1
by brian
clean slate |
1 |
/* Copyright (C) 2000-2006 MySQL AB
|
2 |
||
3 |
This program is free software; you can redistribute it and/or modify
|
|
4 |
it under the terms of the GNU General Public License as published by
|
|
5 |
the Free Software Foundation; version 2 of the License.
|
|
6 |
||
7 |
This program is distributed in the hope that it will be useful,
|
|
8 |
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
9 |
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
10 |
GNU General Public License for more details.
|
|
11 |
||
12 |
You should have received a copy of the GNU General Public License
|
|
13 |
along with this program; if not, write to the Free Software
|
|
14 |
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */
|
|
15 |
||
16 |
||
17 |
/* Functions to handle keys and fields in forms */
|
|
18 |
||
19 |
#include "mysql_priv.h" |
|
20 |
||
21 |
/*
|
|
22 |
Search after a key that starts with 'field'
|
|
23 |
||
24 |
SYNOPSIS
|
|
25 |
find_ref_key()
|
|
26 |
key First key to check
|
|
27 |
key_count How many keys to check
|
|
28 |
record Start of record
|
|
29 |
field Field to search after
|
|
30 |
key_length On partial match, contains length of fields before
|
|
31 |
field
|
|
32 |
keypart key part # of a field
|
|
33 |
||
34 |
NOTES
|
|
35 |
Used when calculating key for NEXT_NUMBER
|
|
36 |
||
37 |
IMPLEMENTATION
|
|
38 |
If no key starts with field test if field is part of some key. If we find
|
|
39 |
one, then return first key and set key_length to the number of bytes
|
|
40 |
preceding 'field'.
|
|
41 |
||
42 |
RETURN
|
|
43 |
-1 field is not part of the key
|
|
44 |
# Key part for key matching key.
|
|
45 |
key_length is set to length of key before (not including) field
|
|
46 |
*/
|
|
47 |
||
48 |
int find_ref_key(KEY *key, uint key_count, uchar *record, Field *field, |
|
49 |
uint *key_length, uint *keypart) |
|
50 |
{
|
|
51 |
register int i; |
|
52 |
register KEY *key_info; |
|
53 |
uint fieldpos; |
|
54 |
||
55 |
fieldpos= field->offset(record); |
|
56 |
||
57 |
/* Test if some key starts as fieldpos */
|
|
58 |
for (i= 0, key_info= key ; |
|
59 |
i < (int) key_count ; |
|
60 |
i++, key_info++) |
|
61 |
{
|
|
62 |
if (key_info->key_part[0].offset == fieldpos) |
|
63 |
{ /* Found key. Calc keylength */ |
|
64 |
*key_length= *keypart= 0; |
|
65 |
return i; /* Use this key */ |
|
66 |
}
|
|
67 |
}
|
|
68 |
||
69 |
/* Test if some key contains fieldpos */
|
|
70 |
for (i= 0, key_info= key; |
|
71 |
i < (int) key_count ; |
|
72 |
i++, key_info++) |
|
73 |
{
|
|
74 |
uint j; |
|
75 |
KEY_PART_INFO *key_part; |
|
76 |
*key_length=0; |
|
77 |
for (j=0, key_part=key_info->key_part ; |
|
78 |
j < key_info->key_parts ; |
|
79 |
j++, key_part++) |
|
80 |
{
|
|
81 |
if (key_part->offset == fieldpos) |
|
82 |
{
|
|
83 |
*keypart= j; |
|
84 |
return i; /* Use this key */ |
|
85 |
}
|
|
86 |
*key_length+= key_part->store_length; |
|
87 |
}
|
|
88 |
}
|
|
89 |
return(-1); /* No key is ok */ |
|
90 |
}
|
|
91 |
||
92 |
||
93 |
/**
|
|
94 |
Copy part of a record that forms a key or key prefix to a buffer.
|
|
95 |
||
96 |
The function takes a complete table record (as e.g. retrieved by
|
|
97 |
handler::index_read()), and a description of an index on the same table,
|
|
98 |
and extracts the first key_length bytes of the record which are part of a
|
|
99 |
key into to_key. If length == 0 then copy all bytes from the record that
|
|
100 |
form a key.
|
|
101 |
||
102 |
@param to_key buffer that will be used as a key
|
|
103 |
@param from_record full record to be copied from
|
|
104 |
@param key_info descriptor of the index
|
|
105 |
@param key_length specifies length of all keyparts that will be copied
|
|
106 |
*/
|
|
107 |
||
108 |
void key_copy(uchar *to_key, uchar *from_record, KEY *key_info, |
|
109 |
uint key_length) |
|
110 |
{
|
|
111 |
uint length; |
|
112 |
KEY_PART_INFO *key_part; |
|
113 |
||
114 |
if (key_length == 0) |
|
115 |
key_length= key_info->key_length; |
|
116 |
for (key_part= key_info->key_part; (int) key_length > 0; key_part++) |
|
117 |
{
|
|
118 |
if (key_part->null_bit) |
|
119 |
{
|
|
120 |
*to_key++= test(from_record[key_part->null_offset] & |
|
121 |
key_part->null_bit); |
|
122 |
key_length--; |
|
123 |
}
|
|
124 |
if (key_part->key_part_flag & HA_BLOB_PART || |
|
125 |
key_part->key_part_flag & HA_VAR_LENGTH_PART) |
|
126 |
{
|
|
127 |
key_length-= HA_KEY_BLOB_LENGTH; |
|
128 |
length= min(key_length, key_part->length); |
|
129 |
key_part->field->get_key_image(to_key, length, Field::itRAW); |
|
130 |
to_key+= HA_KEY_BLOB_LENGTH; |
|
131 |
}
|
|
132 |
else
|
|
133 |
{
|
|
134 |
length= min(key_length, key_part->length); |
|
135 |
Field *field= key_part->field; |
|
136 |
CHARSET_INFO *cs= field->charset(); |
|
137 |
uint bytes= field->get_key_image(to_key, length, Field::itRAW); |
|
138 |
if (bytes < length) |
|
139 |
cs->cset->fill(cs, (char*) to_key + bytes, length - bytes, ' '); |
|
140 |
}
|
|
141 |
to_key+= length; |
|
142 |
key_length-= length; |
|
143 |
}
|
|
144 |
}
|
|
145 |
||
146 |
||
147 |
/**
|
|
148 |
Zero the null components of key tuple.
|
|
149 |
*/
|
|
150 |
||
151 |
void key_zero_nulls(uchar *tuple, KEY *key_info) |
|
152 |
{
|
|
153 |
KEY_PART_INFO *key_part= key_info->key_part; |
|
154 |
KEY_PART_INFO *key_part_end= key_part + key_info->key_parts; |
|
155 |
for (; key_part != key_part_end; key_part++) |
|
156 |
{
|
|
157 |
if (key_part->null_bit && *tuple) |
|
158 |
bzero(tuple+1, key_part->store_length-1); |
|
159 |
tuple+= key_part->store_length; |
|
160 |
}
|
|
161 |
}
|
|
162 |
||
163 |
||
164 |
/*
|
|
165 |
Restore a key from some buffer to record.
|
|
166 |
||
167 |
This function converts a key into record format. It can be used in cases
|
|
168 |
when we want to return a key as a result row.
|
|
169 |
||
170 |
@param to_record record buffer where the key will be restored to
|
|
171 |
@param from_key buffer that contains a key
|
|
172 |
@param key_info descriptor of the index
|
|
173 |
@param key_length specifies length of all keyparts that will be restored
|
|
174 |
*/
|
|
175 |
||
176 |
void key_restore(uchar *to_record, uchar *from_key, KEY *key_info, |
|
177 |
uint key_length) |
|
178 |
{
|
|
179 |
uint length; |
|
180 |
KEY_PART_INFO *key_part; |
|
181 |
||
182 |
if (key_length == 0) |
|
183 |
{
|
|
184 |
key_length= key_info->key_length; |
|
185 |
}
|
|
186 |
for (key_part= key_info->key_part ; (int) key_length > 0 ; key_part++) |
|
187 |
{
|
|
188 |
uchar used_uneven_bits= 0; |
|
189 |
if (key_part->null_bit) |
|
190 |
{
|
|
191 |
if (*from_key++) |
|
192 |
to_record[key_part->null_offset]|= key_part->null_bit; |
|
193 |
else
|
|
194 |
to_record[key_part->null_offset]&= ~key_part->null_bit; |
|
195 |
key_length--; |
|
196 |
}
|
|
197 |
if (key_part->type == HA_KEYTYPE_BIT) |
|
198 |
{
|
|
199 |
Field_bit *field= (Field_bit *) (key_part->field); |
|
200 |
if (field->bit_len) |
|
201 |
{
|
|
202 |
uchar bits= *(from_key + key_part->length - |
|
203 |
field->pack_length_in_rec() - 1); |
|
204 |
set_rec_bits(bits, to_record + key_part->null_offset + |
|
205 |
(key_part->null_bit == 128), |
|
206 |
field->bit_ofs, field->bit_len); |
|
207 |
/* we have now used the byte with 'uneven' bits */
|
|
208 |
used_uneven_bits= 1; |
|
209 |
}
|
|
210 |
}
|
|
211 |
if (key_part->key_part_flag & HA_BLOB_PART) |
|
212 |
{
|
|
213 |
/*
|
|
214 |
This in fact never happens, as we have only partial BLOB
|
|
215 |
keys yet anyway, so it's difficult to find any sence to
|
|
216 |
restore the part of a record.
|
|
217 |
Maybe this branch is to be removed, but now we
|
|
218 |
have to ignore GCov compaining.
|
|
219 |
*/
|
|
220 |
uint blob_length= uint2korr(from_key); |
|
221 |
Field_blob *field= (Field_blob*) key_part->field; |
|
222 |
from_key+= HA_KEY_BLOB_LENGTH; |
|
223 |
key_length-= HA_KEY_BLOB_LENGTH; |
|
224 |
field->set_ptr_offset(to_record - field->table->record[0], |
|
225 |
(ulong) blob_length, from_key); |
|
226 |
length= key_part->length; |
|
227 |
}
|
|
228 |
else if (key_part->key_part_flag & HA_VAR_LENGTH_PART) |
|
229 |
{
|
|
230 |
Field *field= key_part->field; |
|
231 |
my_bitmap_map *old_map; |
|
232 |
my_ptrdiff_t ptrdiff= to_record - field->table->record[0]; |
|
233 |
field->move_field_offset(ptrdiff); |
|
234 |
key_length-= HA_KEY_BLOB_LENGTH; |
|
235 |
length= min(key_length, key_part->length); |
|
236 |
old_map= dbug_tmp_use_all_columns(field->table, field->table->write_set); |
|
237 |
field->set_key_image(from_key, length); |
|
238 |
dbug_tmp_restore_column_map(field->table->write_set, old_map); |
|
239 |
from_key+= HA_KEY_BLOB_LENGTH; |
|
240 |
field->move_field_offset(-ptrdiff); |
|
241 |
}
|
|
242 |
else
|
|
243 |
{
|
|
244 |
length= min(key_length, key_part->length); |
|
245 |
/* skip the byte with 'uneven' bits, if used */
|
|
246 |
memcpy(to_record + key_part->offset, from_key + used_uneven_bits |
|
247 |
, (size_t) length - used_uneven_bits); |
|
248 |
}
|
|
249 |
from_key+= length; |
|
250 |
key_length-= length; |
|
251 |
}
|
|
252 |
}
|
|
253 |
||
254 |
||
255 |
/**
|
|
256 |
Compare if a key has changed.
|
|
257 |
||
258 |
@param table TABLE
|
|
259 |
@param key key to compare to row
|
|
260 |
@param idx Index used
|
|
261 |
@param key_length Length of key
|
|
262 |
||
263 |
@note
|
|
264 |
In theory we could just call field->cmp() for all field types,
|
|
265 |
but as we are only interested if a key has changed (not if the key is
|
|
266 |
larger or smaller than the previous value) we can do things a bit
|
|
267 |
faster by using memcmp() instead.
|
|
268 |
||
269 |
@retval
|
|
270 |
0 If key is equal
|
|
271 |
@retval
|
|
272 |
1 Key has changed
|
|
273 |
*/
|
|
274 |
||
275 |
bool key_cmp_if_same(TABLE *table,const uchar *key,uint idx,uint key_length) |
|
276 |
{
|
|
277 |
uint store_length; |
|
278 |
KEY_PART_INFO *key_part; |
|
279 |
const uchar *key_end= key + key_length;; |
|
280 |
||
281 |
for (key_part=table->key_info[idx].key_part; |
|
282 |
key < key_end ; |
|
283 |
key_part++, key+= store_length) |
|
284 |
{
|
|
285 |
uint length; |
|
286 |
store_length= key_part->store_length; |
|
287 |
||
288 |
if (key_part->null_bit) |
|
289 |
{
|
|
290 |
if (*key != test(table->record[0][key_part->null_offset] & |
|
291 |
key_part->null_bit)) |
|
292 |
return 1; |
|
293 |
if (*key) |
|
294 |
continue; |
|
295 |
key++; |
|
296 |
store_length--; |
|
297 |
}
|
|
298 |
if (key_part->key_part_flag & (HA_BLOB_PART | HA_VAR_LENGTH_PART | |
|
299 |
HA_BIT_PART)) |
|
300 |
{
|
|
301 |
if (key_part->field->key_cmp(key, key_part->length)) |
|
302 |
return 1; |
|
303 |
continue; |
|
304 |
}
|
|
305 |
length= min((uint) (key_end-key), store_length); |
|
306 |
if (!(key_part->key_type & (FIELDFLAG_NUMBER+FIELDFLAG_BINARY+ |
|
307 |
FIELDFLAG_PACK))) |
|
308 |
{
|
|
309 |
CHARSET_INFO *cs= key_part->field->charset(); |
|
310 |
uint char_length= key_part->length / cs->mbmaxlen; |
|
311 |
const uchar *pos= table->record[0] + key_part->offset; |
|
312 |
if (length > char_length) |
|
313 |
{
|
|
314 |
char_length= my_charpos(cs, pos, pos + length, char_length); |
|
315 |
set_if_smaller(char_length, length); |
|
316 |
}
|
|
317 |
if (cs->coll->strnncollsp(cs, |
|
318 |
(const uchar*) key, length, |
|
319 |
(const uchar*) pos, char_length, 0)) |
|
320 |
return 1; |
|
321 |
continue; |
|
322 |
}
|
|
323 |
if (memcmp(key,table->record[0]+key_part->offset,length)) |
|
324 |
return 1; |
|
325 |
}
|
|
326 |
return 0; |
|
327 |
}
|
|
328 |
||
329 |
/*
|
|
330 |
unpack key-fields from record to some buffer.
|
|
331 |
||
332 |
This is used mainly to get a good error message. We temporary
|
|
333 |
change the column bitmap so that all columns are readable.
|
|
334 |
||
335 |
@param
|
|
336 |
to Store value here in an easy to read form
|
|
337 |
@param
|
|
338 |
table Table to use
|
|
339 |
@param
|
|
340 |
idx Key number
|
|
341 |
*/
|
|
342 |
||
343 |
void key_unpack(String *to,TABLE *table,uint idx) |
|
344 |
{
|
|
345 |
KEY_PART_INFO *key_part,*key_part_end; |
|
346 |
Field *field; |
|
347 |
String tmp; |
|
348 |
my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); |
|
349 |
DBUG_ENTER("key_unpack"); |
|
350 |
||
351 |
to->length(0); |
|
352 |
for (key_part=table->key_info[idx].key_part,key_part_end=key_part+ |
|
353 |
table->key_info[idx].key_parts ; |
|
354 |
key_part < key_part_end; |
|
355 |
key_part++) |
|
356 |
{
|
|
357 |
if (to->length()) |
|
358 |
to->append('-'); |
|
359 |
if (key_part->null_bit) |
|
360 |
{
|
|
361 |
if (table->record[0][key_part->null_offset] & key_part->null_bit) |
|
362 |
{
|
|
363 |
to->append(STRING_WITH_LEN("NULL")); |
|
364 |
continue; |
|
365 |
}
|
|
366 |
}
|
|
367 |
if ((field=key_part->field)) |
|
368 |
{
|
|
369 |
CHARSET_INFO *cs= field->charset(); |
|
370 |
field->val_str(&tmp); |
|
371 |
if (cs->mbmaxlen > 1 && |
|
372 |
table->field[key_part->fieldnr - 1]->field_length != |
|
373 |
key_part->length) |
|
374 |
{
|
|
375 |
/*
|
|
376 |
Prefix key, multi-byte charset.
|
|
377 |
For the columns of type CHAR(N), the above val_str()
|
|
378 |
call will return exactly "key_part->length" bytes,
|
|
379 |
which can break a multi-byte characters in the middle.
|
|
380 |
Align, returning not more than "char_length" characters.
|
|
381 |
*/
|
|
382 |
uint charpos, char_length= key_part->length / cs->mbmaxlen; |
|
383 |
if ((charpos= my_charpos(cs, tmp.ptr(), |
|
384 |
tmp.ptr() + tmp.length(), |
|
385 |
char_length)) < key_part->length) |
|
386 |
tmp.length(charpos); |
|
387 |
}
|
|
388 |
||
389 |
if (key_part->length < field->pack_length()) |
|
390 |
tmp.length(min(tmp.length(),key_part->length)); |
|
391 |
to->append(tmp); |
|
392 |
}
|
|
393 |
else
|
|
394 |
to->append(STRING_WITH_LEN("???")); |
|
395 |
}
|
|
396 |
dbug_tmp_restore_column_map(table->read_set, old_map); |
|
397 |
DBUG_VOID_RETURN; |
|
398 |
}
|
|
399 |
||
400 |
||
401 |
/*
|
|
402 |
Check if key uses field that is marked in passed field bitmap.
|
|
403 |
||
404 |
SYNOPSIS
|
|
405 |
is_key_used()
|
|
406 |
table TABLE object with which keys and fields are associated.
|
|
407 |
idx Key to be checked.
|
|
408 |
fields Bitmap of fields to be checked.
|
|
409 |
||
410 |
NOTE
|
|
411 |
This function uses TABLE::tmp_set bitmap so the caller should care
|
|
412 |
about saving/restoring its state if it also uses this bitmap.
|
|
413 |
||
414 |
RETURN VALUE
|
|
415 |
TRUE Key uses field from bitmap
|
|
416 |
FALSE Otherwise
|
|
417 |
*/
|
|
418 |
||
419 |
bool is_key_used(TABLE *table, uint idx, const MY_BITMAP *fields) |
|
420 |
{
|
|
421 |
bitmap_clear_all(&table->tmp_set); |
|
422 |
table->mark_columns_used_by_index_no_reset(idx, &table->tmp_set); |
|
423 |
if (bitmap_is_overlapping(&table->tmp_set, fields)) |
|
424 |
return 1; |
|
425 |
||
426 |
/*
|
|
427 |
If table handler has primary key as part of the index, check that primary
|
|
428 |
key is not updated
|
|
429 |
*/
|
|
430 |
if (idx != table->s->primary_key && table->s->primary_key < MAX_KEY && |
|
431 |
(table->file->ha_table_flags() & HA_PRIMARY_KEY_IN_READ_INDEX)) |
|
432 |
return is_key_used(table, table->s->primary_key, fields); |
|
433 |
return 0; |
|
434 |
}
|
|
435 |
||
436 |
||
437 |
/**
|
|
438 |
Compare key in row to a given key.
|
|
439 |
||
440 |
@param key_part Key part handler
|
|
441 |
@param key Key to compare to value in table->record[0]
|
|
442 |
@param key_length length of 'key'
|
|
443 |
||
444 |
@return
|
|
445 |
The return value is SIGN(key_in_row - range_key):
|
|
446 |
- 0 Key is equal to range or 'range' == 0 (no range)
|
|
447 |
- -1 Key is less than range
|
|
448 |
- 1 Key is larger than range
|
|
449 |
*/
|
|
450 |
||
451 |
int key_cmp(KEY_PART_INFO *key_part, const uchar *key, uint key_length) |
|
452 |
{
|
|
453 |
uint store_length; |
|
454 |
||
455 |
for (const uchar *end=key + key_length; |
|
456 |
key < end; |
|
457 |
key+= store_length, key_part++) |
|
458 |
{
|
|
459 |
int cmp; |
|
460 |
store_length= key_part->store_length; |
|
461 |
if (key_part->null_bit) |
|
462 |
{
|
|
463 |
/* This key part allows null values; NULL is lower than everything */
|
|
464 |
register bool field_is_null= key_part->field->is_null(); |
|
465 |
if (*key) // If range key is null |
|
466 |
{
|
|
467 |
/* the range is expecting a null value */
|
|
468 |
if (!field_is_null) |
|
469 |
return 1; // Found key is > range |
|
470 |
/* null -- exact match, go to next key part */
|
|
471 |
continue; |
|
472 |
}
|
|
473 |
else if (field_is_null) |
|
474 |
return -1; // NULL is less than any value |
|
475 |
key++; // Skip null byte |
|
476 |
store_length--; |
|
477 |
}
|
|
478 |
if ((cmp=key_part->field->key_cmp(key, key_part->length)) < 0) |
|
479 |
return -1; |
|
480 |
if (cmp > 0) |
|
481 |
return 1; |
|
482 |
}
|
|
483 |
return 0; // Keys are equal |
|
484 |
}
|
|
485 |
||
486 |
||
487 |
/*
|
|
488 |
Compare two records in index order
|
|
489 |
SYNOPSIS
|
|
490 |
key_rec_cmp()
|
|
491 |
key Index information
|
|
492 |
rec0 Pointer to table->record[0]
|
|
493 |
first_rec Pointer to record compare with
|
|
494 |
second_rec Pointer to record compare against first_rec
|
|
495 |
||
496 |
DESCRIPTION
|
|
497 |
This method is set-up such that it can be called directly from the
|
|
498 |
priority queue and it is attempted to be optimised as much as possible
|
|
499 |
since this will be called O(N * log N) times while performing a merge
|
|
500 |
sort in various places in the code.
|
|
501 |
||
502 |
We retrieve the pointer to table->record[0] using the fact that key_parts
|
|
503 |
have an offset making it possible to calculate the start of the record.
|
|
504 |
We need to get the diff to the compared record since none of the records
|
|
505 |
being compared are stored in table->record[0].
|
|
506 |
||
507 |
We first check for NULL values, if there are no NULL values we use
|
|
508 |
a compare method that gets two field pointers and a max length
|
|
509 |
and return the result of the comparison.
|
|
510 |
*/
|
|
511 |
||
512 |
int key_rec_cmp(void *key, uchar *first_rec, uchar *second_rec) |
|
513 |
{
|
|
514 |
KEY *key_info= (KEY*)key; |
|
515 |
uint key_parts= key_info->key_parts, i= 0; |
|
516 |
KEY_PART_INFO *key_part= key_info->key_part; |
|
517 |
uchar *rec0= key_part->field->ptr - key_part->offset; |
|
518 |
my_ptrdiff_t first_diff= first_rec - rec0, sec_diff= second_rec - rec0; |
|
519 |
int result= 0; |
|
520 |
DBUG_ENTER("key_rec_cmp"); |
|
521 |
||
522 |
do
|
|
523 |
{
|
|
524 |
Field *field= key_part->field; |
|
525 |
||
526 |
if (key_part->null_bit) |
|
527 |
{
|
|
528 |
/* The key_part can contain NULL values */
|
|
529 |
bool first_is_null= field->is_null_in_record_with_offset(first_diff); |
|
530 |
bool sec_is_null= field->is_null_in_record_with_offset(sec_diff); |
|
531 |
/*
|
|
532 |
NULL is smaller then everything so if first is NULL and the other
|
|
533 |
not then we know that we should return -1 and for the opposite
|
|
534 |
we should return +1. If both are NULL then we call it equality
|
|
535 |
although it is a strange form of equality, we have equally little
|
|
536 |
information of the real value.
|
|
537 |
*/
|
|
538 |
if (!first_is_null) |
|
539 |
{
|
|
540 |
if (!sec_is_null) |
|
541 |
; /* Fall through, no NULL fields */ |
|
542 |
else
|
|
543 |
{
|
|
544 |
DBUG_RETURN(+1); |
|
545 |
}
|
|
546 |
}
|
|
547 |
else if (!sec_is_null) |
|
548 |
{
|
|
549 |
DBUG_RETURN(-1); |
|
550 |
}
|
|
551 |
else
|
|
552 |
goto next_loop; /* Both were NULL */ |
|
553 |
}
|
|
554 |
/*
|
|
555 |
No null values in the fields
|
|
556 |
We use the virtual method cmp_max with a max length parameter.
|
|
557 |
For most field types this translates into a cmp without
|
|
558 |
max length. The exceptions are the BLOB and VARCHAR field types
|
|
559 |
that take the max length into account.
|
|
560 |
*/
|
|
561 |
result= field->cmp_max(field->ptr+first_diff, field->ptr+sec_diff, |
|
562 |
key_part->length); |
|
563 |
next_loop: |
|
564 |
key_part++; |
|
565 |
} while (!result && ++i < key_parts); |
|
566 |
DBUG_RETURN(result); |
|
567 |
}
|