histogram.h revision ddb351dbec246cf1fab5ec20d2d5520909041de1
1// Copyright (c) 2011 The Chromium Authors. All rights reserved.
2// Use of this source code is governed by a BSD-style license that can be
3// found in the LICENSE file.
4
5// Histogram is an object that aggregates statistics, and can summarize them in
6// various forms, including ASCII graphical, HTML, and numerically (as a
7// vector of numbers corresponding to each of the aggregating buckets).
8
9// It supports calls to accumulate either time intervals (which are processed
10// as integral number of milliseconds), or arbitrary integral units.
11
12// The default layout of buckets is exponential.  For example, buckets might
13// contain (sequentially) the count of values in the following intervals:
14// [0,1), [1,2), [2,4), [4,8), [8,16), [16,32), [32,64), [64,infinity)
15// That bucket allocation would actually result from construction of a histogram
16// for values between 1 and 64, with 8 buckets, such as:
17// Histogram count(L"some name", 1, 64, 8);
18// Note that the underflow bucket [0,1) and the overflow bucket [64,infinity)
19// are not counted by the constructor in the user supplied "bucket_count"
20// argument.
21// The above example has an exponential ratio of 2 (doubling the bucket width
22// in each consecutive bucket.  The Histogram class automatically calculates
23// the smallest ratio that it can use to construct the number of buckets
24// selected in the constructor.  An another example, if you had 50 buckets,
25// and millisecond time values from 1 to 10000, then the ratio between
26// consecutive bucket widths will be approximately somewhere around the 50th
27// root of 10000.  This approach provides very fine grain (narrow) buckets
28// at the low end of the histogram scale, but allows the histogram to cover a
29// gigantic range with the addition of very few buckets.
30
31// Histograms use a pattern involving a function static variable, that is a
32// pointer to a histogram.  This static is explicitly initialized on any thread
33// that detects a uninitialized (NULL) pointer.  The potentially racy
34// initialization is not a problem as it is always set to point to the same
35// value (i.e., the FactoryGet always returns the same value).  FactoryGet
36// is also completely thread safe, which results in a completely thread safe,
37// and relatively fast, set of counters.  To avoid races at shutdown, the static
38// pointer is NOT deleted, and we leak the histograms at process termination.
39
40#ifndef BASE_METRICS_HISTOGRAM_H_
41#define BASE_METRICS_HISTOGRAM_H_
42#pragma once
43
44#include <map>
45#include <string>
46#include <vector>
47
48#include "base/base_api.h"
49#include "base/gtest_prod_util.h"
50#include "base/logging.h"
51#include "base/time.h"
52
53class Pickle;
54
55namespace base {
56
57class Lock;
58
59//------------------------------------------------------------------------------
60// Provide easy general purpose histogram in a macro, just like stats counters.
61// The first four macros use 50 buckets.
62
63#define HISTOGRAM_TIMES(name, sample) HISTOGRAM_CUSTOM_TIMES( \
64    name, sample, base::TimeDelta::FromMilliseconds(1), \
65    base::TimeDelta::FromSeconds(10), 50)
66
67#define HISTOGRAM_COUNTS(name, sample) HISTOGRAM_CUSTOM_COUNTS( \
68    name, sample, 1, 1000000, 50)
69
70#define HISTOGRAM_COUNTS_100(name, sample) HISTOGRAM_CUSTOM_COUNTS( \
71    name, sample, 1, 100, 50)
72
73#define HISTOGRAM_COUNTS_10000(name, sample) HISTOGRAM_CUSTOM_COUNTS( \
74    name, sample, 1, 10000, 50)
75
76#define HISTOGRAM_CUSTOM_COUNTS(name, sample, min, max, bucket_count) do { \
77    static base::Histogram* counter(NULL); \
78    if (!counter) \
79      counter = base::Histogram::FactoryGet(name, min, max, bucket_count, \
80                                            base::Histogram::kNoFlags); \
81    DCHECK_EQ(name, counter->histogram_name()); \
82    counter->Add(sample); \
83  } while (0)
84
85#define HISTOGRAM_PERCENTAGE(name, under_one_hundred) \
86    HISTOGRAM_ENUMERATION(name, under_one_hundred, 101)
87
88// For folks that need real specific times, use this to select a precise range
89// of times you want plotted, and the number of buckets you want used.
90#define HISTOGRAM_CUSTOM_TIMES(name, sample, min, max, bucket_count) do { \
91    static base::Histogram* counter(NULL); \
92    if (!counter) \
93      counter = base::Histogram::FactoryTimeGet(name, min, max, bucket_count, \
94                                                base::Histogram::kNoFlags); \
95    DCHECK_EQ(name, counter->histogram_name()); \
96    counter->AddTime(sample); \
97  } while (0)
98
99// DO NOT USE THIS.  It is being phased out, in favor of HISTOGRAM_CUSTOM_TIMES.
100#define HISTOGRAM_CLIPPED_TIMES(name, sample, min, max, bucket_count) do { \
101    static base::Histogram* counter(NULL); \
102    if (!counter) \
103      counter = base::Histogram::FactoryTimeGet(name, min, max, bucket_count, \
104                                                base::Histogram::kNoFlags); \
105    DCHECK_EQ(name, counter->histogram_name()); \
106    if ((sample) < (max)) counter->AddTime(sample); \
107  } while (0)
108
109// Support histograming of an enumerated value.  The samples should always be
110// less than boundary_value.
111
112#define HISTOGRAM_ENUMERATION(name, sample, boundary_value) do { \
113    static base::Histogram* counter(NULL); \
114    if (!counter) \
115      counter = base::LinearHistogram::FactoryGet(name, 1, boundary_value, \
116          boundary_value + 1, base::Histogram::kNoFlags); \
117    DCHECK_EQ(name, counter->histogram_name()); \
118    counter->Add(sample); \
119  } while (0)
120
121#define HISTOGRAM_CUSTOM_ENUMERATION(name, sample, custom_ranges) do { \
122    static base::Histogram* counter(NULL); \
123    if (!counter) \
124      counter = base::CustomHistogram::FactoryGet(name, custom_ranges, \
125                                                  base::Histogram::kNoFlags); \
126    DCHECK_EQ(name, counter->histogram_name()); \
127    counter->Add(sample); \
128  } while (0)
129
130
131//------------------------------------------------------------------------------
132// Define Debug vs non-debug flavors of macros.
133#ifndef NDEBUG
134
135#define DHISTOGRAM_TIMES(name, sample) HISTOGRAM_TIMES(name, sample)
136#define DHISTOGRAM_COUNTS(name, sample) HISTOGRAM_COUNTS(name, sample)
137#define DHISTOGRAM_PERCENTAGE(name, under_one_hundred) HISTOGRAM_PERCENTAGE(\
138    name, under_one_hundred)
139#define DHISTOGRAM_CUSTOM_TIMES(name, sample, min, max, bucket_count) \
140    HISTOGRAM_CUSTOM_TIMES(name, sample, min, max, bucket_count)
141#define DHISTOGRAM_CLIPPED_TIMES(name, sample, min, max, bucket_count) \
142    HISTOGRAM_CLIPPED_TIMES(name, sample, min, max, bucket_count)
143#define DHISTOGRAM_CUSTOM_COUNTS(name, sample, min, max, bucket_count) \
144    HISTOGRAM_CUSTOM_COUNTS(name, sample, min, max, bucket_count)
145#define DHISTOGRAM_ENUMERATION(name, sample, boundary_value) \
146    HISTOGRAM_ENUMERATION(name, sample, boundary_value)
147#define DHISTOGRAM_CUSTOM_ENUMERATION(name, sample, custom_ranges) \
148    HISTOGRAM_CUSTOM_ENUMERATION(name, sample, custom_ranges)
149
150#else  // NDEBUG
151
152#define DHISTOGRAM_TIMES(name, sample) do {} while (0)
153#define DHISTOGRAM_COUNTS(name, sample) do {} while (0)
154#define DHISTOGRAM_PERCENTAGE(name, under_one_hundred) do {} while (0)
155#define DHISTOGRAM_CUSTOM_TIMES(name, sample, min, max, bucket_count) \
156    do {} while (0)
157#define DHISTOGRAM_CLIPPED_TIMES(name, sample, min, max, bucket_count) \
158    do {} while (0)
159#define DHISTOGRAM_CUSTOM_COUNTS(name, sample, min, max, bucket_count) \
160    do {} while (0)
161#define DHISTOGRAM_ENUMERATION(name, sample, boundary_value) do {} while (0)
162#define DHISTOGRAM_CUSTOM_ENUMERATION(name, sample, custom_ranges) \
163    do {} while (0)
164
165#endif  // NDEBUG
166
167//------------------------------------------------------------------------------
168// The following macros provide typical usage scenarios for callers that wish
169// to record histogram data, and have the data submitted/uploaded via UMA.
170// Not all systems support such UMA, but if they do, the following macros
171// should work with the service.
172
173#define UMA_HISTOGRAM_TIMES(name, sample) UMA_HISTOGRAM_CUSTOM_TIMES( \
174    name, sample, base::TimeDelta::FromMilliseconds(1), \
175    base::TimeDelta::FromSeconds(10), 50)
176
177#define UMA_HISTOGRAM_MEDIUM_TIMES(name, sample) UMA_HISTOGRAM_CUSTOM_TIMES( \
178    name, sample, base::TimeDelta::FromMilliseconds(10), \
179    base::TimeDelta::FromMinutes(3), 50)
180
181// Use this macro when times can routinely be much longer than 10 seconds.
182#define UMA_HISTOGRAM_LONG_TIMES(name, sample) UMA_HISTOGRAM_CUSTOM_TIMES( \
183    name, sample, base::TimeDelta::FromMilliseconds(1), \
184    base::TimeDelta::FromHours(1), 50)
185
186#define UMA_HISTOGRAM_CUSTOM_TIMES(name, sample, min, max, bucket_count) do { \
187    static base::Histogram* counter(NULL); \
188    if (!counter) \
189      counter = base::Histogram::FactoryTimeGet(name, min, max, bucket_count, \
190            base::Histogram::kUmaTargetedHistogramFlag); \
191    DCHECK_EQ(name, counter->histogram_name()); \
192    counter->AddTime(sample); \
193  } while (0)
194
195// DO NOT USE THIS.  It is being phased out, in favor of HISTOGRAM_CUSTOM_TIMES.
196#define UMA_HISTOGRAM_CLIPPED_TIMES(name, sample, min, max, bucket_count) do { \
197    static base::Histogram* counter(NULL); \
198    if (!counter) \
199      counter = base::Histogram::FactoryTimeGet(name, min, max, bucket_count, \
200           base::Histogram::kUmaTargetedHistogramFlag); \
201    DCHECK_EQ(name, counter->histogram_name()); \
202    if ((sample) < (max)) counter->AddTime(sample); \
203  } while (0)
204
205#define UMA_HISTOGRAM_COUNTS(name, sample) UMA_HISTOGRAM_CUSTOM_COUNTS( \
206    name, sample, 1, 1000000, 50)
207
208#define UMA_HISTOGRAM_COUNTS_100(name, sample) UMA_HISTOGRAM_CUSTOM_COUNTS( \
209    name, sample, 1, 100, 50)
210
211#define UMA_HISTOGRAM_COUNTS_10000(name, sample) UMA_HISTOGRAM_CUSTOM_COUNTS( \
212    name, sample, 1, 10000, 50)
213
214#define UMA_HISTOGRAM_CUSTOM_COUNTS(name, sample, min, max, bucket_count) do { \
215    static base::Histogram* counter(NULL); \
216    if (!counter) \
217      counter = base::Histogram::FactoryGet(name, min, max, bucket_count, \
218          base::Histogram::kUmaTargetedHistogramFlag); \
219    DCHECK_EQ(name, counter->histogram_name()); \
220    counter->Add(sample); \
221  } while (0)
222
223#define UMA_HISTOGRAM_MEMORY_KB(name, sample) UMA_HISTOGRAM_CUSTOM_COUNTS( \
224    name, sample, 1000, 500000, 50)
225
226#define UMA_HISTOGRAM_MEMORY_MB(name, sample) UMA_HISTOGRAM_CUSTOM_COUNTS( \
227    name, sample, 1, 1000, 50)
228
229#define UMA_HISTOGRAM_PERCENTAGE(name, under_one_hundred) \
230    UMA_HISTOGRAM_ENUMERATION(name, under_one_hundred, 101)
231
232#define UMA_HISTOGRAM_BOOLEAN(name, sample) do { \
233    static base::Histogram* counter(NULL); \
234    if (!counter) \
235      counter = base::BooleanHistogram::FactoryGet(name, \
236          base::Histogram::kUmaTargetedHistogramFlag); \
237    DCHECK_EQ(name, counter->histogram_name()); \
238    counter->AddBoolean(sample); \
239  } while (0)
240
241#define UMA_HISTOGRAM_ENUMERATION(name, sample, boundary_value) do { \
242    static base::Histogram* counter(NULL); \
243    if (!counter) \
244      counter = base::LinearHistogram::FactoryGet(name, 1, boundary_value, \
245          boundary_value + 1, base::Histogram::kUmaTargetedHistogramFlag); \
246    DCHECK_EQ(name, counter->histogram_name()); \
247    counter->Add(sample); \
248  } while (0)
249
250#define UMA_HISTOGRAM_CUSTOM_ENUMERATION(name, sample, custom_ranges) do { \
251    static base::Histogram* counter(NULL); \
252    if (!counter) \
253      counter = base::CustomHistogram::FactoryGet(name, custom_ranges, \
254          base::Histogram::kUmaTargetedHistogramFlag); \
255    DCHECK_EQ(name, counter->histogram_name()); \
256    counter->Add(sample); \
257  } while (0)
258
259//------------------------------------------------------------------------------
260
261class BooleanHistogram;
262class CustomHistogram;
263class Histogram;
264class LinearHistogram;
265
266class BASE_API Histogram {
267 public:
268  typedef int Sample;  // Used for samples (and ranges of samples).
269  typedef int Count;  // Used to count samples in a bucket.
270  static const Sample kSampleType_MAX = INT_MAX;
271  // Initialize maximum number of buckets in histograms as 16,384.
272  static const size_t kBucketCount_MAX;
273
274  typedef std::vector<Count> Counts;
275  typedef std::vector<Sample> Ranges;
276
277  // These enums are used to facilitate deserialization of renderer histograms
278  // into the browser.
279  enum ClassType {
280    HISTOGRAM,
281    LINEAR_HISTOGRAM,
282    BOOLEAN_HISTOGRAM,
283    CUSTOM_HISTOGRAM,
284    NOT_VALID_IN_RENDERER
285  };
286
287  enum BucketLayout {
288    EXPONENTIAL,
289    LINEAR,
290    CUSTOM
291  };
292
293  enum Flags {
294    kNoFlags = 0,
295    kUmaTargetedHistogramFlag = 0x1,  // Histogram should be UMA uploaded.
296
297    // Indicate that the histogram was pickled to be sent across an IPC Channel.
298    // If we observe this flag on a histogram being aggregated into after IPC,
299    // then we are running in a single process mode, and the aggregation should
300    // not take place (as we would be aggregating back into the source
301    // histogram!).
302    kIPCSerializationSourceFlag = 0x10,
303
304    kHexRangePrintingFlag = 0x8000,  // Fancy bucket-naming supported.
305  };
306
307  enum Inconsistencies {
308    NO_INCONSISTENCIES = 0x0,
309    RANGE_CHECKSUM_ERROR = 0x1,
310    BUCKET_ORDER_ERROR = 0x2,
311    COUNT_HIGH_ERROR = 0x4,
312    COUNT_LOW_ERROR = 0x8,
313
314    NEVER_EXCEEDED_VALUE = 0x10
315  };
316
317  struct DescriptionPair {
318    Sample sample;
319    const char* description;  // Null means end of a list of pairs.
320  };
321
322  //----------------------------------------------------------------------------
323  // Statistic values, developed over the life of the histogram.
324
325  class BASE_API SampleSet {
326   public:
327    explicit SampleSet();
328    ~SampleSet();
329
330    // Adjust size of counts_ for use with given histogram.
331    void Resize(const Histogram& histogram);
332    void CheckSize(const Histogram& histogram) const;
333
334    // Accessor for histogram to make routine additions.
335    void Accumulate(Sample value, Count count, size_t index);
336
337    // Accessor methods.
338    Count counts(size_t i) const { return counts_[i]; }
339    Count TotalCount() const;
340    int64 sum() const { return sum_; }
341    int64 redundant_count() const { return redundant_count_; }
342
343    // Arithmetic manipulation of corresponding elements of the set.
344    void Add(const SampleSet& other);
345    void Subtract(const SampleSet& other);
346
347    bool Serialize(Pickle* pickle) const;
348    bool Deserialize(void** iter, const Pickle& pickle);
349
350   protected:
351    // Actual histogram data is stored in buckets, showing the count of values
352    // that fit into each bucket.
353    Counts counts_;
354
355    // Save simple stats locally.  Note that this MIGHT get done in base class
356    // without shared memory at some point.
357    int64 sum_;         // sum of samples.
358
359   private:
360    // Allow tests to corrupt our innards for testing purposes.
361    FRIEND_TEST(HistogramTest, CorruptSampleCounts);
362
363    // To help identify memory corruption, we reduntantly save the number of
364    // samples we've accumulated into all of our buckets.  We can compare this
365    // count to the sum of the counts in all buckets, and detect problems.  Note
366    // that due to races in histogram accumulation (if a histogram is indeed
367    // updated on several threads simultaneously), the tallies might mismatch,
368    // and also the snapshotting code may asynchronously get a mismatch (though
369    // generally either race based mismatch cause is VERY rare).
370    int64 redundant_count_;
371  };
372
373  //----------------------------------------------------------------------------
374  // minimum should start from 1. 0 is invalid as a minimum. 0 is an implicit
375  // default underflow bucket.
376  static Histogram* FactoryGet(const std::string& name,
377                               Sample minimum,
378                               Sample maximum,
379                               size_t bucket_count,
380                               Flags flags);
381  static Histogram* FactoryTimeGet(const std::string& name,
382                                   base::TimeDelta minimum,
383                                   base::TimeDelta maximum,
384                                   size_t bucket_count,
385                                   Flags flags);
386
387  void Add(int value);
388
389  // This method is an interface, used only by BooleanHistogram.
390  virtual void AddBoolean(bool value);
391
392  // Accept a TimeDelta to increment.
393  void AddTime(TimeDelta time) {
394    Add(static_cast<int>(time.InMilliseconds()));
395  }
396
397  void AddSampleSet(const SampleSet& sample);
398
399  // This method is an interface, used only by LinearHistogram.
400  virtual void SetRangeDescriptions(const DescriptionPair descriptions[]);
401
402  // The following methods provide graphical histogram displays.
403  void WriteHTMLGraph(std::string* output) const;
404  void WriteAscii(bool graph_it, const std::string& newline,
405                  std::string* output) const;
406
407  // Support generic flagging of Histograms.
408  // 0x1 Currently used to mark this histogram to be recorded by UMA..
409  // 0x8000 means print ranges in hex.
410  void SetFlags(Flags flags) { flags_ = static_cast<Flags> (flags_ | flags); }
411  void ClearFlags(Flags flags) { flags_ = static_cast<Flags>(flags_ & ~flags); }
412  int flags() const { return flags_; }
413
414  // Convenience methods for serializing/deserializing the histograms.
415  // Histograms from Renderer process are serialized and sent to the browser.
416  // Browser process reconstructs the histogram from the pickled version
417  // accumulates the browser-side shadow copy of histograms (that mirror
418  // histograms created in the renderer).
419
420  // Serialize the given snapshot of a Histogram into a String. Uses
421  // Pickle class to flatten the object.
422  static std::string SerializeHistogramInfo(const Histogram& histogram,
423                                            const SampleSet& snapshot);
424  // The following method accepts a list of pickled histograms and
425  // builds a histogram and updates shadow copy of histogram data in the
426  // browser process.
427  static bool DeserializeHistogramInfo(const std::string& histogram_info);
428
429  // Check to see if bucket ranges, counts and tallies in the snapshot are
430  // consistent with the bucket ranges and checksums in our histogram.  This can
431  // produce a false-alarm if a race occurred in the reading of the data during
432  // a SnapShot process, but should otherwise be false at all times (unless we
433  // have memory over-writes, or DRAM failures).
434  virtual Inconsistencies FindCorruption(const SampleSet& snapshot) const;
435
436  //----------------------------------------------------------------------------
437  // Accessors for factory constuction, serialization and testing.
438  //----------------------------------------------------------------------------
439  virtual ClassType histogram_type() const;
440  const std::string& histogram_name() const { return histogram_name_; }
441  Sample declared_min() const { return declared_min_; }
442  Sample declared_max() const { return declared_max_; }
443  virtual Sample ranges(size_t i) const;
444  uint32 range_checksum() const { return range_checksum_; }
445  virtual size_t bucket_count() const;
446  // Snapshot the current complete set of sample data.
447  // Override with atomic/locked snapshot if needed.
448  virtual void SnapshotSample(SampleSet* sample) const;
449
450  virtual bool HasConstructorArguments(Sample minimum, Sample maximum,
451                                       size_t bucket_count);
452
453  virtual bool HasConstructorTimeDeltaArguments(TimeDelta minimum,
454                                                TimeDelta maximum,
455                                                size_t bucket_count);
456  // Return true iff the range_checksum_ matches current ranges_ vector.
457  bool HasValidRangeChecksum() const;
458
459 protected:
460  Histogram(const std::string& name, Sample minimum,
461            Sample maximum, size_t bucket_count);
462  Histogram(const std::string& name, TimeDelta minimum,
463            TimeDelta maximum, size_t bucket_count);
464
465  virtual ~Histogram();
466
467  // Initialize ranges_ mapping.
468  void InitializeBucketRange();
469
470  // Method to override to skip the display of the i'th bucket if it's empty.
471  virtual bool PrintEmptyBucket(size_t index) const;
472
473  //----------------------------------------------------------------------------
474  // Methods to override to create histogram with different bucket widths.
475  //----------------------------------------------------------------------------
476  // Find bucket to increment for sample value.
477  virtual size_t BucketIndex(Sample value) const;
478  // Get normalized size, relative to the ranges_[i].
479  virtual double GetBucketSize(Count current, size_t i) const;
480
481  // Recalculate range_checksum_.
482  void ResetRangeChecksum();
483
484  // Return a string description of what goes in a given bucket.
485  // Most commonly this is the numeric value, but in derived classes it may
486  // be a name (or string description) given to the bucket.
487  virtual const std::string GetAsciiBucketRange(size_t it) const;
488
489  //----------------------------------------------------------------------------
490  // Methods to override to create thread safe histogram.
491  //----------------------------------------------------------------------------
492  // Update all our internal data, including histogram
493  virtual void Accumulate(Sample value, Count count, size_t index);
494
495  //----------------------------------------------------------------------------
496  // Accessors for derived classes.
497  //----------------------------------------------------------------------------
498  void SetBucketRange(size_t i, Sample value);
499
500  // Validate that ranges_ was created sensibly (top and bottom range
501  // values relate properly to the declared_min_ and declared_max_)..
502  bool ValidateBucketRanges() const;
503
504  virtual uint32 CalculateRangeChecksum() const;
505
506 private:
507  // Allow tests to corrupt our innards for testing purposes.
508  FRIEND_TEST(HistogramTest, CorruptBucketBounds);
509  FRIEND_TEST(HistogramTest, CorruptSampleCounts);
510  FRIEND_TEST(HistogramTest, Crc32SampleHash);
511  FRIEND_TEST(HistogramTest, Crc32TableTest);
512
513  friend class StatisticsRecorder;  // To allow it to delete duplicates.
514
515  // Post constructor initialization.
516  void Initialize();
517
518  // Checksum function for accumulating range values into a checksum.
519  static uint32 Crc32(uint32 sum, Sample range);
520
521  //----------------------------------------------------------------------------
522  // Helpers for emitting Ascii graphic.  Each method appends data to output.
523
524  // Find out how large the (graphically) the largest bucket will appear to be.
525  double GetPeakBucketSize(const SampleSet& snapshot) const;
526
527  // Write a common header message describing this histogram.
528  void WriteAsciiHeader(const SampleSet& snapshot,
529                        Count sample_count, std::string* output) const;
530
531  // Write information about previous, current, and next buckets.
532  // Information such as cumulative percentage, etc.
533  void WriteAsciiBucketContext(const int64 past, const Count current,
534                               const int64 remaining, const size_t i,
535                               std::string* output) const;
536
537  // Write textual description of the bucket contents (relative to histogram).
538  // Output is the count in the buckets, as well as the percentage.
539  void WriteAsciiBucketValue(Count current, double scaled_sum,
540                             std::string* output) const;
541
542  // Produce actual graph (set of blank vs non blank char's) for a bucket.
543  void WriteAsciiBucketGraph(double current_size, double max_size,
544                             std::string* output) const;
545
546  //----------------------------------------------------------------------------
547  // Table for generating Crc32 values.
548  static const uint32 kCrcTable[256];
549  //----------------------------------------------------------------------------
550  // Invariant values set at/near construction time
551
552  // ASCII version of original name given to the constructor.  All identically
553  // named instances will be coalesced cross-project.
554  const std::string histogram_name_;
555  Sample declared_min_;  // Less than this goes into counts_[0]
556  Sample declared_max_;  // Over this goes into counts_[bucket_count_ - 1].
557  size_t bucket_count_;  // Dimension of counts_[].
558
559  // Flag the histogram for recording by UMA via metric_services.h.
560  Flags flags_;
561
562  // For each index, show the least value that can be stored in the
563  // corresponding bucket. We also append one extra element in this array,
564  // containing kSampleType_MAX, to make calculations easy.
565  // The dimension of ranges_ is bucket_count + 1.
566  Ranges ranges_;
567
568  // For redundancy, we store a checksum of all the sample ranges when ranges
569  // are generated.  If ever there is ever a difference, then the histogram must
570  // have been corrupted.
571  uint32 range_checksum_;
572
573  // Finally, provide the state that changes with the addition of each new
574  // sample.
575  SampleSet sample_;
576
577  DISALLOW_COPY_AND_ASSIGN(Histogram);
578};
579
580//------------------------------------------------------------------------------
581
582// LinearHistogram is a more traditional histogram, with evenly spaced
583// buckets.
584class BASE_API LinearHistogram : public Histogram {
585 public:
586  virtual ~LinearHistogram();
587
588  /* minimum should start from 1. 0 is as minimum is invalid. 0 is an implicit
589     default underflow bucket. */
590  static Histogram* FactoryGet(const std::string& name,
591                               Sample minimum,
592                               Sample maximum,
593                               size_t bucket_count,
594                               Flags flags);
595  static Histogram* FactoryTimeGet(const std::string& name,
596                                   TimeDelta minimum,
597                                   TimeDelta maximum,
598                                   size_t bucket_count,
599                                   Flags flags);
600
601  // Overridden from Histogram:
602  virtual ClassType histogram_type() const;
603
604  // Store a list of number/text values for use in rendering the histogram.
605  // The last element in the array has a null in its "description" slot.
606  virtual void SetRangeDescriptions(const DescriptionPair descriptions[]);
607
608 protected:
609  LinearHistogram(const std::string& name, Sample minimum,
610                  Sample maximum, size_t bucket_count);
611
612  LinearHistogram(const std::string& name, TimeDelta minimum,
613                  TimeDelta maximum, size_t bucket_count);
614
615  // Initialize ranges_ mapping.
616  void InitializeBucketRange();
617  virtual double GetBucketSize(Count current, size_t i) const;
618
619  // If we have a description for a bucket, then return that.  Otherwise
620  // let parent class provide a (numeric) description.
621  virtual const std::string GetAsciiBucketRange(size_t i) const;
622
623  // Skip printing of name for numeric range if we have a name (and if this is
624  // an empty bucket).
625  virtual bool PrintEmptyBucket(size_t index) const;
626
627 private:
628  // For some ranges, we store a printable description of a bucket range.
629  // If there is no desciption, then GetAsciiBucketRange() uses parent class
630  // to provide a description.
631  typedef std::map<Sample, std::string> BucketDescriptionMap;
632  BucketDescriptionMap bucket_description_;
633
634  DISALLOW_COPY_AND_ASSIGN(LinearHistogram);
635};
636
637//------------------------------------------------------------------------------
638
639// BooleanHistogram is a histogram for booleans.
640class BASE_API BooleanHistogram : public LinearHistogram {
641 public:
642  static Histogram* FactoryGet(const std::string& name, Flags flags);
643
644  virtual ClassType histogram_type() const;
645
646  virtual void AddBoolean(bool value);
647
648 private:
649  explicit BooleanHistogram(const std::string& name);
650
651  DISALLOW_COPY_AND_ASSIGN(BooleanHistogram);
652};
653
654//------------------------------------------------------------------------------
655
656// CustomHistogram is a histogram for a set of custom integers.
657class BASE_API CustomHistogram : public Histogram {
658 public:
659
660  static Histogram* FactoryGet(const std::string& name,
661                               const std::vector<Sample>& custom_ranges,
662                               Flags flags);
663
664  // Overridden from Histogram:
665  virtual ClassType histogram_type() const;
666
667 protected:
668  CustomHistogram(const std::string& name,
669                  const std::vector<Sample>& custom_ranges);
670
671  // Initialize ranges_ mapping.
672  void InitializedCustomBucketRange(const std::vector<Sample>& custom_ranges);
673  virtual double GetBucketSize(Count current, size_t i) const;
674
675  DISALLOW_COPY_AND_ASSIGN(CustomHistogram);
676};
677
678//------------------------------------------------------------------------------
679// StatisticsRecorder handles all histograms in the system.  It provides a
680// general place for histograms to register, and supports a global API for
681// accessing (i.e., dumping, or graphing) the data in all the histograms.
682
683class BASE_API StatisticsRecorder {
684 public:
685  typedef std::vector<Histogram*> Histograms;
686
687  StatisticsRecorder();
688
689  ~StatisticsRecorder();
690
691  // Find out if histograms can now be registered into our list.
692  static bool IsActive();
693
694  // Register, or add a new histogram to the collection of statistics. If an
695  // identically named histogram is already registered, then the argument
696  // |histogram| will deleted.  The returned value is always the registered
697  // histogram (either the argument, or the pre-existing registered histogram).
698  static Histogram* RegisterOrDeleteDuplicate(Histogram* histogram);
699
700  // Methods for printing histograms.  Only histograms which have query as
701  // a substring are written to output (an empty string will process all
702  // registered histograms).
703  static void WriteHTMLGraph(const std::string& query, std::string* output);
704  static void WriteGraph(const std::string& query, std::string* output);
705
706  // Method for extracting histograms which were marked for use by UMA.
707  static void GetHistograms(Histograms* output);
708
709  // Find a histogram by name. It matches the exact name. This method is thread
710  // safe.  If a matching histogram is not found, then the |histogram| is
711  // not changed.
712  static bool FindHistogram(const std::string& query, Histogram** histogram);
713
714  static bool dump_on_exit() { return dump_on_exit_; }
715
716  static void set_dump_on_exit(bool enable) { dump_on_exit_ = enable; }
717
718  // GetSnapshot copies some of the pointers to registered histograms into the
719  // caller supplied vector (Histograms).  Only histograms with names matching
720  // query are returned. The query must be a substring of histogram name for its
721  // pointer to be copied.
722  static void GetSnapshot(const std::string& query, Histograms* snapshot);
723
724
725 private:
726  // We keep all registered histograms in a map, from name to histogram.
727  typedef std::map<std::string, Histogram*> HistogramMap;
728
729  static HistogramMap* histograms_;
730
731  // lock protects access to the above map.
732  static base::Lock* lock_;
733
734  // Dump all known histograms to log.
735  static bool dump_on_exit_;
736
737  DISALLOW_COPY_AND_ASSIGN(StatisticsRecorder);
738};
739
740}  // namespace base
741
742#endif  // BASE_METRICS_HISTOGRAM_H_
743