1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
|
#ifndef SQL_SORT_INCLUDED
#define SQL_SORT_INCLUDED
/* Copyright (c) 2000, 2010, Oracle and/or its affiliates. All rights reserved.
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; version 2 of the License.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1335 USA */
#include "my_base.h" /* ha_rows */
#include <my_sys.h> /* qsort2_cmp */
#include "queues.h"
struct SORT_FIELD;
class Field;
struct TABLE;
/* Defines used by filesort and uniques */
#define MERGEBUFF 7
#define MERGEBUFF2 15
/*
The structure SORT_ADDON_FIELD describes a fixed layout
for field values appended to sorted values in records to be sorted
in the sort buffer.
Only fixed layout is supported now.
Null bit maps for the appended values is placed before the values
themselves. Offsets are from the last sorted field, that is from the
record referefence, which is still last component of sorted records.
It is preserved for backward compatiblility.
The structure is used tp store values of the additional fields
in the sort buffer. It is used also when these values are read
from a temporary file/buffer. As the reading procedures are beyond the
scope of the 'filesort' code the values have to be retrieved via
the callback function 'unpack_addon_fields'.
*/
typedef struct st_sort_addon_field
{
/* Sort addon packed field */
Field *field; /* Original field */
uint offset; /* Offset from the last sorted field */
uint null_offset; /* Offset to to null bit from the last sorted field */
uint length; /* Length in the sort buffer */
uint8 null_bit; /* Null bit mask for the field */
} SORT_ADDON_FIELD;
struct BUFFPEK_COMPARE_CONTEXT
{
qsort_cmp2 key_compare;
void *key_compare_arg;
};
/**
Descriptor for a merge chunk to be sort-merged.
A merge chunk is a sequence of pre-sorted records, written to a
temporary file. A Merge_chunk instance describes where this chunk is stored
in the file, and where it is located when it is in memory.
It is a POD because
- we read/write them from/to files.
We have accessors (getters/setters) for all struct members.
*/
struct Merge_chunk {
public:
Merge_chunk(): m_current_key(NULL),
m_file_position(0),
m_buffer_start(NULL),
m_buffer_end(NULL),
m_rowcount(0),
m_mem_count(0),
m_max_keys(0)
{}
my_off_t file_position() const { return m_file_position; }
void set_file_position(my_off_t val) { m_file_position= val; }
void advance_file_position(my_off_t val) { m_file_position+= val; }
uchar *buffer_start() { return m_buffer_start; }
const uchar *buffer_end() const { return m_buffer_end; }
void set_buffer(uchar *start, uchar *end)
{
m_buffer_start= start;
m_buffer_end= end;
}
void set_buffer_start(uchar *start)
{
m_buffer_start= start;
}
void set_buffer_end(uchar *end)
{
DBUG_ASSERT(m_buffer_end == NULL || end <= m_buffer_end);
m_buffer_end= end;
}
void init_current_key() { m_current_key= m_buffer_start; }
uchar *current_key() { return m_current_key; }
void advance_current_key(uint val) { m_current_key+= val; }
void decrement_rowcount(ha_rows val) { m_rowcount-= val; }
void set_rowcount(ha_rows val) { m_rowcount= val; }
ha_rows rowcount() const { return m_rowcount; }
ha_rows mem_count() const { return m_mem_count; }
void set_mem_count(ha_rows val) { m_mem_count= val; }
ha_rows decrement_mem_count() { return --m_mem_count; }
ha_rows max_keys() const { return m_max_keys; }
void set_max_keys(ha_rows val) { m_max_keys= val; }
size_t buffer_size() const { return m_buffer_end - m_buffer_start; }
/**
Tries to merge *this with *mc, returns true if successful.
The assumption is that *this is no longer in use,
and the space it has been allocated can be handed over to a
buffer which is adjacent to it.
*/
bool merge_freed_buff(Merge_chunk *mc) const
{
if (mc->m_buffer_end == m_buffer_start)
{
mc->m_buffer_end= m_buffer_end;
mc->m_max_keys+= m_max_keys;
return true;
}
else if (mc->m_buffer_start == m_buffer_end)
{
mc->m_buffer_start= m_buffer_start;
mc->m_max_keys+= m_max_keys;
return true;
}
return false;
}
uchar *m_current_key; /// The current key for this chunk.
my_off_t m_file_position;/// Current position in the file to be sorted.
uchar *m_buffer_start; /// Start of main-memory buffer for this chunk.
uchar *m_buffer_end; /// End of main-memory buffer for this chunk.
ha_rows m_rowcount; /// Number of unread rows in this chunk.
ha_rows m_mem_count; /// Number of rows in the main-memory buffer.
ha_rows m_max_keys; /// If we have fixed-size rows:
/// max number of rows in buffer.
};
typedef Bounds_checked_array<SORT_ADDON_FIELD> Addon_fields_array;
/**
This class wraps information about usage of addon fields.
An Addon_fields object is used both during packing of data in the filesort
buffer, and later during unpacking in 'Filesort_info::unpack_addon_fields'.
@see documentation for the Sort_addon_field struct.
@see documentation for get_addon_fields()
*/
class Addon_fields {
public:
Addon_fields(Addon_fields_array arr)
: m_field_descriptors(arr),
m_addon_buf(),
m_addon_buf_length(),
m_using_packed_addons(false)
{
DBUG_ASSERT(!arr.is_null());
}
SORT_ADDON_FIELD *begin() { return m_field_descriptors.begin(); }
SORT_ADDON_FIELD *end() { return m_field_descriptors.end(); }
/// rr_unpack_from_tempfile needs an extra buffer when unpacking.
uchar *allocate_addon_buf(uint sz)
{
m_addon_buf= (uchar *)my_malloc(sz, MYF(MY_WME | MY_THREAD_SPECIFIC));
if (m_addon_buf)
m_addon_buf_length= sz;
return m_addon_buf;
}
void free_addon_buff()
{
my_free(m_addon_buf);
m_addon_buf= NULL;
m_addon_buf_length= 0;
}
uchar *get_addon_buf() { return m_addon_buf; }
uint get_addon_buf_length() const { return m_addon_buf_length; }
void set_using_packed_addons(bool val)
{
m_using_packed_addons= val;
}
bool using_packed_addons() const
{
return m_using_packed_addons;
}
static bool can_pack_addon_fields(uint record_length)
{
return (record_length <= (0xFFFF));
}
/**
@returns Total number of bytes used for packed addon fields.
the size of the length field + size of null bits + sum of field sizes.
*/
static uint read_addon_length(uchar *p)
{
return size_of_length_field + uint2korr(p);
}
/**
Stores the number of bytes used for packed addon fields.
*/
static void store_addon_length(uchar *p, uint sz)
{
// We actually store the length of everything *after* the length field.
int2store(p, sz - size_of_length_field);
}
static const uint size_of_length_field= 2;
private:
Addon_fields_array m_field_descriptors;
uchar *m_addon_buf; ///< Buffer for unpacking addon fields.
uint m_addon_buf_length; ///< Length of the buffer.
bool m_using_packed_addons; ///< Are we packing the addon fields?
};
/**
There are two record formats for sorting:
|<key a><key b>...|<rowid>|
/ sort_length / ref_l /
or with "addon fields"
|<key a><key b>...|<null bits>|<field a><field b>...|
/ sort_length / addon_length /
The packed format for "addon fields"
|<key a><key b>...|<length>|<null bits>|<field a><field b>...|
/ sort_length / addon_length /
<key> Fields are fixed-size, specially encoded with
Field::make_sort_key() so we can do byte-by-byte compare.
<length> Contains the *actual* packed length (after packing) of
everything after the sort keys.
The size of the length field is 2 bytes,
which should cover most use cases: addon data <= 65535 bytes.
This is the same as max record size in MySQL.
<null bits> One bit for each nullable field, indicating whether the field
is null or not. May have size zero if no fields are nullable.
<field xx> Are stored with field->pack(), and retrieved with
field->unpack(). Addon fields within a record are stored
consecutively, with no "holes" or padding. They will have zero
size for NULL values.
*/
class Sort_param {
public:
uint rec_length; // Length of sorted records.
uint sort_length; // Length of sorted columns.
uint ref_length; // Length of record ref.
uint addon_length; // Length of addon_fields
uint res_length; // Length of records in final sorted file/buffer.
uint max_keys_per_buffer; // Max keys / buffer.
uint min_dupl_count;
ha_rows max_rows; // Select limit, or HA_POS_ERROR if unlimited.
ha_rows examined_rows; // Number of examined rows.
TABLE *sort_form; // For quicker make_sortkey.
/**
ORDER BY list with some precalculated info for filesort.
Array is created and owned by a Filesort instance.
*/
Bounds_checked_array<SORT_FIELD> local_sortorder;
Addon_fields *addon_fields; // Descriptors for companion fields.
bool using_pq;
uchar *unique_buff;
bool not_killable;
char* tmp_buffer;
// The fields below are used only by Unique class.
qsort2_cmp compare;
BUFFPEK_COMPARE_CONTEXT cmp_context;
Sort_param()
{
memset(this, 0, sizeof(*this));
}
void init_for_filesort(uint sortlen, TABLE *table,
ha_rows maxrows, bool sort_positions);
/// Enables the packing of addons if possible.
void try_to_pack_addons(ulong max_length_for_sort_data);
/// Are we packing the "addon fields"?
bool using_packed_addons() const
{
DBUG_ASSERT(m_using_packed_addons ==
(addon_fields != NULL &&
addon_fields->using_packed_addons()));
return m_using_packed_addons;
}
/// Are we using "addon fields"?
bool using_addon_fields() const
{
return addon_fields != NULL;
}
/**
Getter for record length and result length.
@param record_start Pointer to record.
@param [out] recl Store record length here.
@param [out] resl Store result length here.
*/
void get_rec_and_res_len(uchar *record_start, uint *recl, uint *resl)
{
if (!using_packed_addons())
{
*recl= rec_length;
*resl= res_length;
return;
}
uchar *plen= record_start + sort_length;
*resl= Addon_fields::read_addon_length(plen);
DBUG_ASSERT(*resl <= res_length);
const uchar *record_end= plen + *resl;
*recl= static_cast<uint>(record_end - record_start);
}
private:
uint m_packable_length;
bool m_using_packed_addons; ///< caches the value of using_packed_addons()
};
typedef Bounds_checked_array<uchar> Sort_buffer;
int merge_many_buff(Sort_param *param, Sort_buffer sort_buffer,
Merge_chunk *buffpek, uint *maxbuffer, IO_CACHE *t_file);
ulong read_to_buffer(IO_CACHE *fromfile, Merge_chunk *buffpek,
Sort_param *param);
bool merge_buffers(Sort_param *param,IO_CACHE *from_file,
IO_CACHE *to_file, Sort_buffer sort_buffer,
Merge_chunk *lastbuff, Merge_chunk *Fb,
Merge_chunk *Tb, int flag);
int merge_index(Sort_param *param, Sort_buffer sort_buffer,
Merge_chunk *buffpek, uint maxbuffer,
IO_CACHE *tempfile, IO_CACHE *outfile);
void reuse_freed_buff(QUEUE *queue, Merge_chunk *reuse, uint key_length);
#endif /* SQL_SORT_INCLUDED */
|