Skip to content

Commit 6f8be0f

Browse files
refactor: update gammainc implementation to follow Boost v1.88.0
PR-URL: #7619 Reviewed-by: Athan Reines <kgryte@gmail.com>
1 parent 351de2d commit 6f8be0f

19 files changed

+568
-249
lines changed

lib/node_modules/@stdlib/math/base/special/gammainc/docs/types/index.d.ts

Lines changed: 11 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -33,22 +33,27 @@
3333
* @returns function value
3434
*
3535
* @example
36-
* var y = gammainc( 6.0, 2.0 )
36+
* var y = gammainc( 6.0, 2.0 );
3737
* // returns ~0.9826
38+
*
3839
* @example
39-
* var y = gammainc( 1.0, 2.0, true, true )
40+
* var y = gammainc( 1.0, 2.0, true, true );
4041
* // returns ~0.7358
42+
*
4143
* @example
42-
* var y = gammainc( 7.0, 5.0 )
44+
* var y = gammainc( 7.0, 5.0 );
4345
* // returns ~0.8270
46+
*
4447
* @example
45-
* var y = gammainc( 7.0, 5.0, false )
48+
* var y = gammainc( 7.0, 5.0, false );
4649
* // returns ~19.8482
50+
*
4751
* @example
48-
* var y = gammainc( NaN, 2.0 )
52+
* var y = gammainc( NaN, 2.0 );
4953
* // returns NaN
54+
*
5055
* @example
51-
* var y = gammainc( 6.0, NaN )
56+
* var y = gammainc( 6.0, NaN );
5257
* // returns NaN
5358
*/
5459
declare function gammainc( x: number, a: number, regularized?: boolean, upper?: boolean ): number;

lib/node_modules/@stdlib/math/base/special/gammainc/lib/finite_gamma_q.js

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -18,7 +18,7 @@
1818
*
1919
* ## Notice
2020
*
21-
* The original C++ code and copyright notice are from the [Boost library]{@link http://www.boost.org/doc/libs/1_37_0/boost/math/special_functions/gamma.hpp}. The implementation has been modified for JavaScript.
21+
* The original C++ code and copyright notice are from the [Boost library]{@link https://www.boost.org/doc/libs/1_88_0/boost/math/special_functions/gamma.hpp}. The implementation has been modified for JavaScript.
2222
*
2323
* ```text
2424
* (C) Copyright John Maddock 2006.

lib/node_modules/@stdlib/math/base/special/gammainc/lib/finite_half_gamma_q.js

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -18,7 +18,7 @@
1818
*
1919
* ## Notice
2020
*
21-
* The original C++ code and copyright notice are from the [Boost library]{@link http://www.boost.org/doc/libs/1_37_0/boost/math/special_functions/gamma.hpp}. The implementation has been modified for JavaScript.
21+
* The original C++ code and copyright notice are from the [Boost library]{@link https://www.boost.org/doc/libs/1_88_0/boost/math/special_functions/gamma.hpp}. The implementation has been modified for JavaScript.
2222
*
2323
* ```text
2424
* (C) Copyright John Maddock 2006.

lib/node_modules/@stdlib/math/base/special/gammainc/lib/full_igamma_prefix.js

Lines changed: 5 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -18,7 +18,7 @@
1818
*
1919
* ## Notice
2020
*
21-
* The original C++ code and copyright notice are from the [Boost library]{@link http://www.boost.org/doc/libs/1_37_0/boost/math/special_functions/gamma.hpp}. The implementation has been modified for JavaScript.
21+
* The original C++ code and copyright notice are from the [Boost library]{@link https://www.boost.org/doc/libs/1_88_0/boost/math/special_functions/gamma.hpp}. The implementation has been modified for JavaScript.
2222
*
2323
* ```text
2424
* (C) Copyright John Maddock 2006.
@@ -59,20 +59,16 @@ function fullIGammaPrefix( a, z ) {
5959
if ( z >= 1.0 ) {
6060
if ( ( alz < MAX_LN ) && ( -z > MIN_LN ) ) {
6161
prefix = pow( z, a ) * exp( -z );
62-
}
63-
else if ( a >= 1.0 ) {
62+
} else if ( a >= 1.0 ) {
6463
prefix = pow( z / exp(z/a), a );
65-
}
66-
else {
64+
} else {
6765
prefix = exp( alz - z );
6866
}
69-
}
70-
else {
67+
} else {
7168
/* eslint-disable no-lonely-if */
7269
if ( alz > MIN_LN ) {
7370
prefix = pow( z, a ) * exp( -z );
74-
}
75-
else if ( z/a < MAX_LN ) {
71+
} else if ( z/a < MAX_LN ) {
7672
prefix = pow( z / exp(z/a), a );
7773
} else {
7874
prefix = exp( alz - z );
Lines changed: 258 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,258 @@
1+
/**
2+
* @license Apache-2.0
3+
*
4+
* Copyright (c) 2025 The Stdlib Authors.
5+
*
6+
* Licensed under the Apache License, Version 2.0 (the "License");
7+
* you may not use this file except in compliance with the License.
8+
* You may obtain a copy of the License at
9+
*
10+
* http://www.apache.org/licenses/LICENSE-2.0
11+
*
12+
* Unless required by applicable law or agreed to in writing, software
13+
* distributed under the License is distributed on an "AS IS" BASIS,
14+
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
15+
* See the License for the specific language governing permissions and
16+
* limitations under the License.
17+
*
18+
*
19+
* ## Notice
20+
*
21+
* The original C++ code and copyright notice are from the [Boost library]{@link https://www.boost.org/doc/libs/1_88_0/boost/math/special_functions/gamma.hpp}. The implementation has been modified for JavaScript.
22+
*
23+
* ```text
24+
* (C) Copyright John Maddock 2006.
25+
* (C) Copyright Paul A. Bristow 2007.
26+
*
27+
* Use, modification and distribution are subject to the
28+
* Boost Software License, Version 1.0. (See accompanying file
29+
* LICENSE or copy at http://www.boost.org/LICENSE_1_0.txt)
30+
* ```
31+
*/
32+
33+
'use strict';
34+
35+
// MODULES //
36+
37+
var floor = require( '@stdlib/math/base/special/floor' );
38+
var gamma = require( '@stdlib/math/base/special/gamma' );
39+
var abs = require( '@stdlib/math/base/special/abs' );
40+
var pow = require( '@stdlib/math/base/special/pow' );
41+
var ln = require( '@stdlib/math/base/special/ln' );
42+
var SQRT_EPSILON = require( '@stdlib/constants/float64/sqrt-eps' );
43+
var FLOAT64_MAX = require( '@stdlib/constants/float64/max' );
44+
var MAX_LN = require( '@stdlib/constants/float64/max-ln' );
45+
var tgammaILargeX = require( './tgamma_i_large_x.js' );
46+
var finiteGammaQ = require( './finite_gamma_q.js' );
47+
var finiteHalfGammaQ = require( './finite_half_gamma_q.js' );
48+
var fullIGammaPrefix = require( './full_igamma_prefix.js' );
49+
var igammaTemmeLarge = require( './igamma_temme_large.js' );
50+
var lowerGammaSeries = require( './lower_gamma_series.js' );
51+
var regularisedGammaPrefix = require( './regularised_gamma_prefix.js' );
52+
var tgammaSmallUpperPart = require( './tgamma_small_upper_part.js' );
53+
var upperGammaFraction = require( './upper_gamma_fraction.js' );
54+
55+
56+
// MAIN //
57+
58+
/**
59+
* Main entry point for evaluating all the four incomplete gamma functions (lower, upper, regularized lower, and regularized upper).
60+
*
61+
* @private
62+
* @param {number} x - function parameter
63+
* @param {number} a - function parameter
64+
* @param {boolean} regularized - boolean indicating if the function should evaluate the regularized or non-regularized incomplete gamma functions
65+
* @param {boolean} upper - boolean indicating if the function should return the upper tail of the incomplete gamma function
66+
* @returns {number} function value
67+
*/
68+
function igammaFinal( x, a, regularized, upper ) {
69+
var optimisedInvert;
70+
var evalMethod;
71+
var isHalfInt;
72+
var initValue;
73+
var useTemme;
74+
var isSmallA;
75+
var result;
76+
var invert;
77+
var isInt;
78+
var sigma;
79+
var res;
80+
var gam;
81+
var fa;
82+
var g;
83+
84+
result = 0.0;
85+
invert = upper;
86+
isSmallA = ( a < 30 ) && ( a <= x + 1.0 ) && ( x < MAX_LN );
87+
if ( isSmallA ) {
88+
fa = floor( a );
89+
isInt = ( fa === a );
90+
isHalfInt = ( isInt ) ? false : ( abs( fa - a ) === 0.5 );
91+
} else {
92+
isInt = false;
93+
isHalfInt = false;
94+
}
95+
if ( isInt && x > 0.6 ) {
96+
// Calculate Q via finite sum:
97+
invert = !invert;
98+
evalMethod = 0;
99+
} else if ( isHalfInt && x > 0.2 ) {
100+
// Calculate Q via finite sum for half integer a:
101+
invert = !invert;
102+
evalMethod = 1;
103+
} else if ( x < SQRT_EPSILON && a > 1.0 ) {
104+
evalMethod = 6;
105+
} else if ( ( x > 1000.0 ) && ( ( a < x ) || ( abs( a - 50.0 ) / x < 1.0 ) ) ) {
106+
// Calculate Q via asymptotic approximation:
107+
invert = !invert;
108+
evalMethod = 7;
109+
} else if ( x < 0.5 ) {
110+
// Changeover criterion chosen to give a changeover at Q ~ 0.33:
111+
if ( -0.4 / ln( x ) < a ) {
112+
evalMethod = 2;
113+
} else {
114+
evalMethod = 3;
115+
}
116+
} else if ( x < 1.1 ) {
117+
// Changeover here occurs when P ~ 0.75 or Q ~ 0.25:
118+
if ( x * 0.75 < a ) {
119+
evalMethod = 2;
120+
} else {
121+
evalMethod = 3;
122+
}
123+
} else {
124+
// Begin by testing whether we're in the "bad" zone where the result will be near 0.5 and the usual series and continued fractions are slow to converge:
125+
useTemme = false;
126+
if ( regularized && a > 20 ) {
127+
sigma = abs( (x-a)/a );
128+
if ( a > 200 ) {
129+
// Limit chosen so that we use Temme's expansion only if the result would be larger than about 10^-6. Below that the regular series and continued fractions converge OK, and if we use Temme's method we get increasing errors from the dominant erfc term as it's (inexact) argument increases in magnitude.
130+
if ( 20 / a > sigma * sigma ) {
131+
useTemme = true;
132+
}
133+
} else if ( sigma < 0.4 ) {
134+
useTemme = true;
135+
}
136+
}
137+
if ( useTemme ) {
138+
evalMethod = 5;
139+
}
140+
// Regular case where the result will not be too close to 0.5: Changeover occurs at P ~ Q ~ 0.5. Note that series computation of P is about x2 faster than continued fraction calculation of Q, so try and use the CF only when really necessary, especially for small x.
141+
else if ( x - ( 1.0 / (3.0 * x) ) < a ) {
142+
evalMethod = 2;
143+
} else {
144+
evalMethod = 4;
145+
invert = !invert;
146+
}
147+
}
148+
149+
/* eslint-disable default-case */
150+
switch ( evalMethod ) {
151+
case 0:
152+
result = finiteGammaQ( a, x );
153+
if ( regularized === false ) {
154+
result *= gamma( a );
155+
}
156+
break;
157+
case 1:
158+
result = finiteHalfGammaQ( a, x );
159+
if ( regularized === false ) {
160+
result *= gamma( a );
161+
}
162+
break;
163+
case 2:
164+
// Compute P:
165+
result = ( regularized ) ?
166+
regularisedGammaPrefix( a, x ) :
167+
fullIGammaPrefix( a, x );
168+
if ( result !== 0.0 ) {
169+
// If the result will be inverted, we can potentially save computation by initializing the series sum closer to the final result. This reduces the number of iterations needed in the series evaluation. However, care must be taken to avoid spurious numeric overflows. In practice, the more expensive overflow checks below are typically bypassed for well-behaved input values.
170+
initValue = 0.0;
171+
optimisedInvert = false;
172+
if ( invert ) {
173+
initValue = ( regularized ) ? 1.0 : gamma( a );
174+
if (
175+
regularized ||
176+
result >= 1.0 ||
177+
FLOAT64_MAX * result > initValue
178+
) {
179+
initValue /= result;
180+
if (
181+
regularized ||
182+
a < 1.0 ||
183+
( FLOAT64_MAX / a > initValue )
184+
) {
185+
initValue *= -a;
186+
optimisedInvert = true;
187+
} else {
188+
initValue = 0.0;
189+
}
190+
} else {
191+
initValue = 0.0;
192+
}
193+
}
194+
result *= lowerGammaSeries( a, x, initValue ) / a;
195+
if ( optimisedInvert ) {
196+
invert = false;
197+
result = -result;
198+
}
199+
}
200+
break;
201+
case 3:
202+
// Compute Q:
203+
invert = !invert;
204+
res = tgammaSmallUpperPart( a, x, invert );
205+
result = res[ 0 ];
206+
g = res[ 1 ];
207+
invert = false;
208+
if ( regularized ) {
209+
result /= g;
210+
}
211+
break;
212+
case 4:
213+
// Compute Q:
214+
result = ( regularized ) ?
215+
regularisedGammaPrefix( a, x ) :
216+
fullIGammaPrefix( a, x );
217+
if ( result !== 0 ) {
218+
result *= upperGammaFraction( a, x );
219+
}
220+
break;
221+
case 5:
222+
result = igammaTemmeLarge( a, x );
223+
if ( x >= a ) {
224+
invert = !invert;
225+
}
226+
break;
227+
case 6:
228+
// Since `x` is so small that P is necessarily very small too, use http://functions.wolfram.com/GammaBetaErf/GammaRegularized/06/01/05/01/01/
229+
result = ( regularized ) ?
230+
pow( x, a ) / gamma( a + 1.0 ) :
231+
pow( x, a ) / a;
232+
result *= 1.0 - ( a * x / ( a + 1.0 ) );
233+
break;
234+
case 7:
235+
// `x` is large, so compute Q:
236+
result = ( regularized ) ?
237+
regularisedGammaPrefix( a, x ) :
238+
fullIGammaPrefix( a, x );
239+
result /= x;
240+
if ( result !== 0.0 ) {
241+
result *= tgammaILargeX( a, x );
242+
}
243+
break;
244+
}
245+
if ( regularized && result > 1.0 ) {
246+
result = 1.0;
247+
}
248+
if ( invert ) {
249+
gam = ( regularized ) ? 1.0 : gamma( a );
250+
result = gam - result;
251+
}
252+
return result;
253+
}
254+
255+
256+
// EXPORTS //
257+
258+
module.exports = igammaFinal;

lib/node_modules/@stdlib/math/base/special/gammainc/lib/igamma_temme_large.js

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -18,7 +18,7 @@
1818
*
1919
* ## Notice
2020
*
21-
* The original C++ code and copyright notice are from the [Boost library]{@link http://www.boost.org/doc/libs/1_37_0/boost/math/special_functions/gamma.hpp}. The implementation has been modified for JavaScript.
21+
* The original C++ code and copyright notice are from the [Boost library]{@link https://www.boost.org/doc/libs/1_88_0/boost/math/special_functions/gamma.hpp}. The implementation has been modified for JavaScript.
2222
*
2323
* ```text
2424
* (C) Copyright John Maddock 2006.

lib/node_modules/@stdlib/math/base/special/gammainc/lib/lower_gamma_series.js

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -18,7 +18,7 @@
1818
*
1919
* ## Notice
2020
*
21-
* The original C++ code and copyright notice are from the [Boost library]{@link http://www.boost.org/doc/libs/1_37_0/boost/math/special_functions/gamma.hpp}. The implementation has been modified for JavaScript.
21+
* The original C++ code and copyright notice are from the [Boost library]{@link https://www.boost.org/doc/libs/1_88_0/boost/math/special_functions/gamma.hpp}. The implementation has been modified for JavaScript.
2222
*
2323
* ```text
2424
* (C) Copyright John Maddock 2006.
@@ -46,7 +46,7 @@ var lowerIncompleteGammaSeries = require( './lower_incomplete_gamma_series.js' )
4646
* ## Method
4747
*
4848
* - Multiply result by `((z^a) * (e^-z) / a)` to get the full lower incomplete integral.
49-
* - Divide by `tgamma(a)` to get the normalized value.
49+
* - Divide by `gamma(a)` to get the normalized value.
5050
*
5151
* @private
5252
* @param {number} a - function parameter

lib/node_modules/@stdlib/math/base/special/gammainc/lib/lower_incomplete_gamma_series.js

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -18,7 +18,7 @@
1818
*
1919
* ## Notice
2020
*
21-
* The original C++ code and copyright notice are from the [Boost library]{@link http://www.boost.org/doc/libs/1_37_0/boost/math/special_functions/gamma.hpp}. The implementation has been modified for JavaScript.
21+
* The original C++ code and copyright notice are from the [Boost library]{@link https://www.boost.org/doc/libs/1_88_0/boost/math/special_functions/gamma.hpp}. The implementation has been modified for JavaScript.
2222
*
2323
* ```text
2424
* (C) Copyright John Maddock 2006.

0 commit comments

Comments
 (0)