
|
---|
113 | file_pos = lseek (dbf->desc, bucket_adr, L_SET);
|
---|
114 | if (file_pos != bucket_adr)
|
---|
115 | _gdbm_fatal (dbf, "lseek error");
|
---|
116 |
|
---|
117 | num_bytes = read (dbf->desc, dbf->bucket, dbf->header->bucket_size);
|
---|
118 | if (num_bytes != dbf->header->bucket_size)
|
---|
119 | _gdbm_fatal (dbf, "read error");
|
---|
120 |
|
---|
121 | // GREENSTONE CUSTOMISATION: Swap each value in the bucket if the GDBM file is the wrong endianness
|
---|
122 | if (wrong_endianness)
|
---|
123 | {
|
---|
124 | endian_swap_array((int*) dbf->bucket, dbf->header->bucket_size);
|
---|
125 |
|
---|
126 | // There is a 4-byte string in each hash table entry that we need to leave alone (because character
|
---|
127 | // strings are not affected by endianness), so convert these back to how they were originally
|
---|
128 | for (b = 0; b < dbf->header->bucket_elems; b++)
|
---|
129 | {
|
---|
130 | int* key_start = (int*) dbf->bucket->h_table[b].key_start;
|
---|
131 | *key_start = endian_swap (*key_start);
|
---|
132 | }
|
---|
133 | }
|
---|
134 | }
|
---|
135 |
|
---|
136 | return;
|
---|
137 | }
|
---|
138 |
|
---|
139 |
|
---|
140 | /* Split the current bucket. This includes moving all items in the bucket to
|
---|
141 | a new bucket. This doesn't require any disk reads because all hash values
|
---|
142 | are stored in the buckets. Splitting the current bucket may require
|
---|
143 | doubling the size of the hash directory. */
|
---|
144 | void
|
---|
145 | _gdbm_split_bucket (dbf, next_insert)
|
---|
146 | gdbm_file_info *dbf;
|
---|
147 | int next_insert;
|
---|
148 | {
|
---|
149 | hash_bucket *bucket[2]; /* Pointers to the new buckets. */
|
---|
150 |
|
---|
151 | int new_bits; /* The number of bits for the new buckets. */
|
---|
152 | int cache_0; /* Location in the cache for the buckets. */
|
---|
153 | int cache_1;
|
---|
154 | off_t adr_0; /* File address of the new bucket 0. */
|
---|
155 | off_t adr_1; /* File address of the new bucket 1. */
|
---|
156 | avail_elem old_bucket; /* Avail Struct for the old bucket. */
|
---|
157 |
|
---|
158 | off_t dir_start0; /* Used in updating the directory. */
|
---|
159 | off_t dir_start1;
|
---|
160 | off_t dir_end;
|
---|
161 |
|
---|
162 | off_t *new_dir; /* Pointer to the new directory. */
|
---|
163 | off_t dir_adr; /* Address of the new directory. */
|
---|
164 | int dir_size; /* Size of the new directory. */
|
---|
165 | off_t old_adr[31]; /* Address of the old directories. */
|
---|
166 | int old_size[31]; /* Size of the old directories. */
|
---|
167 | int old_count; /* Number of old directories. */
|
---|
168 |
|
---|
169 | int index; /* Used in array indexing. */
|
---|
170 | int index1; /* Used in array indexing. */
|
---|
171 | int elem_loc; /* Location in new bucket to put element. */
|
---|
172 | bucket_element *old_el; /* Pointer into the old bucket. */
|
---|
173 | int select; /* Used to index bucket during movement. */
|
---|
174 |
|
---|
175 |
|
---|
176 | /* No directories are yet old. */
|
---|
177 | old_count = 0;
|
---|
178 |
|
---|
179 | if (dbf->bucket_cache == NULL)
|
---|
180 | {
|
---|
181 | if(_gdbm_init_cache(dbf, DEFAULT_CACHESIZE) == -1)
|
---|
182 | _gdbm_fatal(dbf, "couldn't init cache");
|
---|
183 | }
|
---|
184 |
|
---|
185 | while (dbf->bucket->count == dbf->header->bucket_elems)
|
---|
186 | {
|
---|
187 | /* Initialize the "new" buckets in the cache. */
|
---|
188 | do
|
---|
189 | {
|
---|
190 | dbf->last_read = (dbf->last_read + 1) % dbf->cache_size;
|
---|
191 | cache_0 = dbf->last_read;
|
---|
192 | }
|
---|
193 | while (dbf->bucket_cache[cache_0].ca_bucket == dbf->bucket);
|
---|
194 | bucket[0] = dbf->bucket_cache[cache_0].ca_bucket;
|
---|
195 | if (dbf->bucket_cache[cache_0].ca_changed)
|
---|
196 | _gdbm_write_bucket (dbf, &dbf->bucket_cache[cache_0]);
|
---|
197 | do
|
---|
198 | {
|
---|
199 | dbf->last_read = (dbf->last_read + 1) % dbf->cache_size;
|
---|
200 | cache_1 = dbf->last_read;
|
---|
201 | }
|
---|
202 | while (dbf->bucket_cache[cache_1].ca_bucket == dbf->bucket);
|
---|
203 | bucket[1] = dbf->bucket_cache[cache_1].ca_bucket;
|
---|
204 | if (dbf->bucket_cache[cache_1].ca_changed)
|
---|
205 | _gdbm_write_bucket (dbf, &dbf->bucket_cache[cache_1]);
|
---|
206 | new_bits = dbf->bucket->bucket_bits+1;
|
---|
207 | _gdbm_new_bucket (dbf, bucket[0], new_bits);
|
---|
208 | _gdbm_new_bucket (dbf, bucket[1], new_bits);
|
---|
209 | adr_0 = _gdbm_alloc (dbf, dbf->header->bucket_size);
|
---|
210 | dbf->bucket_cache[cache_0].ca_adr = adr_0;
|
---|
211 | adr_1 = _gdbm_alloc (dbf, dbf->header->bucket_size);
|
---|
212 | dbf->bucket_cache[cache_1].ca_adr = adr_1;
|
---|
213 |
|
---|
214 |
|
---|
215 |
|
---|
216 | /* Double the directory size if necessary. */
|
---|
217 | if (dbf->header->dir_bits == dbf->bucket->bucket_bits)
|
---|
218 | {
|
---|
219 | dir_size = dbf->header->dir_size * 2;
|
---|
220 | dir_adr = _gdbm_alloc (dbf, dir_size);
|
---|
221 | new_dir = (off_t *) malloc (dir_size);
|
---|
222 | if (new_dir == NULL) _gdbm_fatal (dbf, "malloc error");
|
---|
223 | for (index = 0;
|
---|
224 | index < dbf->header->dir_size/sizeof (off_t); index++)
|
---|
225 | {
|
---|
226 | new_dir[2*index] = dbf->dir[index];
|
---|
227 | new_dir[2*index+1] = dbf->dir[index];
|
---|
228 | }
|
---|
229 |
|
---|
230 | /* Update header. */
|
---|
231 | old_adr[old_count] = dbf->header->dir;
|
---|
232 | dbf->header->dir = dir_adr;
|
---|
233 | old_size[old_count] = dbf->header->dir_size;
|
---|
234 | dbf->header->dir_size = dir_size;
|
---|
235 | dbf->header->dir_bits = new_bits;
|
---|
236 | old_count++;
|
---|
237 |
|
---|
238 | /* Now update dbf. */
|
---|
239 | dbf->header_changed = TRUE;
|
---|
240 | dbf->bucket_dir *= 2;
|
---|
241 | free (dbf->dir);
|
---|
242 | dbf->dir = new_dir;
|
---|
243 | }
|
---|
244 |
|
---|
245 | /* Copy all elements in dbf->bucket into the new buckets. */
|
---|
246 | for (index = 0; index < dbf->header->bucket_elems; index++)
|
---|
247 | {
|
---|
248 | old_el = & (dbf->bucket->h_table[index]);
|
---|
249 | select = (old_el->hash_value >> (31-new_bits)) & 1;
|
---|
250 | elem_loc = old_el->hash_value % dbf->header->bucket_elems;
|
---|
251 | while (bucket[select]->h_table[elem_loc].hash_value != -1)
|
---|
252 | elem_loc = (elem_loc + 1) % dbf->header->bucket_elems;
|
---|
253 | bucket[select]->h_table[elem_loc] = *old_el;
|
---|
254 | bucket[select]->count += 1;
|
---|
255 | }
|
---|
256 |
|
---|
257 | /* Allocate avail space for the bucket[1]. */
|
---|
258 | bucket[1]->bucket_avail[0].av_adr
|
---|
259 | = _gdbm_alloc (dbf, dbf->header->block_size);
|
---|
260 | bucket[1]->bucket_avail[0].av_size = dbf->header->block_size;
|
---|
261 | bucket[1]->av_count = 1;
|
---|
262 |
|
---|
263 | /* Copy the avail elements in dbf->bucket to bucket[0]. */
|
---|
264 | bucket[0]->av_count = dbf->bucket->av_count;
|
---|
265 | index = 0;
|
---|
266 | index1 = 0;
|
---|
267 | if (bucket[0]->av_count == BUCKET_AVAIL)
|
---|
268 | {
|
---|
269 | /* The avail is full, move the first one to bucket[1]. */
|
---|
270 | _gdbm_put_av_elem (dbf->bucket->bucket_avail[0],
|
---|
271 | bucket[1]->bucket_avail,
|
---|
272 | &bucket[1]->av_count, FALSE);
|
---|
273 | index = 1;
|
---|
274 | bucket[0]->av_count --;
|
---|
275 | }
|
---|
276 | for (; index < dbf->bucket->av_count; index++)
|
---|
277 | {
|
---|
278 | bucket[0]->bucket_avail[index1++] = dbf->bucket->bucket_avail[index];
|
---|
279 | }
|
---|
280 |
|
---|
281 | /* Update the directory. We have new file addresses for both buckets. */
|
---|
282 | dir_start1 = (dbf->bucket_dir >> (dbf->header->dir_bits - new_bits)) | 1;
|
---|
283 | dir_end = (dir_start1 + 1) << (dbf->header->dir_bits - new_bits);
|
---|
284 | dir_start1 = dir_start1 << (dbf->header->dir_bits - new_bits);
|
---|
285 | dir_start0 = dir_start1 - (dir_end - dir_start1);
|
---|
286 | for (index = dir_start0; index < dir_start1; index++)
|
---|
287 | dbf->dir[index] = adr_0;
|
---|
288 | for (index = dir_start1; index < dir_end; index++)
|
---|
289 | dbf->dir[index] = adr_1;
|
---|
290 |
|
---|
291 |
|
---|
292 | /* Set changed flags. */
|
---|
293 | dbf->bucket_cache[cache_0].ca_changed = TRUE;
|
---|
294 | dbf->bucket_cache[cache_1].ca_changed = TRUE;
|
---|
295 | dbf->bucket_changed = TRUE;
|
---|
296 | dbf->directory_changed = TRUE;
|
---|
297 | dbf->second_changed = TRUE;
|
---|
298 |
|
---|
299 | /* Update the cache! */
|
---|
300 | dbf->bucket_dir = next_insert >> (31-dbf->header->dir_bits);
|
---|
301 |
|
---|
302 | /* Invalidate old cache entry. */
|
---|
303 | old_bucket.av_adr = dbf->cache_entry->ca_adr;
|
---|
304 | old_bucket.av_size = dbf->header->bucket_size;
|
---|
305 | dbf->cache_entry->ca_adr = 0;
|
---|
306 | dbf->cache_entry->ca_changed = FALSE;
|
---|
307 |
|
---|
308 | /* Set dbf->bucket to the proper bucket. */
|
---|
309 | if (dbf->dir[dbf->bucket_dir] == adr_0)
|
---|
310 | {
|
---|
311 | dbf->bucket = bucket[0];
|
---|
312 | dbf->cache_entry = &dbf->bucket_cache[cache_0];
|
---|
313 | _gdbm_put_av_elem (old_bucket,
|
---|
314 | bucket[1]->bucket_avail,
|
---|
315 | &bucket[1]->av_count, FALSE);
|
---|
316 | }
|
---|
317 | else
|
---|
318 | {
|
---|
319 | dbf->bucket = bucket[1];
|
---|
320 | dbf->cache_entry = &dbf->bucket_cache[cache_1];
|
---|
321 | _gdbm_put_av_elem (old_bucket,
|
---|
322 | bucket[0]->bucket_avail,
|
---|
323 | &bucket[0]->av_count, FALSE);
|
---|
324 | }
|
---|
325 |
|
---|
326 | }
|
---|
327 |
|
---|
328 | /* Get rid of old directories. */
|
---|
329 | for (index = 0; index < old_count; index++)
|
---|
330 | _gdbm_free (dbf, old_adr[index], old_size[index]);
|
---|
331 | }
|
---|
332 |
|
---|
333 |
|
---|
334 | /* The only place where a bucket is written. CA_ENTRY is the
|
---|
335 | cache entry containing the bucket to be written. */
|
---|
336 |
|
---|
337 | void
|
---|
338 | _gdbm_write_bucket (dbf, ca_entry)
|
---|
339 | gdbm_file_info *dbf;
|
---|
340 | cache_elem *ca_entry;
|
---|
341 | {
|
---|
342 | int num_bytes; /* The return value for write. */
|
---|
343 | off_t file_pos; /* The return value for lseek. */
|
---|
344 |
|
---|
345 | file_pos = lseek (dbf->desc, ca_entry->ca_adr, L_SET);
|
---|
346 | if (file_pos != ca_entry->ca_adr)
|
---|
347 | _gdbm_fatal (dbf, "lseek error");
|
---|
348 | num_bytes = write (dbf->desc, ca_entry->ca_bucket, dbf->header->bucket_size);
|
---|
349 | if (num_bytes != dbf->header->bucket_size)
|
---|
350 | _gdbm_fatal (dbf, "write error");
|
---|
351 | ca_entry->ca_changed = FALSE;
|
---|
352 | ca_entry->ca_data.hash_val = -1;
|
---|
353 | ca_entry->ca_data.elem_loc = -1;
|
---|
354 | }
|
---|