summaryrefslogtreecommitdiff
path: root/sql/sql_analyze_stmt.h
blob: 990c79fb9ad520c68c1351f58a0b21b054d58c7c (plain)
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
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
/*
   Copyright (c) 2015, 2020, MariaDB Corporation.

   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 Street, Fifth Floor, Boston, MA 02110-1335 USA */

/*

== ANALYZE-stmt classes ==

This file contains classes for supporting "ANALYZE statement" feature. These are 
a set of data structures that can be used to store the data about how the 
statement executed.

There are two kinds of data collection:

1. Various counters. We assume that incrementing counters has very low
overhead. Because of that, execution code increments counters unconditionally
(even when not running "ANALYZE $statement" commands. You run regular SELECT/
UPDATE/DELETE/etc and the counters are incremented).

As a free bonus, this lets us print detailed information into the slow query
log, should the query be slow.

2. Timing data. Measuring the time it took to run parts of query has noticeable
overhead. Because of that, we measure the time only when running "ANALYZE
$stmt").

*/

class Gap_time_tracker;
void attach_gap_time_tracker(THD *thd, Gap_time_tracker *gap_tracker, ulonglong timeval);
void process_gap_time_tracker(THD *thd, ulonglong timeval);

/*
  A class for tracking time it takes to do a certain action
*/
class Exec_time_tracker
{
protected:
  ulonglong count;
  ulonglong cycles;
  ulonglong last_start;

  void cycles_stop_tracking(THD *thd)
  {
    ulonglong end= my_timer_cycles();
    cycles += end - last_start;
    if (unlikely(end < last_start))
      cycles += ULONGLONG_MAX;

    process_gap_time_tracker(thd, end);
    if (my_gap_tracker)
      attach_gap_time_tracker(thd, my_gap_tracker, end);
  }
public:
  Exec_time_tracker() : count(0), cycles(0), my_gap_tracker(NULL) {}

  /*
    The time spent between stop_tracking() call on this object and any
    other time measurement will be billed to this tracker.
  */
  Gap_time_tracker *my_gap_tracker;

  // interface for collecting time
  void start_tracking(THD *thd)
  {
    last_start= my_timer_cycles();
    process_gap_time_tracker(thd, last_start);
  }

  void stop_tracking(THD *thd)
  {
    count++;
    cycles_stop_tracking(thd);
  }

  // interface for getting the time
  ulonglong get_loops() const { return count; }
  double get_time_ms() const
  {
    // convert 'cycles' to milliseconds.
    return 1000.0 * static_cast<double>(cycles) /
      static_cast<double>(sys_timer_info.cycles.frequency);
  }
};


/*
  Tracker for time spent between the calls to Exec_time_tracker's {start|
  stop}_tracking().

  @seealso Gap_time_tracker_data in sql_class.h
*/
class Gap_time_tracker
{
  ulonglong cycles;
public:
  Gap_time_tracker() : cycles(0) {}

  void log_time(ulonglong start, ulonglong end) {
    cycles += end - start;
  }

  double get_time_ms() const
  {
    // convert 'cycles' to milliseconds.
    return 1000.0 * static_cast<double>(cycles) /
      static_cast<double>(sys_timer_info.cycles.frequency);
  }
};



/*
  A class for counting certain actions (in all queries), and optionally
  collecting the timings (in ANALYZE queries).
*/

class Time_and_counter_tracker: public Exec_time_tracker
{
public: 
  const bool timed;
  
  Time_and_counter_tracker(bool timed_arg) : timed(timed_arg)
  {}
   
  /* Loops are counted in both ANALYZE and regular queries, as this is cheap */
  void incr_loops() { count++; }
  
  /*
    Unlike Exec_time_tracker::stop_tracking, we don't increase loops.
  */
  void stop_tracking(THD *thd)
  {
    cycles_stop_tracking(thd);
  }
};

#define ANALYZE_START_TRACKING(thd, tracker) \
  { \
    (tracker)->incr_loops(); \
    if (unlikely((tracker)->timed)) \
    { (tracker)->start_tracking(thd); } \
  }

#define ANALYZE_STOP_TRACKING(thd, tracker) \
  if (unlikely((tracker)->timed)) \
  { (tracker)->stop_tracking(thd); }

/*
  A class for collecting read statistics.
  
  The idea is that we run several scans. Each scans gets rows, and then filters
  some of them out.  We count scans, rows, and rows left after filtering.

  (note: at the moment, the class is not actually tied to a physical table. 
   It can be used to track reading from files, buffers, etc).
*/

class Table_access_tracker 
{
public:
  Table_access_tracker() :
    r_scans(0), r_rows(0), /*r_rows_after_table_cond(0),*/
    r_rows_after_where(0)
  {}

  ha_rows r_scans; /* How many scans were ran on this join_tab */
  ha_rows r_rows; /* How many rows we've got after that */
  ha_rows r_rows_after_where; /* Rows after applying attached part of WHERE */

  bool has_scans() const { return (r_scans != 0); }
  ha_rows get_loops() const { return r_scans; }
  double get_avg_rows() const
  {
    return r_scans
      ? static_cast<double>(r_rows) / static_cast<double>(r_scans)
      : 0;
  }

  double get_filtered_after_where() const
  {
    return r_rows > 0
      ? static_cast<double>(r_rows_after_where) /
        static_cast<double>(r_rows)
      : 1.0;
  }

  inline void on_scan_init() { r_scans++; }
  inline void on_record_read() { r_rows++; }
  inline void on_record_after_where() { r_rows_after_where++; }
};


class Json_writer;

/*
  This stores the data about how filesort executed.

  A few things from here (e.g. r_used_pq, r_limit) belong to the query plan,
  however, these parameters are calculated right during the execution so we 
  can't easily put them into the query plan.

  The class is designed to handle multiple invocations of filesort().
*/

class Filesort_tracker : public Sql_alloc
{
public:
  Filesort_tracker(bool do_timing) :
    time_tracker(do_timing), r_limit(0), r_used_pq(0),
    r_examined_rows(0), r_sorted_rows(0), r_output_rows(0),
    sort_passes(0),
    sort_buffer_size(0),
    r_using_addons(false),
    r_packed_addon_fields(false),
    r_sort_keys_packed(false)
  {}
  
  /* Functions that filesort uses to report various things about its execution */

  inline void report_use(THD *thd, ha_rows r_limit_arg)
  {
    if (!time_tracker.get_loops())
      r_limit= r_limit_arg;
    else
      r_limit= (r_limit != r_limit_arg)? 0: r_limit_arg;

    ANALYZE_START_TRACKING(thd, &time_tracker);
  }
  inline void incr_pq_used() { r_used_pq++; }

  inline void report_row_numbers(ha_rows examined_rows, 
                                 ha_rows sorted_rows,
                                 ha_rows returned_rows) 
  { 
    r_examined_rows += examined_rows;
    r_sorted_rows   += sorted_rows;
    r_output_rows   += returned_rows;
  }

  inline void report_merge_passes_at_start(ulong passes)
  {
    sort_passes -= passes;
  }
  inline void report_merge_passes_at_end(THD *thd, ulong passes)
  {
    ANALYZE_STOP_TRACKING(thd, &time_tracker);
    sort_passes += passes;
  }

  inline void report_sort_buffer_size(size_t bufsize)
  {
    if (sort_buffer_size)
      sort_buffer_size= ulonglong(-1); // multiple buffers of different sizes
    else
      sort_buffer_size= bufsize;
  }

  inline void report_addon_fields_format(bool addons_packed)
  {
    r_using_addons= true;
    r_packed_addon_fields= addons_packed;
  }
  inline void report_sort_keys_format(bool sort_keys_packed)
  {
    r_sort_keys_packed= sort_keys_packed;
  }

  void get_data_format(String *str);

  /* Functions to get the statistics */
  void print_json_members(Json_writer *writer);

  ulonglong get_r_loops() const { return time_tracker.get_loops(); }
  double get_avg_examined_rows() const
  {
    return static_cast<double>(r_examined_rows) /
      static_cast<double>(get_r_loops());
  }
  double get_avg_returned_rows() const
  {
    return static_cast<double>(r_output_rows) /
      static_cast<double>(get_r_loops());
  }
  double get_r_filtered() const
  {
    return r_examined_rows > 0
      ? static_cast<double>(r_sorted_rows) /
        static_cast<double>(r_examined_rows)
      : 1.0;
  }
private:
  Time_and_counter_tracker time_tracker;

  //ulonglong r_loops; /* How many times filesort was invoked */
  /*
    LIMIT is typically a constant. There is never "LIMIT 0".
      HA_POS_ERROR means we never had a limit
      0            means different values of LIMIT were used in 
                   different filesort invocations
      other value  means the same LIMIT value was used every time.
  */
  ulonglong r_limit;
  ulonglong r_used_pq; /* How many times PQ was used */

  /* How many rows were examined (before checking the select->cond) */
  ulonglong r_examined_rows;
  
  /* 
    How many rows were put into sorting (this is examined_rows minus rows that
    didn't pass the WHERE condition)
  */
  ulonglong r_sorted_rows;

  /*
    How many rows were returned. This is equal to r_sorted_rows, unless there
    was a LIMIT N clause in which case filesort would not have returned more
    than N rows.
  */
  ulonglong r_output_rows;

  /* How many sorts in total (divide by r_count to get the average) */
  ulonglong sort_passes;
  
  /* 
    0              - means not used (or not known 
    (ulonglong)-1  - multiple
    other          - value
  */
  ulonglong sort_buffer_size;
  bool r_using_addons;
  bool r_packed_addon_fields;
  bool r_sort_keys_packed;
};


/**
  A class to collect data about how rowid filter is executed.

  It stores information about how rowid filter container is filled,
  containers size and observed selectivity.

  The observed selectivity is calculated in this way.
  Some elements elem_set are checked if they belong to container.
  Observed selectivity is calculated as the count of elem_set
  elements that belong to container devided by all elem_set elements.
*/

class Rowid_filter_tracker : public Sql_alloc
{
private:
  /* A member to track the time to fill the rowid filter */
  Time_and_counter_tracker time_tracker;

  /* Size of the rowid filter container buffer */
  size_t container_buff_size;

  /* Count of elements that were used to fill the rowid filter container */
  uint container_elements;

  /* Elements counts used for observed selectivity calculation */
  uint n_checks;
  uint n_positive_checks;
public:
  Rowid_filter_tracker(bool do_timing) :
    time_tracker(do_timing), container_buff_size(0),
    container_elements(0), n_checks(0), n_positive_checks(0)
  {}

  inline void start_tracking(THD *thd)
  {
    ANALYZE_START_TRACKING(thd, &time_tracker);
  }

  inline void stop_tracking(THD *thd)
  {
    ANALYZE_STOP_TRACKING(thd, &time_tracker);
  }

  /* Save container buffer size in bytes */
  inline void report_container_buff_size(uint elem_size)
  {
   container_buff_size= container_elements * elem_size / 8;
  }

  Time_and_counter_tracker *get_time_tracker()
  {
    return &time_tracker;
  }

  double get_time_fill_container_ms() const
  {
    return time_tracker.get_time_ms();
  }

  void increment_checked_elements_count(bool was_checked)
  {
    n_checks++;
    if (was_checked)
     n_positive_checks++;
  }

  inline void increment_container_elements_count() { container_elements++; }

  uint get_container_elements() const { return container_elements; }

  double get_r_selectivity_pct() const
  {
    return static_cast<double>(n_positive_checks) /
      static_cast<double>(n_checks);
  }

  size_t get_container_buff_size() const { return container_buff_size; }
};