catapult/tracing/tracing/base/running_statistics.html - platform/external/chromium-trace - Git at Google

 <!DOCTYPE html>
 <!--
 Copyright 2016 The Chromium Authors. All rights reserved.
 Use of this source code is governed by a BSD-style license that can be
 found in the LICENSE file.
 -->

 <link rel="import" href="/tracing/base/base.html">

 <script>
 'use strict';

 tr.exportTo('tr.b', function() {
   /***
   * An object of this class computes basic statistics online in O(1).
   * Usage:
   * 1. Create an instance.
   * 2. Add numbers using the |add| method.
   * 3. Query statistics.
   * 4. Repeat from step 2.
   */
   class RunningStatistics {
     constructor() {
       this.mean_ = 0;
       this.count_ = 0;
       this.max_ = -Infinity;
       this.min_ = Infinity;
       this.sum_ = 0;
       this.variance_ = 0;

       // Mean of logarithms of absolute values of samples, or undefined if any
       // samples were <= 0.
       this.meanlogs_ = 0;
     }

     get count() {
       return this.count_;
     }

     get geometricMean() {
       if (this.meanlogs_ === undefined)
         return 0;
       return Math.exp(this.meanlogs_);
     }

     get mean() {
       if (this.count_ == 0)
         return undefined;
       return this.mean_;
     }

     get max() {
       return this.max_;
     }

     get min() {
       return this.min_;
     }

     get sum() {
       return this.sum_;
     }

     get variance() {
       if (this.count_ == 0)
         return undefined;
       if (this.count_ == 1)
         return 0;
       return this.variance_ / (this.count_ - 1);
     }

     get stddev() {
       if (this.count_ == 0)
         return undefined;
       return Math.sqrt(this.variance);
     }

     add(x) {
       this.count_++;
       this.max_ = Math.max(this.max_, x);
       this.min_ = Math.min(this.min_, x);
       this.sum_ += x;

       // The geometric mean is computed using the arithmetic mean of logarithms.
       if (x <= 0)
         this.meanlogs_ = undefined;
       else if (this.meanlogs_ !== undefined)
         this.meanlogs_ += (Math.log(Math.abs(x)) - this.meanlogs_) / this.count;

       // The following uses Welford's algorithm for computing running mean
       // and variance. See http://www.johndcook.com/blog/standard_deviation.
       if (this.count_ === 1) {
         this.mean_ = x;
         this.variance_ = 0;
       } else {
         var oldMean = this.mean_;
         var oldVariance = this.variance_;
         // Using the 2nd formula for updating the mean yields better precision
         // but it doesn't work for the case oldMean is Infinity. Hence we handle
         // that case separately.
         if (oldMean === Infinity || oldMean === -Infinity) {
           this.mean_ = this.sum_ / this.count_;
         } else {
           this.mean_ = oldMean + (x - oldMean) / this.count_;
         }
         this.variance_ = oldVariance + (x - oldMean) * (x - this.mean_);
       }
     }

     merge(other) {
       var result = new RunningStatistics();
       result.count_ = this.count_ + other.count_;
       result.sum_ = this.sum_ + other.sum_;
       result.min_ = Math.min(this.min_, other.min_);
       result.max_ = Math.max(this.max_, other.max_);
       if (result.count === 0) {
         result.mean_ = 0;
         result.variance_ = 0;
         result.meanlogs_ = 0;
       } else {
         // Combine the mean and the variance using the formulas from
         // https://goo.gl/ddcAep.
         result.mean_ = result.sum / result.count;
         var deltaMean = (this.mean || 0) - (other.mean || 0);
         result.variance_ = this.variance_ + other.variance_ +
           (this.count * other.count * deltaMean * deltaMean / result.count);

         // Merge the arithmetic means of logarithms of absolute values of
         // samples, weighted by counts.
         if (this.meanlogs_ === undefined || other.meanlogs_ === undefined) {
           result.meanlogs_ = undefined;
         } else {
           result.meanlogs_ = (this.count * this.meanlogs_ +
               other.count * other.meanlogs_) / result.count;
         }
       }
       return result;
     }

     asDict() {
       return {
         mean: this.mean,
         meanlogs: this.meanlogs_,
         count: this.count,
         max: this.max,
         min: this.min,
         sum: this.sum,
         variance: this.variance_
       };
     }

     static fromDict(d) {
       var result = new RunningStatistics();
       result.mean_ = d.mean;
       result.count_ = d.count;
       result.max_ = d.max;
       result.min_ = d.min;
       result.sum_ = d.sum;
       result.variance_ = d.variance;
       result.meanlogs_ = d.meanlogs;
       return result;
     }
   }

   return {
     RunningStatistics: RunningStatistics
   };
 });
 </script>
	<!DOCTYPE html>
	<!--
	Copyright 2016 The Chromium Authors. All rights reserved.
	Use of this source code is governed by a BSD-style license that can be
	found in the LICENSE file.
	-->

	<link rel="import" href="/tracing/base/base.html">

	<script>
	'use strict';

	tr.exportTo('tr.b', function() {
	/***
	* An object of this class computes basic statistics online in O(1).
	* Usage:
	* 1. Create an instance.
	* 2. Add numbers using the \|add\| method.
	* 3. Query statistics.
	* 4. Repeat from step 2.
	*/
	class RunningStatistics {
	constructor() {
	this.mean_ = 0;
	this.count_ = 0;
	this.max_ = -Infinity;
	this.min_ = Infinity;
	this.sum_ = 0;
	this.variance_ = 0;

	// Mean of logarithms of absolute values of samples, or undefined if any
	// samples were <= 0.
	this.meanlogs_ = 0;
	}

	get count() {
	return this.count_;
	}

	get geometricMean() {
	if (this.meanlogs_ === undefined)
	return 0;
	return Math.exp(this.meanlogs_);
	}

	get mean() {
	if (this.count_ == 0)
	return undefined;
	return this.mean_;
	}

	get max() {
	return this.max_;
	}

	get min() {
	return this.min_;
	}

	get sum() {
	return this.sum_;
	}

	get variance() {
	if (this.count_ == 0)
	return undefined;
	if (this.count_ == 1)
	return 0;
	return this.variance_ / (this.count_ - 1);
	}

	get stddev() {
	if (this.count_ == 0)
	return undefined;
	return Math.sqrt(this.variance);
	}

	add(x) {
	this.count_++;
	this.max_ = Math.max(this.max_, x);
	this.min_ = Math.min(this.min_, x);
	this.sum_ += x;

	// The geometric mean is computed using the arithmetic mean of logarithms.
	if (x <= 0)
	this.meanlogs_ = undefined;
	else if (this.meanlogs_ !== undefined)
	this.meanlogs_ += (Math.log(Math.abs(x)) - this.meanlogs_) / this.count;

	// The following uses Welford's algorithm for computing running mean
	// and variance. See http://www.johndcook.com/blog/standard_deviation.
	if (this.count_ === 1) {
	this.mean_ = x;
	this.variance_ = 0;
	} else {
	var oldMean = this.mean_;
	var oldVariance = this.variance_;
	// Using the 2nd formula for updating the mean yields better precision
	// but it doesn't work for the case oldMean is Infinity. Hence we handle
	// that case separately.
	if (oldMean === Infinity \|\| oldMean === -Infinity) {
	this.mean_ = this.sum_ / this.count_;
	} else {
	this.mean_ = oldMean + (x - oldMean) / this.count_;
	}
	this.variance_ = oldVariance + (x - oldMean) * (x - this.mean_);
	}
	}

	merge(other) {
	var result = new RunningStatistics();
	result.count_ = this.count_ + other.count_;
	result.sum_ = this.sum_ + other.sum_;
	result.min_ = Math.min(this.min_, other.min_);
	result.max_ = Math.max(this.max_, other.max_);
	if (result.count === 0) {
	result.mean_ = 0;
	result.variance_ = 0;
	result.meanlogs_ = 0;
	} else {
	// Combine the mean and the variance using the formulas from
	// https://goo.gl/ddcAep.
	result.mean_ = result.sum / result.count;
	var deltaMean = (this.mean \|\| 0) - (other.mean \|\| 0);
	result.variance_ = this.variance_ + other.variance_ +
	(this.count * other.count * deltaMean * deltaMean / result.count);

	// Merge the arithmetic means of logarithms of absolute values of
	// samples, weighted by counts.
	if (this.meanlogs_ === undefined \|\| other.meanlogs_ === undefined) {
	result.meanlogs_ = undefined;
	} else {
	result.meanlogs_ = (this.count * this.meanlogs_ +
	other.count * other.meanlogs_) / result.count;
	}
	}
	return result;
	}

	asDict() {
	return {
	mean: this.mean,
	meanlogs: this.meanlogs_,
	count: this.count,
	max: this.max,
	min: this.min,
	sum: this.sum,
	variance: this.variance_
	};
	}

	static fromDict(d) {
	var result = new RunningStatistics();
	result.mean_ = d.mean;
	result.count_ = d.count;
	result.max_ = d.max;
	result.min_ = d.min;
	result.sum_ = d.sum;
	result.variance_ = d.variance;
	result.meanlogs_ = d.meanlogs;
	return result;
	}
	}

	return {
	RunningStatistics: RunningStatistics
	};
	});
	</script>