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 |
/**
|
|
18 |
@file
|
|
19 |
||
20 |
@brief
|
|
21 |
Sorts a database
|
|
22 |
*/
|
|
23 |
||
243.1.17
by Jay Pipes
FINAL PHASE removal of mysql_priv.h (Bye, bye my friend.) |
24 |
#include <drizzled/server_includes.h> |
1
by brian
clean slate |
25 |
#include "sql_sort.h" |
202.3.6
by Monty Taylor
First pass at gettexizing the error messages. |
26 |
#include <drizzled/drizzled_error_messages.h> |
1
by brian
clean slate |
27 |
|
28 |
/// How to write record_ref.
|
|
29 |
#define WRITE_REF(file,from) \
|
|
30 |
if (my_b_write((file),(uchar*) (from),param->ref_length)) \
|
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
31 |
return(1);
|
1
by brian
clean slate |
32 |
|
33 |
/* functions defined in this file */
|
|
34 |
||
35 |
static char **make_char_array(char **old_pos, register uint fields, |
|
36 |
uint length, myf my_flag); |
|
37 |
static uchar *read_buffpek_from_file(IO_CACHE *buffer_file, uint count, |
|
38 |
uchar *buf); |
|
39 |
static ha_rows find_all_keys(SORTPARAM *param,SQL_SELECT *select, |
|
40 |
uchar * *sort_keys, IO_CACHE *buffer_file, |
|
41 |
IO_CACHE *tempfile,IO_CACHE *indexfile); |
|
42 |
static int write_keys(SORTPARAM *param,uchar * *sort_keys, |
|
43 |
uint count, IO_CACHE *buffer_file, IO_CACHE *tempfile); |
|
44 |
static void make_sortkey(SORTPARAM *param,uchar *to, uchar *ref_pos); |
|
45 |
static void register_used_fields(SORTPARAM *param); |
|
46 |
static int merge_index(SORTPARAM *param,uchar *sort_buffer, |
|
47 |
BUFFPEK *buffpek, |
|
48 |
uint maxbuffer,IO_CACHE *tempfile, |
|
49 |
IO_CACHE *outfile); |
|
50 |
static bool save_index(SORTPARAM *param,uchar **sort_keys, uint count, |
|
327.2.3
by Brian Aker
Refactoring of class Table |
51 |
filesort_info_st *table_sort); |
308
by Brian Aker
ulong conversion |
52 |
static uint suffix_length(uint32_t string_length); |
1
by brian
clean slate |
53 |
static uint sortlength(THD *thd, SORT_FIELD *sortorder, uint s_length, |
54 |
bool *multi_byte_charset); |
|
55 |
static SORT_ADDON_FIELD *get_addon_fields(THD *thd, Field **ptabfield, |
|
56 |
uint sortlength, uint *plength); |
|
57 |
static void unpack_addon_fields(struct st_sort_addon_field *addon_field, |
|
58 |
uchar *buff); |
|
59 |
/**
|
|
60 |
Sort a table.
|
|
61 |
Creates a set of pointers that can be used to read the rows
|
|
62 |
in sorted order. This should be done with the functions
|
|
63 |
in records.cc.
|
|
64 |
||
65 |
Before calling filesort, one must have done
|
|
66 |
table->file->info(HA_STATUS_VARIABLE)
|
|
67 |
||
68 |
The result set is stored in table->io_cache or
|
|
69 |
table->record_pointers.
|
|
70 |
||
71 |
@param thd Current thread
|
|
72 |
@param table Table to sort
|
|
73 |
@param sortorder How to sort the table
|
|
74 |
@param s_length Number of elements in sortorder
|
|
75 |
@param select condition to apply to the rows
|
|
76 |
@param max_rows Return only this many rows
|
|
77 |
@param sort_positions Set to 1 if we want to force sorting by position
|
|
327.1.5
by Brian Aker
Refactor around classes. TABLE_LIST has been factored out of table.h |
78 |
(Needed by UPDATE/INSERT or ALTER Table)
|
1
by brian
clean slate |
79 |
@param examined_rows Store number of examined rows here
|
80 |
||
81 |
@todo
|
|
82 |
check why we do this (param.keys--)
|
|
83 |
@note
|
|
84 |
If we sort by position (like if sort_positions is 1) filesort() will
|
|
85 |
call table->prepare_for_position().
|
|
86 |
||
87 |
@retval
|
|
88 |
HA_POS_ERROR Error
|
|
89 |
@retval
|
|
90 |
\# Number of rows
|
|
91 |
@retval
|
|
92 |
examined_rows will be set to number of examined rows
|
|
93 |
*/
|
|
94 |
||
327.1.5
by Brian Aker
Refactor around classes. TABLE_LIST has been factored out of table.h |
95 |
ha_rows filesort(THD *thd, Table *table, SORT_FIELD *sortorder, uint s_length, |
1
by brian
clean slate |
96 |
SQL_SELECT *select, ha_rows max_rows, |
97 |
bool sort_positions, ha_rows *examined_rows) |
|
98 |
{
|
|
99 |
int error; |
|
308
by Brian Aker
ulong conversion |
100 |
uint32_t memavl, min_sort_memory; |
1
by brian
clean slate |
101 |
uint maxbuffer; |
102 |
BUFFPEK *buffpek; |
|
103 |
ha_rows records= HA_POS_ERROR; |
|
104 |
uchar **sort_keys= 0; |
|
105 |
IO_CACHE tempfile, buffpek_pointers, *selected_records_file, *outfile; |
|
106 |
SORTPARAM param; |
|
107 |
bool multi_byte_charset; |
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
108 |
|
327.2.3
by Brian Aker
Refactoring of class Table |
109 |
filesort_info_st table_sort; |
327.2.4
by Brian Aker
Refactoring table.h |
110 |
TableList *tab= table->pos_in_table_list; |
1
by brian
clean slate |
111 |
Item_subselect *subselect= tab ? tab->containing_subselect() : 0; |
112 |
||
319.1.1
by Grant Limberg
renamed all instances of MYSQL_ to DRIZZLE_ |
113 |
DRIZZLE_FILESORT_START(); |
1
by brian
clean slate |
114 |
|
115 |
/*
|
|
116 |
Release InnoDB's adaptive hash index latch (if holding) before
|
|
117 |
running a sort.
|
|
118 |
*/
|
|
119 |
ha_release_temporary_latches(thd); |
|
120 |
||
121 |
/*
|
|
122 |
Don't use table->sort in filesort as it is also used by
|
|
123 |
QUICK_INDEX_MERGE_SELECT. Work with a copy and put it back at the end
|
|
124 |
when index_merge select has finished with it.
|
|
125 |
*/
|
|
327.2.3
by Brian Aker
Refactoring of class Table |
126 |
memcpy(&table_sort, &table->sort, sizeof(filesort_info_st)); |
1
by brian
clean slate |
127 |
table->sort.io_cache= NULL; |
128 |
||
129 |
outfile= table_sort.io_cache; |
|
130 |
my_b_clear(&tempfile); |
|
131 |
my_b_clear(&buffpek_pointers); |
|
132 |
buffpek=0; |
|
133 |
error= 1; |
|
212.6.6
by Mats Kindahl
Removing redundant use of casts in drizzled/ for memcmp(), memcpy(), memset(), and memmove(). |
134 |
memset(¶m, 0, sizeof(param)); |
1
by brian
clean slate |
135 |
param.sort_length= sortlength(thd, sortorder, s_length, &multi_byte_charset); |
136 |
param.ref_length= table->file->ref_length; |
|
137 |
param.addon_field= 0; |
|
138 |
param.addon_length= 0; |
|
137
by Brian Aker
Removed dead FT bits. Small refactoring in sql_plugin.cc |
139 |
if (!(table->file->ha_table_flags() & HA_FAST_KEY_READ) && !sort_positions) |
1
by brian
clean slate |
140 |
{
|
141 |
/*
|
|
142 |
Get the descriptors of all fields whose values are appended
|
|
143 |
to sorted fields and get its total length in param.spack_length.
|
|
144 |
*/
|
|
145 |
param.addon_field= get_addon_fields(thd, table->field, |
|
146 |
param.sort_length, |
|
147 |
¶m.addon_length); |
|
148 |
}
|
|
149 |
||
150 |
table_sort.addon_buf= 0; |
|
151 |
table_sort.addon_length= param.addon_length; |
|
152 |
table_sort.addon_field= param.addon_field; |
|
153 |
table_sort.unpack= unpack_addon_fields; |
|
154 |
if (param.addon_field) |
|
155 |
{
|
|
156 |
param.res_length= param.addon_length; |
|
157 |
if (!(table_sort.addon_buf= (uchar *) my_malloc(param.addon_length, |
|
158 |
MYF(MY_WME)))) |
|
159 |
goto err; |
|
160 |
}
|
|
161 |
else
|
|
162 |
{
|
|
163 |
param.res_length= param.ref_length; |
|
164 |
/*
|
|
165 |
The reference to the record is considered
|
|
166 |
as an additional sorted field
|
|
167 |
*/
|
|
168 |
param.sort_length+= param.ref_length; |
|
169 |
}
|
|
170 |
param.rec_length= param.sort_length+param.addon_length; |
|
171 |
param.max_rows= max_rows; |
|
172 |
||
173 |
if (select && select->quick) |
|
174 |
{
|
|
175 |
status_var_increment(thd->status_var.filesort_range_count); |
|
176 |
}
|
|
177 |
else
|
|
178 |
{
|
|
179 |
status_var_increment(thd->status_var.filesort_scan_count); |
|
180 |
}
|
|
181 |
#ifdef CAN_TRUST_RANGE
|
|
182 |
if (select && select->quick && select->quick->records > 0L) |
|
183 |
{
|
|
184 |
records=min((ha_rows) (select->quick->records*2+EXTRA_RECORDS*2), |
|
185 |
table->file->stats.records)+EXTRA_RECORDS; |
|
186 |
selected_records_file=0; |
|
187 |
}
|
|
188 |
else
|
|
189 |
#endif
|
|
190 |
{
|
|
191 |
records= table->file->estimate_rows_upper_bound(); |
|
192 |
/*
|
|
193 |
If number of records is not known, use as much of sort buffer
|
|
194 |
as possible.
|
|
195 |
*/
|
|
196 |
if (records == HA_POS_ERROR) |
|
197 |
records--; // we use 'records+1' below. |
|
198 |
selected_records_file= 0; |
|
199 |
}
|
|
200 |
||
201 |
if (multi_byte_charset && |
|
202 |
!(param.tmp_buffer= (char*) my_malloc(param.sort_length,MYF(MY_WME)))) |
|
203 |
goto err; |
|
204 |
||
205 |
memavl= thd->variables.sortbuff_size; |
|
287.3.8
by Monty Taylor
Oy. Replaced max and min macros with std::max and std::min so that we get |
206 |
min_sort_memory= max((uint)MIN_SORT_MEMORY, param.sort_length*MERGEBUFF2); |
1
by brian
clean slate |
207 |
while (memavl >= min_sort_memory) |
208 |
{
|
|
308
by Brian Aker
ulong conversion |
209 |
uint32_t old_memavl; |
210 |
uint32_t keys= memavl/(param.rec_length+sizeof(char*)); |
|
1
by brian
clean slate |
211 |
param.keys=(uint) min(records+1, keys); |
212 |
if ((table_sort.sort_keys= |
|
213 |
(uchar **) make_char_array((char **) table_sort.sort_keys, |
|
214 |
param.keys, param.rec_length, MYF(0)))) |
|
215 |
break; |
|
216 |
old_memavl=memavl; |
|
217 |
if ((memavl=memavl/4*3) < min_sort_memory && old_memavl > min_sort_memory) |
|
218 |
memavl= min_sort_memory; |
|
219 |
}
|
|
220 |
sort_keys= table_sort.sort_keys; |
|
221 |
if (memavl < min_sort_memory) |
|
222 |
{
|
|
223 |
my_error(ER_OUT_OF_SORTMEMORY,MYF(ME_ERROR+ME_WAITTANG)); |
|
224 |
goto err; |
|
225 |
}
|
|
226 |
if (open_cached_file(&buffpek_pointers,mysql_tmpdir,TEMP_PREFIX, |
|
227 |
DISK_BUFFER_SIZE, MYF(MY_WME))) |
|
228 |
goto err; |
|
229 |
||
230 |
param.keys--; /* TODO: check why we do this */ |
|
231 |
param.sort_form= table; |
|
232 |
param.end=(param.local_sortorder=sortorder)+s_length; |
|
233 |
if ((records=find_all_keys(¶m,select,sort_keys, &buffpek_pointers, |
|
234 |
&tempfile, selected_records_file)) == |
|
235 |
HA_POS_ERROR) |
|
236 |
goto err; |
|
237 |
maxbuffer= (uint) (my_b_tell(&buffpek_pointers)/sizeof(*buffpek)); |
|
238 |
||
239 |
if (maxbuffer == 0) // The whole set is in memory |
|
240 |
{
|
|
241 |
if (save_index(¶m,sort_keys,(uint) records, &table_sort)) |
|
242 |
goto err; |
|
243 |
}
|
|
244 |
else
|
|
245 |
{
|
|
246 |
if (table_sort.buffpek && table_sort.buffpek_len < maxbuffer) |
|
247 |
{
|
|
248 |
x_free(table_sort.buffpek); |
|
249 |
table_sort.buffpek= 0; |
|
250 |
}
|
|
251 |
if (!(table_sort.buffpek= |
|
252 |
(uchar *) read_buffpek_from_file(&buffpek_pointers, maxbuffer, |
|
253 |
table_sort.buffpek))) |
|
254 |
goto err; |
|
255 |
buffpek= (BUFFPEK *) table_sort.buffpek; |
|
256 |
table_sort.buffpek_len= maxbuffer; |
|
257 |
close_cached_file(&buffpek_pointers); |
|
258 |
/* Open cached file if it isn't open */
|
|
259 |
if (! my_b_inited(outfile) && |
|
260 |
open_cached_file(outfile,mysql_tmpdir,TEMP_PREFIX,READ_RECORD_BUFFER, |
|
261 |
MYF(MY_WME))) |
|
262 |
goto err; |
|
263 |
if (reinit_io_cache(outfile,WRITE_CACHE,0L,0,0)) |
|
264 |
goto err; |
|
265 |
||
266 |
/*
|
|
267 |
Use also the space previously used by string pointers in sort_buffer
|
|
268 |
for temporary key storage.
|
|
269 |
*/
|
|
270 |
param.keys=((param.keys*(param.rec_length+sizeof(char*))) / |
|
271 |
param.rec_length-1); |
|
272 |
maxbuffer--; // Offset from 0 |
|
273 |
if (merge_many_buff(¶m,(uchar*) sort_keys,buffpek,&maxbuffer, |
|
274 |
&tempfile)) |
|
275 |
goto err; |
|
276 |
if (flush_io_cache(&tempfile) || |
|
277 |
reinit_io_cache(&tempfile,READ_CACHE,0L,0,0)) |
|
278 |
goto err; |
|
279 |
if (merge_index(¶m,(uchar*) sort_keys,buffpek,maxbuffer,&tempfile, |
|
280 |
outfile)) |
|
281 |
goto err; |
|
282 |
}
|
|
283 |
if (records > param.max_rows) |
|
284 |
records=param.max_rows; |
|
285 |
error =0; |
|
286 |
||
287 |
err: |
|
288 |
if (param.tmp_buffer) |
|
289 |
x_free(param.tmp_buffer); |
|
290 |
if (!subselect || !subselect->is_uncacheable()) |
|
291 |
{
|
|
292 |
x_free((uchar*) sort_keys); |
|
293 |
table_sort.sort_keys= 0; |
|
294 |
x_free((uchar*) buffpek); |
|
295 |
table_sort.buffpek= 0; |
|
296 |
table_sort.buffpek_len= 0; |
|
297 |
}
|
|
298 |
close_cached_file(&tempfile); |
|
299 |
close_cached_file(&buffpek_pointers); |
|
300 |
if (my_b_inited(outfile)) |
|
301 |
{
|
|
302 |
if (flush_io_cache(outfile)) |
|
303 |
error=1; |
|
304 |
{
|
|
305 |
my_off_t save_pos=outfile->pos_in_file; |
|
306 |
/* For following reads */
|
|
307 |
if (reinit_io_cache(outfile,READ_CACHE,0L,0,0)) |
|
308 |
error=1; |
|
309 |
outfile->end_of_file=save_pos; |
|
310 |
}
|
|
311 |
}
|
|
312 |
if (error) |
|
313 |
my_message(ER_FILSORT_ABORT, ER(ER_FILSORT_ABORT), |
|
314 |
MYF(ME_ERROR+ME_WAITTANG)); |
|
315 |
else
|
|
316 |
statistic_add(thd->status_var.filesort_rows, |
|
308
by Brian Aker
ulong conversion |
317 |
(uint32_t) records, &LOCK_status); |
1
by brian
clean slate |
318 |
*examined_rows= param.examined_rows; |
327.2.3
by Brian Aker
Refactoring of class Table |
319 |
memcpy(&table->sort, &table_sort, sizeof(filesort_info_st)); |
319.1.1
by Grant Limberg
renamed all instances of MYSQL_ to DRIZZLE_ |
320 |
DRIZZLE_FILESORT_END(); |
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
321 |
return(error ? HA_POS_ERROR : records); |
1
by brian
clean slate |
322 |
} /* filesort */ |
323 |
||
324 |
||
327.1.5
by Brian Aker
Refactor around classes. TABLE_LIST has been factored out of table.h |
325 |
void filesort_free_buffers(Table *table, bool full) |
1
by brian
clean slate |
326 |
{
|
327 |
if (table->sort.record_pointers) |
|
328 |
{
|
|
329 |
my_free((uchar*) table->sort.record_pointers,MYF(0)); |
|
330 |
table->sort.record_pointers=0; |
|
331 |
}
|
|
332 |
if (full) |
|
333 |
{
|
|
334 |
if (table->sort.sort_keys ) |
|
335 |
{
|
|
336 |
x_free((uchar*) table->sort.sort_keys); |
|
337 |
table->sort.sort_keys= 0; |
|
338 |
}
|
|
339 |
if (table->sort.buffpek) |
|
340 |
{
|
|
341 |
x_free((uchar*) table->sort.buffpek); |
|
342 |
table->sort.buffpek= 0; |
|
343 |
table->sort.buffpek_len= 0; |
|
344 |
}
|
|
345 |
}
|
|
346 |
if (table->sort.addon_buf) |
|
347 |
{
|
|
348 |
my_free((char *) table->sort.addon_buf, MYF(0)); |
|
349 |
my_free((char *) table->sort.addon_field, MYF(MY_ALLOW_ZERO_PTR)); |
|
350 |
table->sort.addon_buf=0; |
|
351 |
table->sort.addon_field=0; |
|
352 |
}
|
|
353 |
}
|
|
354 |
||
355 |
/** Make a array of string pointers. */
|
|
356 |
||
357 |
static char **make_char_array(char **old_pos, register uint fields, |
|
358 |
uint length, myf my_flag) |
|
359 |
{
|
|
360 |
register char **pos; |
|
361 |
char *char_pos; |
|
362 |
||
363 |
if (old_pos || |
|
364 |
(old_pos= (char**) my_malloc((uint) fields*(length+sizeof(char*)), |
|
365 |
my_flag))) |
|
366 |
{
|
|
367 |
pos=old_pos; char_pos=((char*) (pos+fields)) -length; |
|
368 |
while (fields--) *(pos++) = (char_pos+= length); |
|
369 |
}
|
|
370 |
||
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
371 |
return(old_pos); |
1
by brian
clean slate |
372 |
} /* make_char_array */ |
373 |
||
374 |
||
375 |
/** Read 'count' number of buffer pointers into memory. */
|
|
376 |
||
377 |
static uchar *read_buffpek_from_file(IO_CACHE *buffpek_pointers, uint count, |
|
378 |
uchar *buf) |
|
379 |
{
|
|
308
by Brian Aker
ulong conversion |
380 |
uint32_t length= sizeof(BUFFPEK)*count; |
1
by brian
clean slate |
381 |
uchar *tmp= buf; |
382 |
if (count > UINT_MAX/sizeof(BUFFPEK)) |
|
383 |
return 0; /* sizeof(BUFFPEK)*count will overflow */ |
|
384 |
if (!tmp) |
|
385 |
tmp= (uchar *)my_malloc(length, MYF(MY_WME)); |
|
386 |
if (tmp) |
|
387 |
{
|
|
388 |
if (reinit_io_cache(buffpek_pointers,READ_CACHE,0L,0,0) || |
|
389 |
my_b_read(buffpek_pointers, (uchar*) tmp, length)) |
|
390 |
{
|
|
391 |
my_free((char*) tmp, MYF(0)); |
|
392 |
tmp=0; |
|
393 |
}
|
|
394 |
}
|
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
395 |
return(tmp); |
1
by brian
clean slate |
396 |
}
|
397 |
||
398 |
||
399 |
/**
|
|
400 |
Search after sort_keys and write them into tempfile.
|
|
401 |
All produced sequences are guaranteed to be non-empty.
|
|
402 |
||
403 |
@param param Sorting parameter
|
|
404 |
@param select Use this to get source data
|
|
405 |
@param sort_keys Array of pointers to sort key + addon buffers.
|
|
406 |
@param buffpek_pointers File to write BUFFPEKs describing sorted segments
|
|
407 |
in tempfile.
|
|
408 |
@param tempfile File to write sorted sequences of sortkeys to.
|
|
409 |
@param indexfile If !NULL, use it for source data (contains rowids)
|
|
410 |
||
411 |
@note
|
|
412 |
Basic idea:
|
|
413 |
@verbatim
|
|
414 |
while (get_next_sortkey())
|
|
415 |
{
|
|
416 |
if (no free space in sort_keys buffers)
|
|
417 |
{
|
|
418 |
sort sort_keys buffer;
|
|
419 |
dump sorted sequence to 'tempfile';
|
|
420 |
dump BUFFPEK describing sequence location into 'buffpek_pointers';
|
|
421 |
}
|
|
422 |
put sort key into 'sort_keys';
|
|
423 |
}
|
|
424 |
if (sort_keys has some elements && dumped at least once)
|
|
425 |
sort-dump-dump as above;
|
|
426 |
else
|
|
427 |
don't sort, leave sort_keys array to be sorted by caller.
|
|
428 |
@endverbatim
|
|
429 |
||
430 |
@retval
|
|
431 |
Number of records written on success.
|
|
432 |
@retval
|
|
433 |
HA_POS_ERROR on error.
|
|
434 |
*/
|
|
435 |
||
436 |
static ha_rows find_all_keys(SORTPARAM *param, SQL_SELECT *select, |
|
437 |
uchar **sort_keys, |
|
438 |
IO_CACHE *buffpek_pointers, |
|
439 |
IO_CACHE *tempfile, IO_CACHE *indexfile) |
|
440 |
{
|
|
441 |
int error,flag,quick_select; |
|
442 |
uint idx,indexpos,ref_length; |
|
443 |
uchar *ref_pos,*next_pos,ref_buff[MAX_REFLENGTH]; |
|
444 |
my_off_t record; |
|
327.1.5
by Brian Aker
Refactor around classes. TABLE_LIST has been factored out of table.h |
445 |
Table *sort_form; |
1
by brian
clean slate |
446 |
THD *thd= current_thd; |
447 |
volatile THD::killed_state *killed= &thd->killed; |
|
448 |
handler *file; |
|
449 |
MY_BITMAP *save_read_set, *save_write_set; |
|
450 |
||
451 |
idx=indexpos=0; |
|
452 |
error=quick_select=0; |
|
453 |
sort_form=param->sort_form; |
|
454 |
file=sort_form->file; |
|
455 |
ref_length=param->ref_length; |
|
456 |
ref_pos= ref_buff; |
|
457 |
quick_select=select && select->quick; |
|
458 |
record=0; |
|
459 |
flag= ((!indexfile && file->ha_table_flags() & HA_REC_NOT_IN_SEQ) |
|
460 |
|| quick_select); |
|
461 |
if (indexfile || flag) |
|
462 |
ref_pos= &file->ref[0]; |
|
463 |
next_pos=ref_pos; |
|
464 |
if (! indexfile && ! quick_select) |
|
465 |
{
|
|
466 |
next_pos=(uchar*) 0; /* Find records in sequence */ |
|
467 |
file->ha_rnd_init(1); |
|
468 |
file->extra_opt(HA_EXTRA_CACHE, |
|
469 |
current_thd->variables.read_buff_size); |
|
470 |
}
|
|
471 |
||
472 |
READ_RECORD read_record_info; |
|
473 |
if (quick_select) |
|
474 |
{
|
|
475 |
if (select->quick->reset()) |
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
476 |
return(HA_POS_ERROR); |
1
by brian
clean slate |
477 |
init_read_record(&read_record_info, current_thd, select->quick->head, |
478 |
select, 1, 1); |
|
479 |
}
|
|
480 |
||
481 |
/* Remember original bitmaps */
|
|
482 |
save_read_set= sort_form->read_set; |
|
483 |
save_write_set= sort_form->write_set; |
|
484 |
/* Set up temporary column read map for columns used by sort */
|
|
485 |
bitmap_clear_all(&sort_form->tmp_set); |
|
486 |
/* Temporary set for register_used_fields and register_field_in_read_map */
|
|
487 |
sort_form->read_set= &sort_form->tmp_set; |
|
488 |
register_used_fields(param); |
|
489 |
if (select && select->cond) |
|
490 |
select->cond->walk(&Item::register_field_in_read_map, 1, |
|
491 |
(uchar*) sort_form); |
|
492 |
sort_form->column_bitmaps_set(&sort_form->tmp_set, &sort_form->tmp_set); |
|
493 |
||
494 |
for (;;) |
|
495 |
{
|
|
496 |
if (quick_select) |
|
497 |
{
|
|
498 |
if ((error= read_record_info.read_record(&read_record_info))) |
|
499 |
{
|
|
500 |
error= HA_ERR_END_OF_FILE; |
|
501 |
break; |
|
502 |
}
|
|
503 |
file->position(sort_form->record[0]); |
|
504 |
}
|
|
505 |
else /* Not quick-select */ |
|
506 |
{
|
|
507 |
if (indexfile) |
|
508 |
{
|
|
509 |
if (my_b_read(indexfile,(uchar*) ref_pos,ref_length)) /* purecov: deadcode */ |
|
510 |
{
|
|
511 |
error= my_errno ? my_errno : -1; /* Abort */ |
|
512 |
break; |
|
513 |
}
|
|
514 |
error=file->rnd_pos(sort_form->record[0],next_pos); |
|
515 |
}
|
|
516 |
else
|
|
517 |
{
|
|
518 |
error=file->rnd_next(sort_form->record[0]); |
|
519 |
if (!flag) |
|
520 |
{
|
|
521 |
my_store_ptr(ref_pos,ref_length,record); // Position to row |
|
522 |
record+= sort_form->s->db_record_offset; |
|
523 |
}
|
|
524 |
else if (!error) |
|
525 |
file->position(sort_form->record[0]); |
|
526 |
}
|
|
527 |
if (error && error != HA_ERR_RECORD_DELETED) |
|
528 |
break; |
|
529 |
}
|
|
530 |
||
531 |
if (*killed) |
|
532 |
{
|
|
533 |
if (!indexfile && !quick_select) |
|
534 |
{
|
|
535 |
(void) file->extra(HA_EXTRA_NO_CACHE); |
|
536 |
file->ha_rnd_end(); |
|
537 |
}
|
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
538 |
return(HA_POS_ERROR); /* purecov: inspected */ |
1
by brian
clean slate |
539 |
}
|
540 |
if (error == 0) |
|
541 |
param->examined_rows++; |
|
542 |
if (error == 0 && (!select || select->skip_record() == 0)) |
|
543 |
{
|
|
544 |
if (idx == param->keys) |
|
545 |
{
|
|
546 |
if (write_keys(param,sort_keys,idx,buffpek_pointers,tempfile)) |
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
547 |
return(HA_POS_ERROR); |
1
by brian
clean slate |
548 |
idx=0; |
549 |
indexpos++; |
|
550 |
}
|
|
551 |
make_sortkey(param,sort_keys[idx++],ref_pos); |
|
552 |
}
|
|
553 |
else
|
|
554 |
file->unlock_row(); |
|
555 |
/* It does not make sense to read more keys in case of a fatal error */
|
|
556 |
if (thd->is_error()) |
|
557 |
break; |
|
558 |
}
|
|
559 |
if (quick_select) |
|
560 |
{
|
|
561 |
/*
|
|
562 |
index_merge quick select uses table->sort when retrieving rows, so free
|
|
563 |
resoures it has allocated.
|
|
564 |
*/
|
|
565 |
end_read_record(&read_record_info); |
|
566 |
}
|
|
567 |
else
|
|
568 |
{
|
|
569 |
(void) file->extra(HA_EXTRA_NO_CACHE); /* End cacheing of records */ |
|
570 |
if (!next_pos) |
|
571 |
file->ha_rnd_end(); |
|
572 |
}
|
|
573 |
||
574 |
if (thd->is_error()) |
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
575 |
return(HA_POS_ERROR); |
1
by brian
clean slate |
576 |
|
577 |
/* Signal we should use orignal column read and write maps */
|
|
578 |
sort_form->column_bitmaps_set(save_read_set, save_write_set); |
|
579 |
||
580 |
if (error != HA_ERR_END_OF_FILE) |
|
581 |
{
|
|
582 |
file->print_error(error,MYF(ME_ERROR | ME_WAITTANG)); /* purecov: inspected */ |
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
583 |
return(HA_POS_ERROR); /* purecov: inspected */ |
1
by brian
clean slate |
584 |
}
|
585 |
if (indexpos && idx && |
|
586 |
write_keys(param,sort_keys,idx,buffpek_pointers,tempfile)) |
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
587 |
return(HA_POS_ERROR); /* purecov: inspected */ |
588 |
return(my_b_inited(tempfile) ? |
|
1
by brian
clean slate |
589 |
(ha_rows) (my_b_tell(tempfile)/param->rec_length) : |
590 |
idx); |
|
591 |
} /* find_all_keys */ |
|
592 |
||
593 |
||
594 |
/**
|
|
595 |
@details
|
|
596 |
Sort the buffer and write:
|
|
597 |
-# the sorted sequence to tempfile
|
|
598 |
-# a BUFFPEK describing the sorted sequence position to buffpek_pointers
|
|
599 |
||
600 |
(was: Skriver en buffert med nycklar till filen)
|
|
601 |
||
602 |
@param param Sort parameters
|
|
603 |
@param sort_keys Array of pointers to keys to sort
|
|
604 |
@param count Number of elements in sort_keys array
|
|
605 |
@param buffpek_pointers One 'BUFFPEK' struct will be written into this file.
|
|
606 |
The BUFFPEK::{file_pos, count} will indicate where
|
|
607 |
the sorted data was stored.
|
|
608 |
@param tempfile The sorted sequence will be written into this file.
|
|
609 |
||
610 |
@retval
|
|
611 |
0 OK
|
|
612 |
@retval
|
|
613 |
1 Error
|
|
614 |
*/
|
|
615 |
||
616 |
static int |
|
617 |
write_keys(SORTPARAM *param, register uchar **sort_keys, uint count, |
|
618 |
IO_CACHE *buffpek_pointers, IO_CACHE *tempfile) |
|
619 |
{
|
|
620 |
size_t sort_length, rec_length; |
|
621 |
uchar **end; |
|
622 |
BUFFPEK buffpek; |
|
623 |
||
624 |
sort_length= param->sort_length; |
|
625 |
rec_length= param->rec_length; |
|
626 |
my_string_ptr_sort((uchar*) sort_keys, (uint) count, sort_length); |
|
627 |
if (!my_b_inited(tempfile) && |
|
628 |
open_cached_file(tempfile, mysql_tmpdir, TEMP_PREFIX, DISK_BUFFER_SIZE, |
|
629 |
MYF(MY_WME))) |
|
630 |
goto err; /* purecov: inspected */ |
|
631 |
/* check we won't have more buffpeks than we can possibly keep in memory */
|
|
151
by Brian Aker
Ulonglong to uint64_t |
632 |
if (my_b_tell(buffpek_pointers) + sizeof(BUFFPEK) > (uint64_t)UINT_MAX) |
1
by brian
clean slate |
633 |
goto err; |
634 |
buffpek.file_pos= my_b_tell(tempfile); |
|
635 |
if ((ha_rows) count > param->max_rows) |
|
636 |
count=(uint) param->max_rows; /* purecov: inspected */ |
|
637 |
buffpek.count=(ha_rows) count; |
|
638 |
for (end=sort_keys+count ; sort_keys != end ; sort_keys++) |
|
639 |
if (my_b_write(tempfile, (uchar*) *sort_keys, (uint) rec_length)) |
|
640 |
goto err; |
|
641 |
if (my_b_write(buffpek_pointers, (uchar*) &buffpek, sizeof(buffpek))) |
|
642 |
goto err; |
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
643 |
return(0); |
1
by brian
clean slate |
644 |
|
645 |
err: |
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
646 |
return(1); |
1
by brian
clean slate |
647 |
} /* write_keys */ |
648 |
||
649 |
||
650 |
/**
|
|
651 |
Store length as suffix in high-byte-first order.
|
|
652 |
*/
|
|
653 |
||
654 |
static inline void store_length(uchar *to, uint length, uint pack_length) |
|
655 |
{
|
|
656 |
switch (pack_length) { |
|
657 |
case 1: |
|
658 |
*to= (uchar) length; |
|
659 |
break; |
|
660 |
case 2: |
|
661 |
mi_int2store(to, length); |
|
662 |
break; |
|
663 |
case 3: |
|
664 |
mi_int3store(to, length); |
|
665 |
break; |
|
666 |
default: |
|
667 |
mi_int4store(to, length); |
|
668 |
break; |
|
669 |
}
|
|
670 |
}
|
|
671 |
||
672 |
||
673 |
/** Make a sort-key from record. */
|
|
674 |
||
675 |
static void make_sortkey(register SORTPARAM *param, |
|
676 |
register uchar *to, uchar *ref_pos) |
|
677 |
{
|
|
678 |
register Field *field; |
|
679 |
register SORT_FIELD *sort_field; |
|
680 |
register uint length; |
|
681 |
||
682 |
for (sort_field=param->local_sortorder ; |
|
683 |
sort_field != param->end ; |
|
684 |
sort_field++) |
|
685 |
{
|
|
686 |
bool maybe_null=0; |
|
687 |
if ((field=sort_field->field)) |
|
688 |
{ // Field |
|
689 |
if (field->maybe_null()) |
|
690 |
{
|
|
691 |
if (field->is_null()) |
|
692 |
{
|
|
693 |
if (sort_field->reverse) |
|
212.6.3
by Mats Kindahl
Removing deprecated functions from code and replacing them with C99 equivalents: |
694 |
memset(to, 255, sort_field->length+1); |
1
by brian
clean slate |
695 |
else
|
212.6.6
by Mats Kindahl
Removing redundant use of casts in drizzled/ for memcmp(), memcpy(), memset(), and memmove(). |
696 |
memset(to, 0, sort_field->length+1); |
1
by brian
clean slate |
697 |
to+= sort_field->length+1; |
698 |
continue; |
|
699 |
}
|
|
700 |
else
|
|
701 |
*to++=1; |
|
702 |
}
|
|
703 |
field->sort_string(to, sort_field->length); |
|
704 |
}
|
|
705 |
else
|
|
706 |
{ // Item |
|
707 |
Item *item=sort_field->item; |
|
708 |
maybe_null= item->maybe_null; |
|
709 |
switch (sort_field->result_type) { |
|
710 |
case STRING_RESULT: |
|
711 |
{
|
|
264.2.6
by Andrey Hristov
Constify the usage of CHARSET_INFO almost to the last place in the code. |
712 |
const CHARSET_INFO * const cs=item->collation.collation; |
1
by brian
clean slate |
713 |
char fill_char= ((cs->state & MY_CS_BINSORT) ? (char) 0 : ' '); |
714 |
int diff; |
|
715 |
uint sort_field_length; |
|
716 |
||
717 |
if (maybe_null) |
|
718 |
*to++=1; |
|
719 |
/* All item->str() to use some extra byte for end null.. */
|
|
720 |
String tmp((char*) to,sort_field->length+4,cs); |
|
721 |
String *res= item->str_result(&tmp); |
|
722 |
if (!res) |
|
723 |
{
|
|
724 |
if (maybe_null) |
|
212.6.6
by Mats Kindahl
Removing redundant use of casts in drizzled/ for memcmp(), memcpy(), memset(), and memmove(). |
725 |
memset(to-1, 0, sort_field->length+1); |
1
by brian
clean slate |
726 |
else
|
727 |
{
|
|
728 |
/* purecov: begin deadcode */
|
|
729 |
/*
|
|
730 |
This should only happen during extreme conditions if we run out
|
|
731 |
of memory or have an item marked not null when it can be null.
|
|
732 |
This code is here mainly to avoid a hard crash in this case.
|
|
733 |
*/
|
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
734 |
assert(0); |
212.6.6
by Mats Kindahl
Removing redundant use of casts in drizzled/ for memcmp(), memcpy(), memset(), and memmove(). |
735 |
memset(to, 0, sort_field->length); // Avoid crash |
1
by brian
clean slate |
736 |
/* purecov: end */
|
737 |
}
|
|
738 |
break; |
|
739 |
}
|
|
740 |
length= res->length(); |
|
741 |
sort_field_length= sort_field->length - sort_field->suffix_length; |
|
742 |
diff=(int) (sort_field_length - length); |
|
743 |
if (diff < 0) |
|
744 |
{
|
|
745 |
diff=0; |
|
746 |
length= sort_field_length; |
|
747 |
}
|
|
748 |
if (sort_field->suffix_length) |
|
749 |
{
|
|
750 |
/* Store length last in result_string */
|
|
751 |
store_length(to + sort_field_length, length, |
|
752 |
sort_field->suffix_length); |
|
753 |
}
|
|
754 |
if (sort_field->need_strxnfrm) |
|
755 |
{
|
|
756 |
char *from=(char*) res->ptr(); |
|
757 |
uint tmp_length; |
|
758 |
if ((uchar*) from == to) |
|
759 |
{
|
|
760 |
set_if_smaller(length,sort_field->length); |
|
761 |
memcpy(param->tmp_buffer,from,length); |
|
762 |
from=param->tmp_buffer; |
|
763 |
}
|
|
764 |
tmp_length= my_strnxfrm(cs,to,sort_field->length, |
|
765 |
(uchar*) from, length); |
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
766 |
assert(tmp_length == sort_field->length); |
1
by brian
clean slate |
767 |
}
|
768 |
else
|
|
769 |
{
|
|
770 |
my_strnxfrm(cs,(uchar*)to,length,(const uchar*)res->ptr(),length); |
|
771 |
cs->cset->fill(cs, (char *)to+length,diff,fill_char); |
|
772 |
}
|
|
773 |
break; |
|
774 |
}
|
|
775 |
case INT_RESULT: |
|
776 |
{
|
|
152
by Brian Aker
longlong replacement |
777 |
int64_t value= item->val_int_result(); |
1
by brian
clean slate |
778 |
if (maybe_null) |
779 |
{
|
|
780 |
*to++=1; /* purecov: inspected */ |
|
781 |
if (item->null_value) |
|
782 |
{
|
|
783 |
if (maybe_null) |
|
212.6.6
by Mats Kindahl
Removing redundant use of casts in drizzled/ for memcmp(), memcpy(), memset(), and memmove(). |
784 |
memset(to-1, 0, sort_field->length+1); |
1
by brian
clean slate |
785 |
else
|
786 |
{
|
|
212.6.6
by Mats Kindahl
Removing redundant use of casts in drizzled/ for memcmp(), memcpy(), memset(), and memmove(). |
787 |
memset(to, 0, sort_field->length); |
1
by brian
clean slate |
788 |
}
|
789 |
break; |
|
790 |
}
|
|
791 |
}
|
|
792 |
#if SIZEOF_LONG_LONG > 4
|
|
793 |
to[7]= (uchar) value; |
|
794 |
to[6]= (uchar) (value >> 8); |
|
795 |
to[5]= (uchar) (value >> 16); |
|
796 |
to[4]= (uchar) (value >> 24); |
|
797 |
to[3]= (uchar) (value >> 32); |
|
798 |
to[2]= (uchar) (value >> 40); |
|
799 |
to[1]= (uchar) (value >> 48); |
|
800 |
if (item->unsigned_flag) /* Fix sign */ |
|
801 |
to[0]= (uchar) (value >> 56); |
|
802 |
else
|
|
803 |
to[0]= (uchar) (value >> 56) ^ 128; /* Reverse signbit */ |
|
804 |
#else
|
|
805 |
to[3]= (uchar) value; |
|
806 |
to[2]= (uchar) (value >> 8); |
|
807 |
to[1]= (uchar) (value >> 16); |
|
808 |
if (item->unsigned_flag) /* Fix sign */ |
|
809 |
to[0]= (uchar) (value >> 24); |
|
810 |
else
|
|
811 |
to[0]= (uchar) (value >> 24) ^ 128; /* Reverse signbit */ |
|
812 |
#endif
|
|
813 |
break; |
|
814 |
}
|
|
815 |
case DECIMAL_RESULT: |
|
816 |
{
|
|
817 |
my_decimal dec_buf, *dec_val= item->val_decimal_result(&dec_buf); |
|
818 |
if (maybe_null) |
|
819 |
{
|
|
820 |
if (item->null_value) |
|
821 |
{
|
|
212.6.6
by Mats Kindahl
Removing redundant use of casts in drizzled/ for memcmp(), memcpy(), memset(), and memmove(). |
822 |
memset(to, 0, sort_field->length+1); |
1
by brian
clean slate |
823 |
to++; |
824 |
break; |
|
825 |
}
|
|
826 |
*to++=1; |
|
827 |
}
|
|
828 |
my_decimal2binary(E_DEC_FATAL_ERROR, dec_val, to, |
|
829 |
item->max_length - (item->decimals ? 1:0), |
|
830 |
item->decimals); |
|
831 |
break; |
|
832 |
}
|
|
833 |
case REAL_RESULT: |
|
834 |
{
|
|
835 |
double value= item->val_result(); |
|
836 |
if (maybe_null) |
|
837 |
{
|
|
838 |
if (item->null_value) |
|
839 |
{
|
|
212.6.6
by Mats Kindahl
Removing redundant use of casts in drizzled/ for memcmp(), memcpy(), memset(), and memmove(). |
840 |
memset(to, 0, sort_field->length+1); |
1
by brian
clean slate |
841 |
to++; |
842 |
break; |
|
843 |
}
|
|
844 |
*to++=1; |
|
845 |
}
|
|
846 |
change_double_for_sort(value,(uchar*) to); |
|
847 |
break; |
|
848 |
}
|
|
849 |
case ROW_RESULT: |
|
850 |
default: |
|
851 |
// This case should never be choosen
|
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
852 |
assert(0); |
1
by brian
clean slate |
853 |
break; |
854 |
}
|
|
855 |
}
|
|
856 |
if (sort_field->reverse) |
|
857 |
{ /* Revers key */ |
|
858 |
if (maybe_null) |
|
859 |
to[-1]= ~to[-1]; |
|
860 |
length=sort_field->length; |
|
861 |
while (length--) |
|
862 |
{
|
|
863 |
*to = (uchar) (~ *to); |
|
864 |
to++; |
|
865 |
}
|
|
866 |
}
|
|
867 |
else
|
|
868 |
to+= sort_field->length; |
|
869 |
}
|
|
870 |
||
871 |
if (param->addon_field) |
|
872 |
{
|
|
873 |
/*
|
|
874 |
Save field values appended to sorted fields.
|
|
875 |
First null bit indicators are appended then field values follow.
|
|
876 |
In this implementation we use fixed layout for field values -
|
|
877 |
the same for all records.
|
|
878 |
*/
|
|
879 |
SORT_ADDON_FIELD *addonf= param->addon_field; |
|
880 |
uchar *nulls= to; |
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
881 |
assert(addonf != 0); |
212.6.6
by Mats Kindahl
Removing redundant use of casts in drizzled/ for memcmp(), memcpy(), memset(), and memmove(). |
882 |
memset(nulls, 0, addonf->offset); |
1
by brian
clean slate |
883 |
to+= addonf->offset; |
884 |
for ( ; (field= addonf->field) ; addonf++) |
|
885 |
{
|
|
886 |
if (addonf->null_bit && field->is_null()) |
|
887 |
{
|
|
888 |
nulls[addonf->null_offset]|= addonf->null_bit; |
|
889 |
#ifdef HAVE_purify
|
|
212.6.1
by Mats Kindahl
Replacing all bzero() calls with memset() calls and removing the bzero.c file. |
890 |
memset(to, 0, addonf->length); |
1
by brian
clean slate |
891 |
#endif
|
892 |
}
|
|
893 |
else
|
|
894 |
{
|
|
895 |
#ifdef HAVE_purify
|
|
896 |
uchar *end= field->pack(to, field->ptr); |
|
897 |
uint length= (uint) ((to + addonf->length) - end); |
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
898 |
assert((int) length >= 0); |
1
by brian
clean slate |
899 |
if (length) |
212.6.1
by Mats Kindahl
Replacing all bzero() calls with memset() calls and removing the bzero.c file. |
900 |
memset(end, 0, length); |
1
by brian
clean slate |
901 |
#else
|
902 |
(void) field->pack(to, field->ptr); |
|
903 |
#endif
|
|
904 |
}
|
|
905 |
to+= addonf->length; |
|
906 |
}
|
|
907 |
}
|
|
908 |
else
|
|
909 |
{
|
|
910 |
/* Save filepos last */
|
|
212.6.6
by Mats Kindahl
Removing redundant use of casts in drizzled/ for memcmp(), memcpy(), memset(), and memmove(). |
911 |
memcpy(to, ref_pos, (size_t) param->ref_length); |
1
by brian
clean slate |
912 |
}
|
913 |
return; |
|
914 |
}
|
|
915 |
||
916 |
||
917 |
/*
|
|
918 |
Register fields used by sorting in the sorted table's read set
|
|
919 |
*/
|
|
920 |
||
921 |
static void register_used_fields(SORTPARAM *param) |
|
922 |
{
|
|
923 |
register SORT_FIELD *sort_field; |
|
327.1.5
by Brian Aker
Refactor around classes. TABLE_LIST has been factored out of table.h |
924 |
Table *table=param->sort_form; |
1
by brian
clean slate |
925 |
MY_BITMAP *bitmap= table->read_set; |
926 |
||
927 |
for (sort_field= param->local_sortorder ; |
|
928 |
sort_field != param->end ; |
|
929 |
sort_field++) |
|
930 |
{
|
|
931 |
Field *field; |
|
932 |
if ((field= sort_field->field)) |
|
933 |
{
|
|
934 |
if (field->table == table) |
|
935 |
bitmap_set_bit(bitmap, field->field_index); |
|
936 |
}
|
|
937 |
else
|
|
938 |
{ // Item |
|
939 |
sort_field->item->walk(&Item::register_field_in_read_map, 1, |
|
940 |
(uchar *) table); |
|
941 |
}
|
|
942 |
}
|
|
943 |
||
944 |
if (param->addon_field) |
|
945 |
{
|
|
946 |
SORT_ADDON_FIELD *addonf= param->addon_field; |
|
947 |
Field *field; |
|
948 |
for ( ; (field= addonf->field) ; addonf++) |
|
949 |
bitmap_set_bit(bitmap, field->field_index); |
|
950 |
}
|
|
951 |
else
|
|
952 |
{
|
|
953 |
/* Save filepos last */
|
|
954 |
table->prepare_for_position(); |
|
955 |
}
|
|
956 |
}
|
|
957 |
||
958 |
||
959 |
static bool save_index(SORTPARAM *param, uchar **sort_keys, uint count, |
|
327.2.3
by Brian Aker
Refactoring of class Table |
960 |
filesort_info_st *table_sort) |
1
by brian
clean slate |
961 |
{
|
962 |
uint offset,res_length; |
|
963 |
uchar *to; |
|
964 |
||
965 |
my_string_ptr_sort((uchar*) sort_keys, (uint) count, param->sort_length); |
|
966 |
res_length= param->res_length; |
|
967 |
offset= param->rec_length-res_length; |
|
968 |
if ((ha_rows) count > param->max_rows) |
|
969 |
count=(uint) param->max_rows; |
|
970 |
if (!(to= table_sort->record_pointers= |
|
971 |
(uchar*) my_malloc(res_length*count, MYF(MY_WME)))) |
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
972 |
return(1); /* purecov: inspected */ |
1
by brian
clean slate |
973 |
for (uchar **end= sort_keys+count ; sort_keys != end ; sort_keys++) |
974 |
{
|
|
975 |
memcpy(to, *sort_keys+offset, res_length); |
|
976 |
to+= res_length; |
|
977 |
}
|
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
978 |
return(0); |
1
by brian
clean slate |
979 |
}
|
980 |
||
981 |
||
982 |
/** Merge buffers to make < MERGEBUFF2 buffers. */
|
|
983 |
||
984 |
int merge_many_buff(SORTPARAM *param, uchar *sort_buffer, |
|
985 |
BUFFPEK *buffpek, uint *maxbuffer, IO_CACHE *t_file) |
|
986 |
{
|
|
987 |
register uint i; |
|
988 |
IO_CACHE t_file2,*from_file,*to_file,*temp; |
|
989 |
BUFFPEK *lastbuff; |
|
990 |
||
991 |
if (*maxbuffer < MERGEBUFF2) |
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
992 |
return(0); /* purecov: inspected */ |
1
by brian
clean slate |
993 |
if (flush_io_cache(t_file) || |
994 |
open_cached_file(&t_file2,mysql_tmpdir,TEMP_PREFIX,DISK_BUFFER_SIZE, |
|
995 |
MYF(MY_WME))) |
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
996 |
return(1); /* purecov: inspected */ |
1
by brian
clean slate |
997 |
|
998 |
from_file= t_file ; to_file= &t_file2; |
|
999 |
while (*maxbuffer >= MERGEBUFF2) |
|
1000 |
{
|
|
1001 |
if (reinit_io_cache(from_file,READ_CACHE,0L,0,0)) |
|
1002 |
goto cleanup; |
|
1003 |
if (reinit_io_cache(to_file,WRITE_CACHE,0L,0,0)) |
|
1004 |
goto cleanup; |
|
1005 |
lastbuff=buffpek; |
|
1006 |
for (i=0 ; i <= *maxbuffer-MERGEBUFF*3/2 ; i+=MERGEBUFF) |
|
1007 |
{
|
|
1008 |
if (merge_buffers(param,from_file,to_file,sort_buffer,lastbuff++, |
|
1009 |
buffpek+i,buffpek+i+MERGEBUFF-1,0)) |
|
1010 |
goto cleanup; |
|
1011 |
}
|
|
1012 |
if (merge_buffers(param,from_file,to_file,sort_buffer,lastbuff++, |
|
1013 |
buffpek+i,buffpek+ *maxbuffer,0)) |
|
1014 |
break; /* purecov: inspected */ |
|
1015 |
if (flush_io_cache(to_file)) |
|
1016 |
break; /* purecov: inspected */ |
|
1017 |
temp=from_file; from_file=to_file; to_file=temp; |
|
1018 |
setup_io_cache(from_file); |
|
1019 |
setup_io_cache(to_file); |
|
1020 |
*maxbuffer= (uint) (lastbuff-buffpek)-1; |
|
1021 |
}
|
|
1022 |
cleanup: |
|
1023 |
close_cached_file(to_file); // This holds old result |
|
1024 |
if (to_file == t_file) |
|
1025 |
{
|
|
1026 |
*t_file=t_file2; // Copy result file |
|
1027 |
setup_io_cache(t_file); |
|
1028 |
}
|
|
1029 |
||
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
1030 |
return(*maxbuffer >= MERGEBUFF2); /* Return 1 if interrupted */ |
1
by brian
clean slate |
1031 |
} /* merge_many_buff */ |
1032 |
||
1033 |
||
1034 |
/**
|
|
1035 |
Read data to buffer.
|
|
1036 |
||
1037 |
@retval
|
|
1038 |
(uint)-1 if something goes wrong
|
|
1039 |
*/
|
|
1040 |
||
1041 |
uint read_to_buffer(IO_CACHE *fromfile, BUFFPEK *buffpek, |
|
1042 |
uint rec_length) |
|
1043 |
{
|
|
1044 |
register uint count; |
|
1045 |
uint length; |
|
1046 |
||
1047 |
if ((count=(uint) min((ha_rows) buffpek->max_keys,buffpek->count))) |
|
1048 |
{
|
|
31
by Brian Aker
Removed my versions of pread/pwrite from the Kernel |
1049 |
if (pread(fromfile->file,(uchar*) buffpek->base, (length= rec_length*count),buffpek->file_pos) == 0) |
1
by brian
clean slate |
1050 |
return((uint) -1); /* purecov: inspected */ |
1051 |
buffpek->key=buffpek->base; |
|
1052 |
buffpek->file_pos+= length; /* New filepos */ |
|
1053 |
buffpek->count-= count; |
|
1054 |
buffpek->mem_count= count; |
|
1055 |
}
|
|
1056 |
return (count*rec_length); |
|
1057 |
} /* read_to_buffer */ |
|
1058 |
||
1059 |
||
1060 |
/**
|
|
1061 |
Put all room used by freed buffer to use in adjacent buffer.
|
|
1062 |
||
1063 |
Note, that we can't simply distribute memory evenly between all buffers,
|
|
1064 |
because new areas must not overlap with old ones.
|
|
1065 |
||
1066 |
@param[in] queue list of non-empty buffers, without freed buffer
|
|
1067 |
@param[in] reuse empty buffer
|
|
1068 |
@param[in] key_length key length
|
|
1069 |
*/
|
|
1070 |
||
1071 |
void reuse_freed_buff(QUEUE *queue, BUFFPEK *reuse, uint key_length) |
|
1072 |
{
|
|
1073 |
uchar *reuse_end= reuse->base + reuse->max_keys * key_length; |
|
1074 |
for (uint i= 0; i < queue->elements; ++i) |
|
1075 |
{
|
|
1076 |
BUFFPEK *bp= (BUFFPEK *) queue_element(queue, i); |
|
1077 |
if (bp->base + bp->max_keys * key_length == reuse->base) |
|
1078 |
{
|
|
1079 |
bp->max_keys+= reuse->max_keys; |
|
1080 |
return; |
|
1081 |
}
|
|
1082 |
else if (bp->base == reuse_end) |
|
1083 |
{
|
|
1084 |
bp->base= reuse->base; |
|
1085 |
bp->max_keys+= reuse->max_keys; |
|
1086 |
return; |
|
1087 |
}
|
|
1088 |
}
|
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
1089 |
assert(0); |
1
by brian
clean slate |
1090 |
}
|
1091 |
||
1092 |
||
1093 |
/**
|
|
1094 |
Merge buffers to one buffer.
|
|
1095 |
||
1096 |
@param param Sort parameter
|
|
1097 |
@param from_file File with source data (BUFFPEKs point to this file)
|
|
1098 |
@param to_file File to write the sorted result data.
|
|
1099 |
@param sort_buffer Buffer for data to store up to MERGEBUFF2 sort keys.
|
|
1100 |
@param lastbuff OUT Store here BUFFPEK describing data written to to_file
|
|
1101 |
@param Fb First element in source BUFFPEKs array
|
|
1102 |
@param Tb Last element in source BUFFPEKs array
|
|
1103 |
@param flag
|
|
1104 |
||
1105 |
@retval
|
|
1106 |
0 OK
|
|
1107 |
@retval
|
|
1108 |
other error
|
|
1109 |
*/
|
|
1110 |
||
1111 |
int merge_buffers(SORTPARAM *param, IO_CACHE *from_file, |
|
1112 |
IO_CACHE *to_file, uchar *sort_buffer, |
|
1113 |
BUFFPEK *lastbuff, BUFFPEK *Fb, BUFFPEK *Tb, |
|
1114 |
int flag) |
|
1115 |
{
|
|
1116 |
int error; |
|
1117 |
uint rec_length,res_length,offset; |
|
1118 |
size_t sort_length; |
|
308
by Brian Aker
ulong conversion |
1119 |
uint32_t maxcount; |
1
by brian
clean slate |
1120 |
ha_rows max_rows,org_max_rows; |
1121 |
my_off_t to_start_filepos; |
|
1122 |
uchar *strpos; |
|
1123 |
BUFFPEK *buffpek; |
|
1124 |
QUEUE queue; |
|
1125 |
qsort2_cmp cmp; |
|
1126 |
void *first_cmp_arg; |
|
1127 |
volatile THD::killed_state *killed= ¤t_thd->killed; |
|
1128 |
THD::killed_state not_killable; |
|
1129 |
||
1130 |
status_var_increment(current_thd->status_var.filesort_merge_passes); |
|
1131 |
if (param->not_killable) |
|
1132 |
{
|
|
1133 |
killed= ¬_killable; |
|
1134 |
not_killable= THD::NOT_KILLED; |
|
1135 |
}
|
|
1136 |
||
1137 |
error=0; |
|
1138 |
rec_length= param->rec_length; |
|
1139 |
res_length= param->res_length; |
|
1140 |
sort_length= param->sort_length; |
|
1141 |
offset= rec_length-res_length; |
|
308
by Brian Aker
ulong conversion |
1142 |
maxcount= (uint32_t) (param->keys/((uint) (Tb-Fb) +1)); |
1
by brian
clean slate |
1143 |
to_start_filepos= my_b_tell(to_file); |
1144 |
strpos= (uchar*) sort_buffer; |
|
1145 |
org_max_rows=max_rows= param->max_rows; |
|
1146 |
||
1147 |
/* The following will fire if there is not enough space in sort_buffer */
|
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
1148 |
assert(maxcount!=0); |
1
by brian
clean slate |
1149 |
|
1150 |
if (param->unique_buff) |
|
1151 |
{
|
|
1152 |
cmp= param->compare; |
|
1153 |
first_cmp_arg= (void *) ¶m->cmp_context; |
|
1154 |
}
|
|
1155 |
else
|
|
1156 |
{
|
|
1157 |
cmp= get_ptr_compare(sort_length); |
|
1158 |
first_cmp_arg= (void*) &sort_length; |
|
1159 |
}
|
|
1160 |
if (init_queue(&queue, (uint) (Tb-Fb)+1, offsetof(BUFFPEK,key), 0, |
|
1161 |
(queue_compare) cmp, first_cmp_arg)) |
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
1162 |
return(1); /* purecov: inspected */ |
1
by brian
clean slate |
1163 |
for (buffpek= Fb ; buffpek <= Tb ; buffpek++) |
1164 |
{
|
|
1165 |
buffpek->base= strpos; |
|
1166 |
buffpek->max_keys= maxcount; |
|
1167 |
strpos+= (uint) (error= (int) read_to_buffer(from_file, buffpek, |
|
1168 |
rec_length)); |
|
1169 |
if (error == -1) |
|
1170 |
goto err; /* purecov: inspected */ |
|
1171 |
buffpek->max_keys= buffpek->mem_count; // If less data in buffers than expected |
|
1172 |
queue_insert(&queue, (uchar*) buffpek); |
|
1173 |
}
|
|
1174 |
||
1175 |
if (param->unique_buff) |
|
1176 |
{
|
|
1177 |
/*
|
|
1178 |
Called by Unique::get()
|
|
1179 |
Copy the first argument to param->unique_buff for unique removal.
|
|
1180 |
Store it also in 'to_file'.
|
|
1181 |
||
1182 |
This is safe as we know that there is always more than one element
|
|
1183 |
in each block to merge (This is guaranteed by the Unique:: algorithm
|
|
1184 |
*/
|
|
1185 |
buffpek= (BUFFPEK*) queue_top(&queue); |
|
1186 |
memcpy(param->unique_buff, buffpek->key, rec_length); |
|
1187 |
if (my_b_write(to_file, (uchar*) buffpek->key, rec_length)) |
|
1188 |
{
|
|
1189 |
error=1; goto err; /* purecov: inspected */ |
|
1190 |
}
|
|
1191 |
buffpek->key+= rec_length; |
|
1192 |
buffpek->mem_count--; |
|
1193 |
if (!--max_rows) |
|
1194 |
{
|
|
1195 |
error= 0; /* purecov: inspected */ |
|
1196 |
goto end; /* purecov: inspected */ |
|
1197 |
}
|
|
1198 |
queue_replaced(&queue); // Top element has been used |
|
1199 |
}
|
|
1200 |
else
|
|
1201 |
cmp= 0; // Not unique |
|
1202 |
||
1203 |
while (queue.elements > 1) |
|
1204 |
{
|
|
1205 |
if (*killed) |
|
1206 |
{
|
|
1207 |
error= 1; goto err; /* purecov: inspected */ |
|
1208 |
}
|
|
1209 |
for (;;) |
|
1210 |
{
|
|
1211 |
buffpek= (BUFFPEK*) queue_top(&queue); |
|
1212 |
if (cmp) // Remove duplicates |
|
1213 |
{
|
|
1214 |
if (!(*cmp)(first_cmp_arg, &(param->unique_buff), |
|
1215 |
(uchar**) &buffpek->key)) |
|
1216 |
goto skip_duplicate; |
|
212.6.6
by Mats Kindahl
Removing redundant use of casts in drizzled/ for memcmp(), memcpy(), memset(), and memmove(). |
1217 |
memcpy(param->unique_buff, buffpek->key, rec_length); |
1
by brian
clean slate |
1218 |
}
|
1219 |
if (flag == 0) |
|
1220 |
{
|
|
1221 |
if (my_b_write(to_file,(uchar*) buffpek->key, rec_length)) |
|
1222 |
{
|
|
1223 |
error=1; goto err; /* purecov: inspected */ |
|
1224 |
}
|
|
1225 |
}
|
|
1226 |
else
|
|
1227 |
{
|
|
1228 |
if (my_b_write(to_file, (uchar*) buffpek->key+offset, res_length)) |
|
1229 |
{
|
|
1230 |
error=1; goto err; /* purecov: inspected */ |
|
1231 |
}
|
|
1232 |
}
|
|
1233 |
if (!--max_rows) |
|
1234 |
{
|
|
1235 |
error= 0; /* purecov: inspected */ |
|
1236 |
goto end; /* purecov: inspected */ |
|
1237 |
}
|
|
1238 |
||
1239 |
skip_duplicate: |
|
1240 |
buffpek->key+= rec_length; |
|
1241 |
if (! --buffpek->mem_count) |
|
1242 |
{
|
|
1243 |
if (!(error= (int) read_to_buffer(from_file,buffpek, |
|
1244 |
rec_length))) |
|
1245 |
{
|
|
1246 |
VOID(queue_remove(&queue,0)); |
|
1247 |
reuse_freed_buff(&queue, buffpek, rec_length); |
|
1248 |
break; /* One buffer have been removed */ |
|
1249 |
}
|
|
1250 |
else if (error == -1) |
|
1251 |
goto err; /* purecov: inspected */ |
|
1252 |
}
|
|
1253 |
queue_replaced(&queue); /* Top element has been replaced */ |
|
1254 |
}
|
|
1255 |
}
|
|
1256 |
buffpek= (BUFFPEK*) queue_top(&queue); |
|
1257 |
buffpek->base= sort_buffer; |
|
1258 |
buffpek->max_keys= param->keys; |
|
1259 |
||
1260 |
/*
|
|
1261 |
As we know all entries in the buffer are unique, we only have to
|
|
1262 |
check if the first one is the same as the last one we wrote
|
|
1263 |
*/
|
|
1264 |
if (cmp) |
|
1265 |
{
|
|
1266 |
if (!(*cmp)(first_cmp_arg, &(param->unique_buff), (uchar**) &buffpek->key)) |
|
1267 |
{
|
|
1268 |
buffpek->key+= rec_length; // Remove duplicate |
|
1269 |
--buffpek->mem_count; |
|
1270 |
}
|
|
1271 |
}
|
|
1272 |
||
1273 |
do
|
|
1274 |
{
|
|
1275 |
if ((ha_rows) buffpek->mem_count > max_rows) |
|
1276 |
{ /* Don't write too many records */ |
|
1277 |
buffpek->mem_count= (uint) max_rows; |
|
1278 |
buffpek->count= 0; /* Don't read more */ |
|
1279 |
}
|
|
1280 |
max_rows-= buffpek->mem_count; |
|
1281 |
if (flag == 0) |
|
1282 |
{
|
|
1283 |
if (my_b_write(to_file,(uchar*) buffpek->key, |
|
1284 |
(rec_length*buffpek->mem_count))) |
|
1285 |
{
|
|
1286 |
error= 1; goto err; /* purecov: inspected */ |
|
1287 |
}
|
|
1288 |
}
|
|
1289 |
else
|
|
1290 |
{
|
|
1291 |
register uchar *end; |
|
1292 |
strpos= buffpek->key+offset; |
|
1293 |
for (end= strpos+buffpek->mem_count*rec_length ; |
|
1294 |
strpos != end ; |
|
1295 |
strpos+= rec_length) |
|
1296 |
{
|
|
1297 |
if (my_b_write(to_file, (uchar *) strpos, res_length)) |
|
1298 |
{
|
|
1299 |
error=1; goto err; |
|
1300 |
}
|
|
1301 |
}
|
|
1302 |
}
|
|
1303 |
}
|
|
1304 |
while ((error=(int) read_to_buffer(from_file,buffpek, rec_length)) |
|
1305 |
!= -1 && error != 0); |
|
1306 |
||
1307 |
end: |
|
1308 |
lastbuff->count= min(org_max_rows-max_rows, param->max_rows); |
|
1309 |
lastbuff->file_pos= to_start_filepos; |
|
1310 |
err: |
|
1311 |
delete_queue(&queue); |
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
1312 |
return(error); |
1
by brian
clean slate |
1313 |
} /* merge_buffers */ |
1314 |
||
1315 |
||
1316 |
/* Do a merge to output-file (save only positions) */
|
|
1317 |
||
1318 |
static int merge_index(SORTPARAM *param, uchar *sort_buffer, |
|
1319 |
BUFFPEK *buffpek, uint maxbuffer, |
|
1320 |
IO_CACHE *tempfile, IO_CACHE *outfile) |
|
1321 |
{
|
|
1322 |
if (merge_buffers(param,tempfile,outfile,sort_buffer,buffpek,buffpek, |
|
1323 |
buffpek+maxbuffer,1)) |
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
1324 |
return(1); /* purecov: inspected */ |
1325 |
return(0); |
|
1
by brian
clean slate |
1326 |
} /* merge_index */ |
1327 |
||
1328 |
||
308
by Brian Aker
ulong conversion |
1329 |
static uint suffix_length(uint32_t string_length) |
1
by brian
clean slate |
1330 |
{
|
1331 |
if (string_length < 256) |
|
1332 |
return 1; |
|
1333 |
if (string_length < 256L*256L) |
|
1334 |
return 2; |
|
1335 |
if (string_length < 256L*256L*256L) |
|
1336 |
return 3; |
|
1337 |
return 4; // Can't sort longer than 4G |
|
1338 |
}
|
|
1339 |
||
1340 |
||
1341 |
||
1342 |
/**
|
|
1343 |
Calculate length of sort key.
|
|
1344 |
||
1345 |
@param thd Thread handler
|
|
1346 |
@param sortorder Order of items to sort
|
|
1347 |
@param s_length Number of items to sort
|
|
1348 |
@param[out] multi_byte_charset Set to 1 if we are using multi-byte charset
|
|
1349 |
(In which case we have to use strxnfrm())
|
|
1350 |
||
1351 |
@note
|
|
1352 |
sortorder->length is updated for each sort item.
|
|
1353 |
@n
|
|
1354 |
sortorder->need_strxnfrm is set 1 if we have to use strxnfrm
|
|
1355 |
||
1356 |
@return
|
|
1357 |
Total length of sort buffer in bytes
|
|
1358 |
*/
|
|
1359 |
||
1360 |
static uint |
|
1361 |
sortlength(THD *thd, SORT_FIELD *sortorder, uint s_length, |
|
1362 |
bool *multi_byte_charset) |
|
1363 |
{
|
|
1364 |
register uint length; |
|
264.2.6
by Andrey Hristov
Constify the usage of CHARSET_INFO almost to the last place in the code. |
1365 |
const CHARSET_INFO *cs; |
1
by brian
clean slate |
1366 |
*multi_byte_charset= 0; |
1367 |
||
1368 |
length=0; |
|
1369 |
for (; s_length-- ; sortorder++) |
|
1370 |
{
|
|
1371 |
sortorder->need_strxnfrm= 0; |
|
1372 |
sortorder->suffix_length= 0; |
|
1373 |
if (sortorder->field) |
|
1374 |
{
|
|
1375 |
cs= sortorder->field->sort_charset(); |
|
1376 |
sortorder->length= sortorder->field->sort_length(); |
|
1377 |
||
1378 |
if (use_strnxfrm((cs=sortorder->field->sort_charset()))) |
|
1379 |
{
|
|
1380 |
sortorder->need_strxnfrm= 1; |
|
1381 |
*multi_byte_charset= 1; |
|
1382 |
sortorder->length= cs->coll->strnxfrmlen(cs, sortorder->length); |
|
1383 |
}
|
|
1384 |
if (sortorder->field->maybe_null()) |
|
1385 |
length++; // Place for NULL marker |
|
1386 |
}
|
|
1387 |
else
|
|
1388 |
{
|
|
1389 |
sortorder->result_type= sortorder->item->result_type(); |
|
152
by Brian Aker
longlong replacement |
1390 |
if (sortorder->item->result_as_int64_t()) |
1
by brian
clean slate |
1391 |
sortorder->result_type= INT_RESULT; |
1392 |
switch (sortorder->result_type) { |
|
1393 |
case STRING_RESULT: |
|
1394 |
sortorder->length=sortorder->item->max_length; |
|
1395 |
set_if_smaller(sortorder->length, thd->variables.max_sort_length); |
|
1396 |
if (use_strnxfrm((cs=sortorder->item->collation.collation))) |
|
1397 |
{
|
|
1398 |
sortorder->length= cs->coll->strnxfrmlen(cs, sortorder->length); |
|
1399 |
sortorder->need_strxnfrm= 1; |
|
1400 |
*multi_byte_charset= 1; |
|
1401 |
}
|
|
1402 |
else if (cs == &my_charset_bin) |
|
1403 |
{
|
|
1404 |
/* Store length last to be able to sort blob/varbinary */
|
|
1405 |
sortorder->suffix_length= suffix_length(sortorder->length); |
|
1406 |
sortorder->length+= sortorder->suffix_length; |
|
1407 |
}
|
|
1408 |
break; |
|
1409 |
case INT_RESULT: |
|
1410 |
#if SIZEOF_LONG_LONG > 4
|
|
152
by Brian Aker
longlong replacement |
1411 |
sortorder->length=8; // Size of intern int64_t |
1
by brian
clean slate |
1412 |
#else
|
1413 |
sortorder->length=4; |
|
1414 |
#endif
|
|
1415 |
break; |
|
1416 |
case DECIMAL_RESULT: |
|
1417 |
sortorder->length= |
|
1418 |
my_decimal_get_binary_size(sortorder->item->max_length - |
|
1419 |
(sortorder->item->decimals ? 1 : 0), |
|
1420 |
sortorder->item->decimals); |
|
1421 |
break; |
|
1422 |
case REAL_RESULT: |
|
1423 |
sortorder->length=sizeof(double); |
|
1424 |
break; |
|
1425 |
case ROW_RESULT: |
|
1426 |
default: |
|
1427 |
// This case should never be choosen
|
|
51.1.12
by Jay Pipes
Removed/replaced DBUG symbols |
1428 |
assert(0); |
1
by brian
clean slate |
1429 |
break; |
1430 |
}
|
|
1431 |
if (sortorder->item->maybe_null) |
|
1432 |
length++; // Place for NULL marker |
|
1433 |
}
|
|
1434 |
set_if_smaller(sortorder->length, thd->variables.max_sort_length); |
|
1435 |
length+=sortorder->length; |
|
1436 |
}
|
|
1437 |
sortorder->field= (Field*) 0; // end marker |
|
1438 |
return length; |
|
1439 |
}
|
|
1440 |
||
1441 |
||
1442 |
/**
|
|
1443 |
Get descriptors of fields appended to sorted fields and
|
|
1444 |
calculate its total length.
|
|
1445 |
||
1446 |
The function first finds out what fields are used in the result set.
|
|
1447 |
Then it calculates the length of the buffer to store the values of
|
|
1448 |
these fields together with the value of sort values.
|
|
1449 |
If the calculated length is not greater than max_length_for_sort_data
|
|
1450 |
the function allocates memory for an array of descriptors containing
|
|
1451 |
layouts for the values of the non-sorted fields in the buffer and
|
|
1452 |
fills them.
|
|
1453 |
||
1454 |
@param thd Current thread
|
|
1455 |
@param ptabfield Array of references to the table fields
|
|
1456 |
@param sortlength Total length of sorted fields
|
|
1457 |
@param[out] plength Total length of appended fields
|
|
1458 |
||
1459 |
@note
|
|
1460 |
The null bits for the appended values are supposed to be put together
|
|
1461 |
and stored the buffer just ahead of the value of the first field.
|
|
1462 |
||
1463 |
@return
|
|
1464 |
Pointer to the layout descriptors for the appended fields, if any
|
|
1465 |
@retval
|
|
1466 |
NULL if we do not store field values with sort data.
|
|
1467 |
*/
|
|
1468 |
||
1469 |
static SORT_ADDON_FIELD * |
|
1470 |
get_addon_fields(THD *thd, Field **ptabfield, uint sortlength, uint *plength) |
|
1471 |
{
|
|
1472 |
Field **pfield; |
|
1473 |
Field *field; |
|
1474 |
SORT_ADDON_FIELD *addonf; |
|
1475 |
uint length= 0; |
|
1476 |
uint fields= 0; |
|
1477 |
uint null_fields= 0; |
|
1478 |
MY_BITMAP *read_set= (*ptabfield)->table->read_set; |
|
1479 |
||
1480 |
/*
|
|
1481 |
If there is a reference to a field in the query add it
|
|
1482 |
to the the set of appended fields.
|
|
1483 |
Note for future refinement:
|
|
1484 |
This this a too strong condition.
|
|
1485 |
Actually we need only the fields referred in the
|
|
1486 |
result set. And for some of them it makes sense to use
|
|
1487 |
the values directly from sorted fields.
|
|
1488 |
*/
|
|
1489 |
*plength= 0; |
|
1490 |
||
1491 |
for (pfield= ptabfield; (field= *pfield) ; pfield++) |
|
1492 |
{
|
|
1493 |
if (!bitmap_is_set(read_set, field->field_index)) |
|
1494 |
continue; |
|
1495 |
if (field->flags & BLOB_FLAG) |
|
1496 |
return 0; |
|
1497 |
length+= field->max_packed_col_length(field->pack_length()); |
|
1498 |
if (field->maybe_null()) |
|
1499 |
null_fields++; |
|
1500 |
fields++; |
|
1501 |
}
|
|
1502 |
if (!fields) |
|
1503 |
return 0; |
|
1504 |
length+= (null_fields+7)/8; |
|
1505 |
||
1506 |
if (length+sortlength > thd->variables.max_length_for_sort_data || |
|
1507 |
!(addonf= (SORT_ADDON_FIELD *) my_malloc(sizeof(SORT_ADDON_FIELD)* |
|
1508 |
(fields+1), MYF(MY_WME)))) |
|
1509 |
return 0; |
|
1510 |
||
1511 |
*plength= length; |
|
1512 |
length= (null_fields+7)/8; |
|
1513 |
null_fields= 0; |
|
1514 |
for (pfield= ptabfield; (field= *pfield) ; pfield++) |
|
1515 |
{
|
|
1516 |
if (!bitmap_is_set(read_set, field->field_index)) |
|
1517 |
continue; |
|
1518 |
addonf->field= field; |
|
1519 |
addonf->offset= length; |
|
1520 |
if (field->maybe_null()) |
|
1521 |
{
|
|
1522 |
addonf->null_offset= null_fields/8; |
|
1523 |
addonf->null_bit= 1<<(null_fields & 7); |
|
1524 |
null_fields++; |
|
1525 |
}
|
|
1526 |
else
|
|
1527 |
{
|
|
1528 |
addonf->null_offset= 0; |
|
1529 |
addonf->null_bit= 0; |
|
1530 |
}
|
|
1531 |
addonf->length= field->max_packed_col_length(field->pack_length()); |
|
1532 |
length+= addonf->length; |
|
1533 |
addonf++; |
|
1534 |
}
|
|
1535 |
addonf->field= 0; // Put end marker |
|
1536 |
||
1537 |
return (addonf-fields); |
|
1538 |
}
|
|
1539 |
||
1540 |
||
1541 |
/**
|
|
1542 |
Copy (unpack) values appended to sorted fields from a buffer back to
|
|
1543 |
their regular positions specified by the Field::ptr pointers.
|
|
1544 |
||
1545 |
@param addon_field Array of descriptors for appended fields
|
|
1546 |
@param buff Buffer which to unpack the value from
|
|
1547 |
||
1548 |
@note
|
|
1549 |
The function is supposed to be used only as a callback function
|
|
1550 |
when getting field values for the sorted result set.
|
|
1551 |
||
1552 |
@return
|
|
1553 |
void.
|
|
1554 |
*/
|
|
1555 |
||
1556 |
static void |
|
1557 |
unpack_addon_fields(struct st_sort_addon_field *addon_field, uchar *buff) |
|
1558 |
{
|
|
1559 |
Field *field; |
|
1560 |
SORT_ADDON_FIELD *addonf= addon_field; |
|
1561 |
||
1562 |
for ( ; (field= addonf->field) ; addonf++) |
|
1563 |
{
|
|
1564 |
if (addonf->null_bit && (addonf->null_bit & buff[addonf->null_offset])) |
|
1565 |
{
|
|
1566 |
field->set_null(); |
|
1567 |
continue; |
|
1568 |
}
|
|
1569 |
field->set_notnull(); |
|
1570 |
field->unpack(field->ptr, buff + addonf->offset); |
|
1571 |
}
|
|
1572 |
}
|
|
1573 |
||
1574 |
/*
|
|
1575 |
** functions to change a double or float to a sortable string
|
|
1576 |
** The following should work for IEEE
|
|
1577 |
*/
|
|
1578 |
||
1579 |
#define DBL_EXP_DIG (sizeof(double)*8-DBL_MANT_DIG)
|
|
1580 |
||
1581 |
void change_double_for_sort(double nr,uchar *to) |
|
1582 |
{
|
|
1583 |
uchar *tmp=(uchar*) to; |
|
1584 |
if (nr == 0.0) |
|
1585 |
{ /* Change to zero string */ |
|
1586 |
tmp[0]=(uchar) 128; |
|
212.6.6
by Mats Kindahl
Removing redundant use of casts in drizzled/ for memcmp(), memcpy(), memset(), and memmove(). |
1587 |
memset(tmp+1, 0, sizeof(nr)-1); |
1
by brian
clean slate |
1588 |
}
|
1589 |
else
|
|
1590 |
{
|
|
1591 |
#ifdef WORDS_BIGENDIAN
|
|
212.6.3
by Mats Kindahl
Removing deprecated functions from code and replacing them with C99 equivalents: |
1592 |
memcpy(tmp,&nr,sizeof(nr)); |
1
by brian
clean slate |
1593 |
#else
|
1594 |
{
|
|
1595 |
uchar *ptr= (uchar*) &nr; |
|
1596 |
#if defined(__FLOAT_WORD_ORDER) && (__FLOAT_WORD_ORDER == __BIG_ENDIAN)
|
|
1597 |
tmp[0]= ptr[3]; tmp[1]=ptr[2]; tmp[2]= ptr[1]; tmp[3]=ptr[0]; |
|
1598 |
tmp[4]= ptr[7]; tmp[5]=ptr[6]; tmp[6]= ptr[5]; tmp[7]=ptr[4]; |
|
1599 |
#else
|
|
1600 |
tmp[0]= ptr[7]; tmp[1]=ptr[6]; tmp[2]= ptr[5]; tmp[3]=ptr[4]; |
|
1601 |
tmp[4]= ptr[3]; tmp[5]=ptr[2]; tmp[6]= ptr[1]; tmp[7]=ptr[0]; |
|
1602 |
#endif
|
|
1603 |
}
|
|
1604 |
#endif
|
|
1605 |
if (tmp[0] & 128) /* Negative */ |
|
1606 |
{ /* make complement */ |
|
1607 |
uint i; |
|
1608 |
for (i=0 ; i < sizeof(nr); i++) |
|
1609 |
tmp[i]=tmp[i] ^ (uchar) 255; |
|
1610 |
}
|
|
1611 |
else
|
|
1612 |
{ /* Set high and move exponent one up */ |
|
1613 |
ushort exp_part=(((ushort) tmp[0] << 8) | (ushort) tmp[1] | |
|
1614 |
(ushort) 32768); |
|
1615 |
exp_part+= (ushort) 1 << (16-1-DBL_EXP_DIG); |
|
1616 |
tmp[0]= (uchar) (exp_part >> 8); |
|
1617 |
tmp[1]= (uchar) exp_part; |
|
1618 |
}
|
|
1619 |
}
|
|
1620 |
}
|