1## Changes in 0.3.4.2
2
3  * Fixed crash in `invIncompleteBeta` (#68) for some inputs initial approximation
4    was computed incorrectly.
5
6
7## Changes in 0.3.4.1
8
9  * Precision of `incompleteGamma` improved.
10
11
12## Changes in 0.3.4.0
13
14  * Dependency on `vector-th-unbox` is dropped. All instances are written by
15    hand now.
16
17
18## Changes in 0.3.3.0
19
20  * New implementation for `logGamma`. Now it's precise within 2 ulps at full
21    range including zeros at 1 and 2.
22
23  * Bug with precision loss of `invErf` for parameters near zero is fixed.
24
25  * Fixed loss of precision in bundled `erf` near zero. (Affect primarily GHCJS)
26
27  * `factorial` for now uses lookup table
28
29  * `logFactorial` is optimized a bit (less number of terms is used)
30
31  * `m_sqrt_eps` constant added.
32
33  * Module `Numeric.SpecFunctions.Internal` is exposed.
34
35  * Many improvements for test suite. Tables of expected function values are
36    generated with mpmath, error estimates improved. Test suite itself is
37    migrated to `tasty` from `test-framework`.
38
39
40## Changes in 0.3.2.1
41
42  * Fixes build on windows for GHC<8.0
43
44
45## Changes in 0.3.2.0
46
47  * GHCJS is now supported
48
49  * Flag `system-expm1` is set to true by default. Only affects GHC<8.0
50
51## Changes in 0.3.1.0
52
53  * Exported data types for iteration steps in root finding
54
55  * Defaults for root finding algorithm are documented
56
57
58## Changes in 0.3.0.2
59
60  * Fix license field in cabal file
61
62
63## Changes in 0.3.0.0
64
65  * `Semigroup` and `Monoid` instances added for data types from `Numeric.Sum`
66
67  * API for finding roots of real functions reworked. 1) All algorithm
68    parameters are now tweakable. 2) Functions for getting list of iterations
69    added.
70
71  * `Foldable` and `Traversable` instances for `Root` were added.
72
73## Changes in 0.2.1.0
74
75  * `log1p` and `expm1` are simply reexported from `GHC.Float`. They're methods
76     of `Floating` type class.
77
78  * On windows `expm1` is implemented in pure haskell for older GHCs.
79
80
81## Changes in 0.2.0.0
82
83  * Bug fixes and documentation tweaks
84
85
86## Changes in 0.2.0.0
87
88  * `logGamma` now uses Lancsoz approximation and same as `logGammaL`.  Old
89     implementation of `logGamma` moved to `Numeric.SpecFunctions.Extra.logGammaAS245`.
90
91  * Precision of `logGamma` for z<1 improved.
92
93  * New much more precise implementation for `incompleteGamma`
94
95  * Dependency on `erf` package dropped. `erf` and `erfc` just do direct calls
96    to C.
97
98  * `Numeric.SpecFunctions.expm1` added
99
100  * `Numeric.SpecFunctions.log1pmx` added.
101
102  * `logGammaCorrection` exported in `Numeric.SpecFunctions.Extra`.
103
104  * Module `Numeric.Series` added for working with infinite sequences, series
105    summation and evaluation of continued fractions.
106
107  * Module `statistics: Statistics.Math.RootFinding` copied to
108    `Numeric.RootFinding`. Instances for `binary` and `aeson` dropped.
109
110  * Root-finding using Newton-Raphson added
111
112  * `Numeric.MathFunctions.Comparison.ulpDelta` added. It calculates signed
113    distance between two doubles.
114
115  * Other bug fixes.
116
117
118
119## Changes in 0.1.7.0
120
121  * Module `statistics: Statistics.Function.Comparison` moved to
122    `Numeric.MathFunctions.Comparison`. Old implementation if `within` compared
123    negative numbers incorrectly.
124
125  * `addUlps` and `ulpDistance` added to `Numeric.MathFunctions.Comparison`.
126
127  * `relativeError` and `eqRelErr` added to `Numeric.MathFunctions.Comparison`.
128
129  * Precision of `logFactorial` is slightly improved.
130
131
132## Changes in 0.1.6.0
133
134  * `logChoose` added for calculation of logarithm of binomial coefficient
135
136  * `chooseExact` and `logChooseFast` added
137
138  * `sinc` added
139
140
141## Changes in 0.1.5.3
142
143  * Fix for test suite on 32bit platform
144
145
146## Changes in 0.1.5
147
148  * Numeric.Sum: new module adds accurate floating point summation.
149
150
151## Changes in 0.1.4
152
153  * logFactorial type is genberalized. It accepts any `Integral` type
154
155  * Evaluation of polynomials using Horner's method where coefficients
156    are store in lists added
157
158
159## Changes in 0.1.3
160
161  * Error function and its inverse added.
162
163  * Digamma function added
164
165  * Evaluation of polynomials using Horner's method added.
166
167  * Crash bug in the inverse incomplete beta fixed.
168