1 | /* Copyright (c) 2000-2002, 2004-2007 MySQL AB, 2009 Sun Microsystems, Inc. |
2 | Use is subject to license terms. |
3 | |
4 | This program is free software; you can redistribute it and/or modify |
5 | it under the terms of the GNU General Public License as published by |
6 | the Free Software Foundation; version 2 of the License. |
7 | |
8 | This program is distributed in the hope that it will be useful, |
9 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
10 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
11 | GNU General Public License for more details. |
12 | |
13 | You should have received a copy of the GNU General Public License |
14 | along with this program; if not, write to the Free Software |
15 | Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA */ |
16 | |
17 | /* Write a record to heap-databas */ |
18 | |
19 | #include "heapdef.h" |
20 | #ifdef __WIN__ |
21 | #include <fcntl.h> |
22 | #endif |
23 | |
24 | #define LOWFIND 1 |
25 | #define LOWUSED 2 |
26 | #define HIGHFIND 4 |
27 | #define HIGHUSED 8 |
28 | |
29 | static uchar *next_free_record_pos(HP_SHARE *info); |
30 | static HASH_INFO *hp_find_free_hash(HP_SHARE *info, HP_BLOCK *block, |
31 | ulong records); |
32 | |
33 | int heap_write(HP_INFO *info, const uchar *record) |
34 | { |
35 | HP_KEYDEF *keydef, *end; |
36 | uchar *pos; |
37 | HP_SHARE *share=info->s; |
38 | DBUG_ENTER("heap_write" ); |
39 | #ifndef DBUG_OFF |
40 | if (info->mode & O_RDONLY) |
41 | { |
42 | DBUG_RETURN(my_errno=EACCES); |
43 | } |
44 | #endif |
45 | if (!(pos=next_free_record_pos(share))) |
46 | DBUG_RETURN(my_errno); |
47 | share->changed=1; |
48 | |
49 | for (keydef = share->keydef, end = keydef + share->keys; keydef < end; |
50 | keydef++) |
51 | { |
52 | if ((*keydef->write_key)(info, keydef, record, pos)) |
53 | goto err; |
54 | } |
55 | |
56 | memcpy(pos,record,(size_t) share->reclength); |
57 | pos[share->visible]= 1; /* Mark record as not deleted */ |
58 | if (++share->records == share->blength) |
59 | share->blength+= share->blength; |
60 | info->s->key_version++; |
61 | info->current_ptr=pos; |
62 | info->current_hash_ptr=0; |
63 | info->update|=HA_STATE_AKTIV; |
64 | #if !defined(DBUG_OFF) && defined(EXTRA_HEAP_DEBUG) |
65 | DBUG_EXECUTE("check_heap" ,heap_check_heap(info, 0);); |
66 | #endif |
67 | if (share->auto_key) |
68 | heap_update_auto_increment(info, record); |
69 | DBUG_RETURN(0); |
70 | |
71 | err: |
72 | if (my_errno == HA_ERR_FOUND_DUPP_KEY) |
73 | DBUG_PRINT("info" ,("Duplicate key: %d" , (int) (keydef - share->keydef))); |
74 | info->errkey= (int) (keydef - share->keydef); |
75 | /* |
76 | We don't need to delete non-inserted key from rb-tree. Also, if |
77 | we got ENOMEM, the key wasn't inserted, so don't try to delete it |
78 | either. Otherwise for HASH index on HA_ERR_FOUND_DUPP_KEY the key |
79 | was inserted and we have to delete it. |
80 | */ |
81 | if (keydef->algorithm == HA_KEY_ALG_BTREE || my_errno == ENOMEM) |
82 | { |
83 | keydef--; |
84 | } |
85 | while (keydef >= share->keydef) |
86 | { |
87 | if ((*keydef->delete_key)(info, keydef, record, pos, 0)) |
88 | break; |
89 | keydef--; |
90 | } |
91 | |
92 | share->deleted++; |
93 | *((uchar**) pos)=share->del_link; |
94 | share->del_link=pos; |
95 | pos[share->visible]= 0; /* Record deleted */ |
96 | |
97 | DBUG_RETURN(my_errno); |
98 | } /* heap_write */ |
99 | |
100 | /* |
101 | Write a key to rb_tree-index |
102 | */ |
103 | |
104 | int hp_rb_write_key(HP_INFO *info, HP_KEYDEF *keyinfo, const uchar *record, |
105 | uchar *recpos) |
106 | { |
107 | heap_rb_param custom_arg; |
108 | size_t old_allocated; |
109 | |
110 | custom_arg.keyseg= keyinfo->seg; |
111 | custom_arg.key_length= hp_rb_make_key(keyinfo, info->recbuf, record, recpos); |
112 | if (keyinfo->flag & HA_NOSAME) |
113 | { |
114 | custom_arg.search_flag= SEARCH_FIND | SEARCH_UPDATE | SEARCH_INSERT; |
115 | keyinfo->rb_tree.flag= TREE_NO_DUPS; |
116 | } |
117 | else |
118 | { |
119 | custom_arg.search_flag= SEARCH_SAME; |
120 | keyinfo->rb_tree.flag= 0; |
121 | } |
122 | old_allocated= keyinfo->rb_tree.allocated; |
123 | if (!tree_insert(&keyinfo->rb_tree, (void*)info->recbuf, |
124 | custom_arg.key_length, &custom_arg)) |
125 | { |
126 | my_errno= HA_ERR_FOUND_DUPP_KEY; |
127 | return 1; |
128 | } |
129 | info->s->index_length+= (keyinfo->rb_tree.allocated-old_allocated); |
130 | return 0; |
131 | } |
132 | |
133 | /* Find where to place new record */ |
134 | |
135 | static uchar *next_free_record_pos(HP_SHARE *info) |
136 | { |
137 | int block_pos; |
138 | uchar *pos; |
139 | size_t length; |
140 | DBUG_ENTER("next_free_record_pos" ); |
141 | |
142 | if (info->del_link) |
143 | { |
144 | pos=info->del_link; |
145 | info->del_link= *((uchar**) pos); |
146 | info->deleted--; |
147 | DBUG_PRINT("exit" ,("Used old position: %p" , pos)); |
148 | DBUG_RETURN(pos); |
149 | } |
150 | if (!(block_pos=(info->records % info->block.records_in_block))) |
151 | { |
152 | if ((info->records > info->max_records && info->max_records) || |
153 | (info->data_length + info->index_length >= info->max_table_size)) |
154 | { |
155 | DBUG_PRINT("error" , |
156 | ("record file full. records: %lu max_records: %lu " |
157 | "data_length: %llu index_length: %llu " |
158 | "max_table_size: %llu" , |
159 | info->records, info->max_records, |
160 | info->data_length, info->index_length, |
161 | info->max_table_size)); |
162 | my_errno=HA_ERR_RECORD_FILE_FULL; |
163 | DBUG_RETURN(NULL); |
164 | } |
165 | if (hp_get_new_block(info, &info->block,&length)) |
166 | DBUG_RETURN(NULL); |
167 | info->data_length+=length; |
168 | } |
169 | DBUG_PRINT("exit" ,("Used new position: %p" , |
170 | ((uchar*) info->block.level_info[0].last_blocks+ |
171 | block_pos * info->block.recbuffer))); |
172 | DBUG_RETURN((uchar*) info->block.level_info[0].last_blocks+ |
173 | block_pos*info->block.recbuffer); |
174 | } |
175 | |
176 | |
177 | /* |
178 | Write a hash-key to the hash-index |
179 | SYNOPSIS |
180 | info Heap table info |
181 | keyinfo Key info |
182 | record Table record to added |
183 | recpos Memory buffer where the table record will be stored if added |
184 | successfully |
185 | NOTE |
186 | Hash index uses HP_BLOCK structure as a 'growable array' of HASH_INFO |
187 | structs. Array size == number of entries in hash index. |
188 | hp_mask(hp_rec_hashnr()) maps hash entries values to hash array positions. |
189 | If there are several hash entries with the same hash array position P, |
190 | they are connected in a linked list via HASH_INFO::next_key. The first |
191 | list element is located at position P, next elements are located at |
192 | positions for which there is no record that should be located at that |
193 | position. The order of elements in the list is arbitrary. |
194 | |
195 | RETURN |
196 | 0 - OK |
197 | -1 - Out of memory |
198 | HA_ERR_FOUND_DUPP_KEY - Duplicate record on unique key. The record was |
199 | still added and the caller must call hp_delete_key for it. |
200 | */ |
201 | |
202 | int hp_write_key(HP_INFO *info, HP_KEYDEF *keyinfo, |
203 | const uchar *record, uchar *recpos) |
204 | { |
205 | HP_SHARE *share = info->s; |
206 | int flag; |
207 | ulong halfbuff,hashnr,first_index; |
208 | ulong UNINIT_VAR(hash_of_key), UNINIT_VAR(hash_of_key2); |
209 | uchar *UNINIT_VAR(ptr_to_rec),*UNINIT_VAR(ptr_to_rec2); |
210 | HASH_INFO *empty,*UNINIT_VAR(gpos),*UNINIT_VAR(gpos2),*pos; |
211 | DBUG_ENTER("hp_write_key" ); |
212 | |
213 | flag=0; |
214 | if (!(empty= hp_find_free_hash(share,&keyinfo->block,share->records))) |
215 | DBUG_RETURN(-1); /* No more memory */ |
216 | halfbuff= (long) share->blength >> 1; |
217 | pos= hp_find_hash(&keyinfo->block,(first_index=share->records-halfbuff)); |
218 | |
219 | /* |
220 | We're about to add one more hash array position, with hash_mask=#records. |
221 | The number of hash positions will change and some entries might need to |
222 | be relocated to the newly added position. Those entries are currently |
223 | members of the list that starts at #first_index position (this is |
224 | guaranteed by properties of hp_mask(hp_rec_hashnr(X)) mapping function) |
225 | At #first_index position currently there may be either: |
226 | a) An entry with hashnr != first_index. We don't need to move it. |
227 | or |
228 | b) A list of items with hash_mask=first_index. The list contains entries |
229 | of 2 types: |
230 | 1) entries that should be relocated to the list that starts at new |
231 | position we're adding ('uppper' list) |
232 | 2) entries that should be left in the list starting at #first_index |
233 | position ('lower' list) |
234 | */ |
235 | if (pos != empty) /* If some records */ |
236 | { |
237 | do |
238 | { |
239 | hashnr = pos->hash_of_key; |
240 | if (flag == 0) |
241 | { |
242 | /* |
243 | First loop, bail out if we're dealing with case a) from above |
244 | comment |
245 | */ |
246 | if (hp_mask(hashnr, share->blength, share->records) != first_index) |
247 | break; |
248 | } |
249 | /* |
250 | flag & LOWFIND - found a record that should be put into lower position |
251 | flag & LOWUSED - lower position occupied by the record |
252 | Same for HIGHFIND and HIGHUSED and 'upper' position |
253 | |
254 | gpos - ptr to last element in lower position's list |
255 | gpos2 - ptr to last element in upper position's list |
256 | |
257 | ptr_to_rec - ptr to last entry that should go into lower list. |
258 | ptr_to_rec2 - same for upper list. |
259 | */ |
260 | if (!(hashnr & halfbuff)) |
261 | { |
262 | /* Key should be put into 'lower' list */ |
263 | if (!(flag & LOWFIND)) |
264 | { |
265 | /* key is the first element to go into lower position */ |
266 | if (flag & HIGHFIND) |
267 | { |
268 | flag=LOWFIND | HIGHFIND; |
269 | /* key shall be moved to the current empty position */ |
270 | gpos=empty; |
271 | empty=pos; /* This place is now free */ |
272 | } |
273 | else |
274 | { |
275 | /* |
276 | We can only get here at first iteration: key is at 'lower' |
277 | position pos and should be left here. |
278 | */ |
279 | flag=LOWFIND | LOWUSED; |
280 | gpos=pos; |
281 | } |
282 | } |
283 | else |
284 | { |
285 | /* Already have another key for lower position */ |
286 | if (!(flag & LOWUSED)) |
287 | { |
288 | /* Change link of previous lower-list key */ |
289 | gpos->ptr_to_rec= ptr_to_rec; |
290 | gpos->next_key= pos; |
291 | gpos->hash_of_key= hash_of_key; |
292 | flag= (flag & HIGHFIND) | (LOWFIND | LOWUSED); |
293 | } |
294 | gpos=pos; |
295 | } |
296 | ptr_to_rec= pos->ptr_to_rec; |
297 | hash_of_key= pos->hash_of_key; |
298 | } |
299 | else |
300 | { |
301 | /* key will be put into 'higher' list */ |
302 | if (!(flag & HIGHFIND)) |
303 | { |
304 | flag= (flag & LOWFIND) | HIGHFIND; |
305 | /* key shall be moved to the last (empty) position */ |
306 | gpos2= empty; |
307 | empty= pos; |
308 | } |
309 | else |
310 | { |
311 | if (!(flag & HIGHUSED)) |
312 | { |
313 | /* Change link of previous upper-list key and save */ |
314 | gpos2->ptr_to_rec= ptr_to_rec2; |
315 | gpos2->next_key= pos; |
316 | gpos2->hash_of_key= hash_of_key2; |
317 | flag= (flag & LOWFIND) | (HIGHFIND | HIGHUSED); |
318 | } |
319 | gpos2=pos; |
320 | } |
321 | ptr_to_rec2= pos->ptr_to_rec; |
322 | hash_of_key2= pos->hash_of_key; |
323 | } |
324 | } |
325 | while ((pos=pos->next_key)); |
326 | |
327 | if ((flag & (LOWFIND | HIGHFIND)) == (LOWFIND | HIGHFIND)) |
328 | { |
329 | /* |
330 | If both 'higher' and 'lower' list have at least one element, now |
331 | there are two hash buckets instead of one. |
332 | */ |
333 | keyinfo->hash_buckets++; |
334 | } |
335 | |
336 | if ((flag & (LOWFIND | LOWUSED)) == LOWFIND) |
337 | { |
338 | gpos->ptr_to_rec= ptr_to_rec; |
339 | gpos->hash_of_key= hash_of_key; |
340 | gpos->next_key= 0; |
341 | } |
342 | if ((flag & (HIGHFIND | HIGHUSED)) == HIGHFIND) |
343 | { |
344 | gpos2->ptr_to_rec= ptr_to_rec2; |
345 | gpos2->hash_of_key= hash_of_key2; |
346 | gpos2->next_key= 0; |
347 | } |
348 | } |
349 | /* Check if we are at the empty position */ |
350 | |
351 | hash_of_key= hp_rec_hashnr(keyinfo, record); |
352 | pos=hp_find_hash(&keyinfo->block, |
353 | hp_mask(hash_of_key, share->blength, share->records + 1)); |
354 | if (pos == empty) |
355 | { |
356 | pos->ptr_to_rec= recpos; |
357 | pos->hash_of_key= hash_of_key; |
358 | pos->next_key= 0; |
359 | keyinfo->hash_buckets++; |
360 | } |
361 | else |
362 | { |
363 | /* Check if more records in same hash-nr family */ |
364 | empty[0]=pos[0]; |
365 | gpos=hp_find_hash(&keyinfo->block, |
366 | hp_mask(pos->hash_of_key, |
367 | share->blength, share->records + 1)); |
368 | |
369 | pos->ptr_to_rec= recpos; |
370 | pos->hash_of_key= hash_of_key; |
371 | if (pos == gpos) |
372 | pos->next_key=empty; |
373 | else |
374 | { |
375 | keyinfo->hash_buckets++; |
376 | pos->next_key= 0; |
377 | hp_movelink(pos, gpos, empty); |
378 | } |
379 | |
380 | /* Check if duplicated keys */ |
381 | if ((keyinfo->flag & HA_NOSAME) && pos == gpos && |
382 | (!(keyinfo->flag & HA_NULL_PART_KEY) || |
383 | !hp_if_null_in_key(keyinfo, record))) |
384 | { |
385 | pos=empty; |
386 | do |
387 | { |
388 | if (pos->hash_of_key == hash_of_key && |
389 | ! hp_rec_key_cmp(keyinfo, record, pos->ptr_to_rec)) |
390 | { |
391 | DBUG_RETURN(my_errno=HA_ERR_FOUND_DUPP_KEY); |
392 | } |
393 | } while ((pos=pos->next_key)); |
394 | } |
395 | } |
396 | DBUG_RETURN(0); |
397 | } |
398 | |
399 | /* Returns ptr to block, and allocates block if neaded */ |
400 | |
401 | static HASH_INFO *hp_find_free_hash(HP_SHARE *info, |
402 | HP_BLOCK *block, ulong records) |
403 | { |
404 | ulong block_pos; |
405 | size_t length; |
406 | |
407 | if (records < block->last_allocated) |
408 | return hp_find_hash(block,records); |
409 | if (!(block_pos=(records % block->records_in_block))) |
410 | { |
411 | if (hp_get_new_block(info, block, &length)) |
412 | return(NULL); |
413 | info->index_length+=length; |
414 | } |
415 | block->last_allocated=records+1; |
416 | return((HASH_INFO*) ((uchar*) block->level_info[0].last_blocks+ |
417 | block_pos*block->recbuffer)); |
418 | } |
419 | |