mirror of
https://git.savannah.gnu.org/git/emacs.git
synced 2024-11-27 07:37:33 +00:00
12f1951a27
Change all toplevel `setq' forms to `defvar' forms, and move them before their first use. Use `when', `unless'. Remove trailing periods from error forms. Add description and headers suggested by Emacs Lisp coding conventions.
976 lines
30 KiB
EmacsLisp
976 lines
30 KiB
EmacsLisp
;;; calc-funcs.el --- well-known functions for Calc
|
|
|
|
;; Copyright (C) 1990, 1991, 1992, 1993, 2001 Free Software Foundation, Inc.
|
|
|
|
;; Author: David Gillespie <daveg@synaptics.com>
|
|
;; Maintainer: Colin Walters <walters@debian.org>
|
|
|
|
;; This file is part of GNU Emacs.
|
|
|
|
;; GNU Emacs is distributed in the hope that it will be useful,
|
|
;; but WITHOUT ANY WARRANTY. No author or distributor
|
|
;; accepts responsibility to anyone for the consequences of using it
|
|
;; or for whether it serves any particular purpose or works at all,
|
|
;; unless he says so in writing. Refer to the GNU Emacs General Public
|
|
;; License for full details.
|
|
|
|
;; Everyone is granted permission to copy, modify and redistribute
|
|
;; GNU Emacs, but only under the conditions described in the
|
|
;; GNU Emacs General Public License. A copy of this license is
|
|
;; supposed to have been given to you along with GNU Emacs so you
|
|
;; can know your rights and responsibilities. It should be in a
|
|
;; file named COPYING. Among other things, the copyright notice
|
|
;; and this notice must be preserved on all copies.
|
|
|
|
;;; Commentary:
|
|
|
|
;;; Code:
|
|
|
|
;; This file is autoloaded from calc-ext.el.
|
|
(require 'calc-ext)
|
|
|
|
(require 'calc-macs)
|
|
|
|
(defun calc-Need-calc-funcs () nil)
|
|
|
|
|
|
(defun calc-inc-gamma (arg)
|
|
(interactive "P")
|
|
(calc-slow-wrapper
|
|
(if (calc-is-inverse)
|
|
(if (calc-is-hyperbolic)
|
|
(calc-binary-op "gamG" 'calcFunc-gammaG arg)
|
|
(calc-binary-op "gamQ" 'calcFunc-gammaQ arg))
|
|
(if (calc-is-hyperbolic)
|
|
(calc-binary-op "gamg" 'calcFunc-gammag arg)
|
|
(calc-binary-op "gamP" 'calcFunc-gammaP arg)))))
|
|
|
|
(defun calc-erf (arg)
|
|
(interactive "P")
|
|
(calc-slow-wrapper
|
|
(if (calc-is-inverse)
|
|
(calc-unary-op "erfc" 'calcFunc-erfc arg)
|
|
(calc-unary-op "erf" 'calcFunc-erf arg))))
|
|
|
|
(defun calc-erfc (arg)
|
|
(interactive "P")
|
|
(calc-invert-func)
|
|
(calc-erf arg))
|
|
|
|
(defun calc-beta (arg)
|
|
(interactive "P")
|
|
(calc-slow-wrapper
|
|
(calc-binary-op "beta" 'calcFunc-beta arg)))
|
|
|
|
(defun calc-inc-beta ()
|
|
(interactive)
|
|
(calc-slow-wrapper
|
|
(if (calc-is-hyperbolic)
|
|
(calc-enter-result 3 "betB" (cons 'calcFunc-betaB (calc-top-list-n 3)))
|
|
(calc-enter-result 3 "betI" (cons 'calcFunc-betaI (calc-top-list-n 3))))))
|
|
|
|
(defun calc-bessel-J (arg)
|
|
(interactive "P")
|
|
(calc-slow-wrapper
|
|
(calc-binary-op "besJ" 'calcFunc-besJ arg)))
|
|
|
|
(defun calc-bessel-Y (arg)
|
|
(interactive "P")
|
|
(calc-slow-wrapper
|
|
(calc-binary-op "besY" 'calcFunc-besY arg)))
|
|
|
|
(defun calc-bernoulli-number (arg)
|
|
(interactive "P")
|
|
(calc-slow-wrapper
|
|
(if (calc-is-hyperbolic)
|
|
(calc-binary-op "bern" 'calcFunc-bern arg)
|
|
(calc-unary-op "bern" 'calcFunc-bern arg))))
|
|
|
|
(defun calc-euler-number (arg)
|
|
(interactive "P")
|
|
(calc-slow-wrapper
|
|
(if (calc-is-hyperbolic)
|
|
(calc-binary-op "eulr" 'calcFunc-euler arg)
|
|
(calc-unary-op "eulr" 'calcFunc-euler arg))))
|
|
|
|
(defun calc-stirling-number (arg)
|
|
(interactive "P")
|
|
(calc-slow-wrapper
|
|
(if (calc-is-hyperbolic)
|
|
(calc-binary-op "str2" 'calcFunc-stir2 arg)
|
|
(calc-binary-op "str1" 'calcFunc-stir1 arg))))
|
|
|
|
(defun calc-utpb ()
|
|
(interactive)
|
|
(calc-prob-dist "b" 3))
|
|
|
|
(defun calc-utpc ()
|
|
(interactive)
|
|
(calc-prob-dist "c" 2))
|
|
|
|
(defun calc-utpf ()
|
|
(interactive)
|
|
(calc-prob-dist "f" 3))
|
|
|
|
(defun calc-utpn ()
|
|
(interactive)
|
|
(calc-prob-dist "n" 3))
|
|
|
|
(defun calc-utpp ()
|
|
(interactive)
|
|
(calc-prob-dist "p" 2))
|
|
|
|
(defun calc-utpt ()
|
|
(interactive)
|
|
(calc-prob-dist "t" 2))
|
|
|
|
(defun calc-prob-dist (letter nargs)
|
|
(calc-slow-wrapper
|
|
(if (calc-is-inverse)
|
|
(calc-enter-result nargs (concat "ltp" letter)
|
|
(append (list (intern (concat "calcFunc-ltp" letter))
|
|
(calc-top-n 1))
|
|
(calc-top-list-n (1- nargs) 2)))
|
|
(calc-enter-result nargs (concat "utp" letter)
|
|
(append (list (intern (concat "calcFunc-utp" letter))
|
|
(calc-top-n 1))
|
|
(calc-top-list-n (1- nargs) 2))))))
|
|
|
|
|
|
|
|
|
|
;;; Sources: Numerical Recipes, Press et al;
|
|
;;; Handbook of Mathematical Functions, Abramowitz & Stegun.
|
|
|
|
|
|
;;; Gamma function.
|
|
|
|
(defun calcFunc-gamma (x)
|
|
(or (math-numberp x) (math-reject-arg x 'numberp))
|
|
(calcFunc-fact (math-add x -1)))
|
|
|
|
(defun math-gammap1-raw (x &optional fprec nfprec) ; compute gamma(1 + x)
|
|
(or fprec
|
|
(setq fprec (math-float calc-internal-prec)
|
|
nfprec (math-float (- calc-internal-prec))))
|
|
(cond ((math-lessp-float (calcFunc-re x) fprec)
|
|
(if (math-lessp-float (calcFunc-re x) nfprec)
|
|
(math-neg (math-div
|
|
(math-pi)
|
|
(math-mul (math-gammap1-raw
|
|
(math-add (math-neg x)
|
|
'(float -1 0))
|
|
fprec nfprec)
|
|
(math-sin-raw
|
|
(math-mul (math-pi) x)))))
|
|
(let ((xplus1 (math-add x '(float 1 0))))
|
|
(math-div (math-gammap1-raw xplus1 fprec nfprec) xplus1))))
|
|
((and (math-realp x)
|
|
(math-lessp-float '(float 736276 0) x))
|
|
(math-overflow))
|
|
(t ; re(x) now >= 10.0
|
|
(let ((xinv (math-div 1 x))
|
|
(lnx (math-ln-raw x)))
|
|
(math-mul (math-sqrt-two-pi)
|
|
(math-exp-raw
|
|
(math-gamma-series
|
|
(math-sub (math-mul (math-add x '(float 5 -1))
|
|
lnx)
|
|
x)
|
|
xinv
|
|
(math-sqr xinv)
|
|
'(float 0 0)
|
|
2)))))))
|
|
|
|
(defun math-gamma-series (sum x xinvsqr oterm n)
|
|
(math-working "gamma" sum)
|
|
(let* ((bn (math-bernoulli-number n))
|
|
(term (math-mul (math-div-float (math-float (nth 1 bn))
|
|
(math-float (* (nth 2 bn)
|
|
(* n (1- n)))))
|
|
x))
|
|
(next (math-add sum term)))
|
|
(if (math-nearly-equal sum next)
|
|
next
|
|
(if (> n (* 2 calc-internal-prec))
|
|
(progn
|
|
;; Need this because series eventually diverges for large enough n.
|
|
(calc-record-why
|
|
"*Gamma computation stopped early, not all digits may be valid")
|
|
next)
|
|
(math-gamma-series next (math-mul x xinvsqr) xinvsqr term (+ n 2))))))
|
|
|
|
|
|
;;; Incomplete gamma function.
|
|
|
|
(defvar math-current-gamma-value nil)
|
|
(defun calcFunc-gammaP (a x)
|
|
(if (equal x '(var inf var-inf))
|
|
'(float 1 0)
|
|
(math-inexact-result)
|
|
(or (Math-numberp a) (math-reject-arg a 'numberp))
|
|
(or (math-numberp x) (math-reject-arg x 'numberp))
|
|
(if (and (math-num-integerp a)
|
|
(integerp (setq a (math-trunc a)))
|
|
(> a 0) (< a 20))
|
|
(math-sub 1 (calcFunc-gammaQ a x))
|
|
(let ((math-current-gamma-value (calcFunc-gamma a)))
|
|
(math-div (calcFunc-gammag a x) math-current-gamma-value)))))
|
|
|
|
(defun calcFunc-gammaQ (a x)
|
|
(if (equal x '(var inf var-inf))
|
|
'(float 0 0)
|
|
(math-inexact-result)
|
|
(or (Math-numberp a) (math-reject-arg a 'numberp))
|
|
(or (math-numberp x) (math-reject-arg x 'numberp))
|
|
(if (and (math-num-integerp a)
|
|
(integerp (setq a (math-trunc a)))
|
|
(> a 0) (< a 20))
|
|
(let ((n 0)
|
|
(sum '(float 1 0))
|
|
(term '(float 1 0)))
|
|
(math-with-extra-prec 1
|
|
(while (< (setq n (1+ n)) a)
|
|
(setq term (math-div (math-mul term x) n)
|
|
sum (math-add sum term))
|
|
(math-working "gamma" sum))
|
|
(math-mul sum (calcFunc-exp (math-neg x)))))
|
|
(let ((math-current-gamma-value (calcFunc-gamma a)))
|
|
(math-div (calcFunc-gammaG a x) math-current-gamma-value)))))
|
|
|
|
(defun calcFunc-gammag (a x)
|
|
(if (equal x '(var inf var-inf))
|
|
(calcFunc-gamma a)
|
|
(math-inexact-result)
|
|
(or (Math-numberp a) (math-reject-arg a 'numberp))
|
|
(or (Math-numberp x) (math-reject-arg x 'numberp))
|
|
(math-with-extra-prec 2
|
|
(setq a (math-float a))
|
|
(setq x (math-float x))
|
|
(if (or (math-negp (calcFunc-re a))
|
|
(math-lessp-float (calcFunc-re x)
|
|
(math-add-float (calcFunc-re a)
|
|
'(float 1 0))))
|
|
(math-inc-gamma-series a x)
|
|
(math-sub (or math-current-gamma-value (calcFunc-gamma a))
|
|
(math-inc-gamma-cfrac a x))))))
|
|
|
|
(defun calcFunc-gammaG (a x)
|
|
(if (equal x '(var inf var-inf))
|
|
'(float 0 0)
|
|
(math-inexact-result)
|
|
(or (Math-numberp a) (math-reject-arg a 'numberp))
|
|
(or (Math-numberp x) (math-reject-arg x 'numberp))
|
|
(math-with-extra-prec 2
|
|
(setq a (math-float a))
|
|
(setq x (math-float x))
|
|
(if (or (math-negp (calcFunc-re a))
|
|
(math-lessp-float (calcFunc-re x)
|
|
(math-add-float (math-abs-approx a)
|
|
'(float 1 0))))
|
|
(math-sub (or math-current-gamma-value (calcFunc-gamma a))
|
|
(math-inc-gamma-series a x))
|
|
(math-inc-gamma-cfrac a x)))))
|
|
|
|
(defun math-inc-gamma-series (a x)
|
|
(if (Math-zerop x)
|
|
'(float 0 0)
|
|
(math-mul (math-exp-raw (math-sub (math-mul a (math-ln-raw x)) x))
|
|
(math-with-extra-prec 2
|
|
(let ((start (math-div '(float 1 0) a)))
|
|
(math-inc-gamma-series-step start start a x))))))
|
|
|
|
(defun math-inc-gamma-series-step (sum term a x)
|
|
(math-working "gamma" sum)
|
|
(setq a (math-add a '(float 1 0))
|
|
term (math-div (math-mul term x) a))
|
|
(let ((next (math-add sum term)))
|
|
(if (math-nearly-equal sum next)
|
|
next
|
|
(math-inc-gamma-series-step next term a x))))
|
|
|
|
(defun math-inc-gamma-cfrac (a x)
|
|
(if (Math-zerop x)
|
|
(or math-current-gamma-value (calcFunc-gamma a))
|
|
(math-mul (math-exp-raw (math-sub (math-mul a (math-ln-raw x)) x))
|
|
(math-inc-gamma-cfrac-step '(float 1 0) x
|
|
'(float 0 0) '(float 1 0)
|
|
'(float 1 0) '(float 1 0) '(float 0 0)
|
|
a x))))
|
|
|
|
(defun math-inc-gamma-cfrac-step (a0 a1 b0 b1 n fac g a x)
|
|
(let ((ana (math-sub n a))
|
|
(anf (math-mul n fac)))
|
|
(setq n (math-add n '(float 1 0))
|
|
a0 (math-mul (math-add a1 (math-mul a0 ana)) fac)
|
|
b0 (math-mul (math-add b1 (math-mul b0 ana)) fac)
|
|
a1 (math-add (math-mul x a0) (math-mul anf a1))
|
|
b1 (math-add (math-mul x b0) (math-mul anf b1)))
|
|
(if (math-zerop a1)
|
|
(math-inc-gamma-cfrac-step a0 a1 b0 b1 n fac g a x)
|
|
(setq fac (math-div '(float 1 0) a1))
|
|
(let ((next (math-mul b1 fac)))
|
|
(math-working "gamma" next)
|
|
(if (math-nearly-equal next g)
|
|
next
|
|
(math-inc-gamma-cfrac-step a0 a1 b0 b1 n fac next a x))))))
|
|
|
|
|
|
;;; Error function.
|
|
|
|
(defun calcFunc-erf (x)
|
|
(if (equal x '(var inf var-inf))
|
|
'(float 1 0)
|
|
(if (equal x '(neg (var inf var-inf)))
|
|
'(float -1 0)
|
|
(if (Math-zerop x)
|
|
x
|
|
(let ((math-current-gamma-value (math-sqrt-pi)))
|
|
(math-to-same-complex-quad
|
|
(math-div (calcFunc-gammag '(float 5 -1)
|
|
(math-sqr (math-to-complex-quad-one x)))
|
|
math-current-gamma-value)
|
|
x))))))
|
|
|
|
(defun calcFunc-erfc (x)
|
|
(if (equal x '(var inf var-inf))
|
|
'(float 0 0)
|
|
(if (math-posp x)
|
|
(let ((math-current-gamma-value (math-sqrt-pi)))
|
|
(math-div (calcFunc-gammaG '(float 5 -1) (math-sqr x))
|
|
math-current-gamma-value))
|
|
(math-sub 1 (calcFunc-erf x)))))
|
|
|
|
(defun math-to-complex-quad-one (x)
|
|
(if (eq (car-safe x) 'polar) (setq x (math-complex x)))
|
|
(if (eq (car-safe x) 'cplx)
|
|
(list 'cplx (math-abs (nth 1 x)) (math-abs (nth 2 x)))
|
|
x))
|
|
|
|
(defun math-to-same-complex-quad (x y)
|
|
(if (eq (car-safe y) 'cplx)
|
|
(if (eq (car-safe x) 'cplx)
|
|
(list 'cplx
|
|
(if (math-negp (nth 1 y)) (math-neg (nth 1 x)) (nth 1 x))
|
|
(if (math-negp (nth 2 y)) (math-neg (nth 2 x)) (nth 2 x)))
|
|
(if (math-negp (nth 1 y)) (math-neg x) x))
|
|
(if (math-negp y)
|
|
(if (eq (car-safe x) 'cplx)
|
|
(list 'cplx (math-neg (nth 1 x)) (nth 2 x))
|
|
(math-neg x))
|
|
x)))
|
|
|
|
|
|
;;; Beta function.
|
|
|
|
(defun calcFunc-beta (a b)
|
|
(if (math-num-integerp a)
|
|
(let ((am (math-add a -1)))
|
|
(or (math-numberp b) (math-reject-arg b 'numberp))
|
|
(math-div 1 (math-mul b (calcFunc-choose (math-add b am) am))))
|
|
(if (math-num-integerp b)
|
|
(calcFunc-beta b a)
|
|
(math-div (math-mul (calcFunc-gamma a) (calcFunc-gamma b))
|
|
(calcFunc-gamma (math-add a b))))))
|
|
|
|
|
|
;;; Incomplete beta function.
|
|
|
|
(defvar math-current-beta-value nil)
|
|
(defun calcFunc-betaI (x a b)
|
|
(cond ((math-zerop x)
|
|
'(float 0 0))
|
|
((math-equal-int x 1)
|
|
'(float 1 0))
|
|
((or (math-zerop a)
|
|
(and (math-num-integerp a)
|
|
(math-negp a)))
|
|
(if (or (math-zerop b)
|
|
(and (math-num-integerp b)
|
|
(math-negp b)))
|
|
(math-reject-arg b 'range)
|
|
'(float 1 0)))
|
|
((or (math-zerop b)
|
|
(and (math-num-integerp b)
|
|
(math-negp b)))
|
|
'(float 0 0))
|
|
((not (math-numberp a)) (math-reject-arg a 'numberp))
|
|
((not (math-numberp b)) (math-reject-arg b 'numberp))
|
|
((math-inexact-result))
|
|
(t (let ((math-current-beta-value (calcFunc-beta a b)))
|
|
(math-div (calcFunc-betaB x a b) math-current-beta-value)))))
|
|
|
|
(defun calcFunc-betaB (x a b)
|
|
(cond
|
|
((math-zerop x)
|
|
'(float 0 0))
|
|
((math-equal-int x 1)
|
|
(calcFunc-beta a b))
|
|
((not (math-numberp x)) (math-reject-arg x 'numberp))
|
|
((not (math-numberp a)) (math-reject-arg a 'numberp))
|
|
((not (math-numberp b)) (math-reject-arg b 'numberp))
|
|
((math-zerop a) (math-reject-arg a 'nonzerop))
|
|
((math-zerop b) (math-reject-arg b 'nonzerop))
|
|
((and (math-num-integerp b)
|
|
(if (math-negp b)
|
|
(math-reject-arg b 'range)
|
|
(Math-natnum-lessp (setq b (math-trunc b)) 20)))
|
|
(and calc-symbolic-mode (or (math-floatp a) (math-floatp b))
|
|
(math-inexact-result))
|
|
(math-mul
|
|
(math-with-extra-prec 2
|
|
(let* ((i 0)
|
|
(term 1)
|
|
(sum (math-div term a)))
|
|
(while (< (setq i (1+ i)) b)
|
|
(setq term (math-mul (math-div (math-mul term (- i b)) i) x)
|
|
sum (math-add sum (math-div term (math-add a i))))
|
|
(math-working "beta" sum))
|
|
sum))
|
|
(math-pow x a)))
|
|
((and (math-num-integerp a)
|
|
(if (math-negp a)
|
|
(math-reject-arg a 'range)
|
|
(Math-natnum-lessp (setq a (math-trunc a)) 20)))
|
|
(math-sub (or math-current-beta-value (calcFunc-beta a b))
|
|
(calcFunc-betaB (math-sub 1 x) b a)))
|
|
(t
|
|
(math-inexact-result)
|
|
(math-with-extra-prec 2
|
|
(setq x (math-float x))
|
|
(setq a (math-float a))
|
|
(setq b (math-float b))
|
|
(let ((bt (math-exp-raw (math-add (math-mul a (math-ln-raw x))
|
|
(math-mul b (math-ln-raw
|
|
(math-sub '(float 1 0)
|
|
x)))))))
|
|
(if (Math-lessp x (math-div (math-add a '(float 1 0))
|
|
(math-add (math-add a b) '(float 2 0))))
|
|
(math-div (math-mul bt (math-beta-cfrac a b x)) a)
|
|
(math-sub (or math-current-beta-value (calcFunc-beta a b))
|
|
(math-div (math-mul bt
|
|
(math-beta-cfrac b a (math-sub 1 x)))
|
|
b))))))))
|
|
|
|
(defun math-beta-cfrac (a b x)
|
|
(let ((qab (math-add a b))
|
|
(qap (math-add a '(float 1 0)))
|
|
(qam (math-add a '(float -1 0))))
|
|
(math-beta-cfrac-step '(float 1 0)
|
|
(math-sub '(float 1 0)
|
|
(math-div (math-mul qab x) qap))
|
|
'(float 1 0) '(float 1 0)
|
|
'(float 1 0)
|
|
qab qap qam a b x)))
|
|
|
|
(defun math-beta-cfrac-step (az bz am bm m qab qap qam a b x)
|
|
(let* ((two-m (math-mul m '(float 2 0)))
|
|
(d (math-div (math-mul (math-mul (math-sub b m) m) x)
|
|
(math-mul (math-add qam two-m) (math-add a two-m))))
|
|
(ap (math-add az (math-mul d am)))
|
|
(bp (math-add bz (math-mul d bm)))
|
|
(d2 (math-neg
|
|
(math-div (math-mul (math-mul (math-add a m) (math-add qab m)) x)
|
|
(math-mul (math-add qap two-m) (math-add a two-m)))))
|
|
(app (math-add ap (math-mul d2 az)))
|
|
(bpp (math-add bp (math-mul d2 bz)))
|
|
(next (math-div app bpp)))
|
|
(math-working "beta" next)
|
|
(if (math-nearly-equal next az)
|
|
next
|
|
(math-beta-cfrac-step next '(float 1 0)
|
|
(math-div ap bpp) (math-div bp bpp)
|
|
(math-add m '(float 1 0))
|
|
qab qap qam a b x))))
|
|
|
|
|
|
;;; Bessel functions.
|
|
|
|
;;; Should generalize this to handle arbitrary precision!
|
|
|
|
(defun calcFunc-besJ (v x)
|
|
(or (math-numberp v) (math-reject-arg v 'numberp))
|
|
(or (math-numberp x) (math-reject-arg x 'numberp))
|
|
(let ((calc-internal-prec (min 8 calc-internal-prec)))
|
|
(math-with-extra-prec 3
|
|
(setq x (math-float (math-normalize x)))
|
|
(setq v (math-float (math-normalize v)))
|
|
(cond ((math-zerop x)
|
|
(if (math-zerop v)
|
|
'(float 1 0)
|
|
'(float 0 0)))
|
|
((math-inexact-result))
|
|
((not (math-num-integerp v))
|
|
(let ((start (math-div 1 (calcFunc-fact v))))
|
|
(math-mul (math-besJ-series start start
|
|
0
|
|
(math-mul '(float -25 -2)
|
|
(math-sqr x))
|
|
v)
|
|
(math-pow (math-div x 2) v))))
|
|
((math-negp (setq v (math-trunc v)))
|
|
(if (math-oddp v)
|
|
(math-neg (calcFunc-besJ (math-neg v) x))
|
|
(calcFunc-besJ (math-neg v) x)))
|
|
((eq v 0)
|
|
(math-besJ0 x))
|
|
((eq v 1)
|
|
(math-besJ1 x))
|
|
((Math-lessp v (math-abs-approx x))
|
|
(let ((j 0)
|
|
(bjm (math-besJ0 x))
|
|
(bj (math-besJ1 x))
|
|
(two-over-x (math-div 2 x))
|
|
bjp)
|
|
(while (< (setq j (1+ j)) v)
|
|
(setq bjp (math-sub (math-mul (math-mul j two-over-x) bj)
|
|
bjm)
|
|
bjm bj
|
|
bj bjp))
|
|
bj))
|
|
(t
|
|
(if (Math-lessp 100 v) (math-reject-arg v 'range))
|
|
(let* ((j (logior (+ v (math-isqrt-small (* 40 v))) 1))
|
|
(two-over-x (math-div 2 x))
|
|
(jsum nil)
|
|
(bjp '(float 0 0))
|
|
(sum '(float 0 0))
|
|
(bj '(float 1 0))
|
|
bjm ans)
|
|
(while (> (setq j (1- j)) 0)
|
|
(setq bjm (math-sub (math-mul (math-mul j two-over-x) bj)
|
|
bjp)
|
|
bjp bj
|
|
bj bjm)
|
|
(if (> (nth 2 (math-abs-approx bj)) 10)
|
|
(setq bj (math-mul bj '(float 1 -10))
|
|
bjp (math-mul bjp '(float 1 -10))
|
|
ans (and ans (math-mul ans '(float 1 -10)))
|
|
sum (math-mul sum '(float 1 -10))))
|
|
(or (setq jsum (not jsum))
|
|
(setq sum (math-add sum bj)))
|
|
(if (= j v)
|
|
(setq ans bjp)))
|
|
(math-div ans (math-sub (math-mul 2 sum) bj))))))))
|
|
|
|
(defun math-besJ-series (sum term k zz vk)
|
|
(math-working "besJ" sum)
|
|
(setq k (1+ k)
|
|
vk (math-add 1 vk)
|
|
term (math-div (math-mul term zz) (math-mul k vk)))
|
|
(let ((next (math-add sum term)))
|
|
(if (math-nearly-equal next sum)
|
|
next
|
|
(math-besJ-series next term k zz vk))))
|
|
|
|
(defun math-besJ0 (x &optional yflag)
|
|
(cond ((and (not yflag) (math-negp (calcFunc-re x)))
|
|
(math-besJ0 (math-neg x)))
|
|
((Math-lessp '(float 8 0) (math-abs-approx x))
|
|
(let* ((z (math-div '(float 8 0) x))
|
|
(y (math-sqr z))
|
|
(xx (math-add x '(float (bigneg 164 398 785) -9)))
|
|
(a1 (math-poly-eval y
|
|
'((float (bigpos 211 887 093 2) -16)
|
|
(float (bigneg 639 370 073 2) -15)
|
|
(float (bigpos 407 510 734 2) -14)
|
|
(float (bigneg 627 628 098 1) -12)
|
|
(float 1 0))))
|
|
(a2 (math-poly-eval y
|
|
'((float (bigneg 152 935 934) -16)
|
|
(float (bigpos 161 095 621 7) -16)
|
|
(float (bigneg 651 147 911 6) -15)
|
|
(float (bigpos 765 488 430 1) -13)
|
|
(float (bigneg 995 499 562 1) -11))))
|
|
(sc (math-sin-cos-raw xx)))
|
|
(if yflag
|
|
(setq sc (cons (math-neg (cdr sc)) (car sc))))
|
|
(math-mul (math-sqrt
|
|
(math-div '(float (bigpos 722 619 636) -9) x))
|
|
(math-sub (math-mul (cdr sc) a1)
|
|
(math-mul (car sc) (math-mul z a2))))))
|
|
(t
|
|
(let ((y (math-sqr x)))
|
|
(math-div (math-poly-eval y
|
|
'((float (bigneg 456 052 849 1) -7)
|
|
(float (bigpos 017 233 739 7) -5)
|
|
(float (bigneg 418 442 121 1) -2)
|
|
(float (bigpos 407 196 516 6) -1)
|
|
(float (bigneg 354 590 362 13) 0)
|
|
(float (bigpos 574 490 568 57) 0)))
|
|
(math-poly-eval y
|
|
'((float 1 0)
|
|
(float (bigpos 712 532 678 2) -7)
|
|
(float (bigpos 853 264 927 5) -5)
|
|
(float (bigpos 718 680 494 9) -3)
|
|
(float (bigpos 985 532 029 1) 0)
|
|
(float (bigpos 411 490 568 57) 0))))))))
|
|
|
|
(defun math-besJ1 (x &optional yflag)
|
|
(cond ((and (math-negp (calcFunc-re x)) (not yflag))
|
|
(math-neg (math-besJ1 (math-neg x))))
|
|
((Math-lessp '(float 8 0) (math-abs-approx x))
|
|
(let* ((z (math-div '(float 8 0) x))
|
|
(y (math-sqr z))
|
|
(xx (math-add x '(float (bigneg 491 194 356 2) -9)))
|
|
(a1 (math-poly-eval y
|
|
'((float (bigneg 019 337 240) -15)
|
|
(float (bigpos 174 520 457 2) -15)
|
|
(float (bigneg 496 396 516 3) -14)
|
|
(float 183105 -8)
|
|
(float 1 0))))
|
|
(a2 (math-poly-eval y
|
|
'((float (bigpos 412 787 105) -15)
|
|
(float (bigneg 987 228 88) -14)
|
|
(float (bigpos 096 199 449 8) -15)
|
|
(float (bigneg 873 690 002 2) -13)
|
|
(float (bigpos 995 499 687 4) -11))))
|
|
(sc (math-sin-cos-raw xx)))
|
|
(if yflag
|
|
(setq sc (cons (math-neg (cdr sc)) (car sc)))
|
|
(if (math-negp x)
|
|
(setq sc (cons (math-neg (car sc)) (math-neg (cdr sc))))))
|
|
(math-mul (math-sqrt (math-div '(float (bigpos 722 619 636) -9) x))
|
|
(math-sub (math-mul (cdr sc) a1)
|
|
(math-mul (car sc) (math-mul z a2))))))
|
|
(t
|
|
(let ((y (math-sqr x)))
|
|
(math-mul
|
|
x
|
|
(math-div (math-poly-eval y
|
|
'((float (bigneg 606 036 016 3) -8)
|
|
(float (bigpos 826 044 157) -4)
|
|
(float (bigneg 439 611 972 2) -3)
|
|
(float (bigpos 531 968 423 2) -1)
|
|
(float (bigneg 235 059 895 7) 0)
|
|
(float (bigpos 232 614 362 72) 0)))
|
|
(math-poly-eval y
|
|
'((float 1 0)
|
|
(float (bigpos 397 991 769 3) -7)
|
|
(float (bigpos 394 743 944 9) -5)
|
|
(float (bigpos 474 330 858 1) -2)
|
|
(float (bigpos 178 535 300 2) 0)
|
|
(float (bigpos 442 228 725 144)
|
|
0)))))))))
|
|
|
|
(defun calcFunc-besY (v x)
|
|
(math-inexact-result)
|
|
(or (math-numberp v) (math-reject-arg v 'numberp))
|
|
(or (math-numberp x) (math-reject-arg x 'numberp))
|
|
(let ((calc-internal-prec (min 8 calc-internal-prec)))
|
|
(math-with-extra-prec 3
|
|
(setq x (math-float (math-normalize x)))
|
|
(setq v (math-float (math-normalize v)))
|
|
(cond ((not (math-num-integerp v))
|
|
(let ((sc (math-sin-cos-raw (math-mul v (math-pi)))))
|
|
(math-div (math-sub (math-mul (calcFunc-besJ v x) (cdr sc))
|
|
(calcFunc-besJ (math-neg v) x))
|
|
(car sc))))
|
|
((math-negp (setq v (math-trunc v)))
|
|
(if (math-oddp v)
|
|
(math-neg (calcFunc-besY (math-neg v) x))
|
|
(calcFunc-besY (math-neg v) x)))
|
|
((eq v 0)
|
|
(math-besY0 x))
|
|
((eq v 1)
|
|
(math-besY1 x))
|
|
(t
|
|
(let ((j 0)
|
|
(bym (math-besY0 x))
|
|
(by (math-besY1 x))
|
|
(two-over-x (math-div 2 x))
|
|
byp)
|
|
(while (< (setq j (1+ j)) v)
|
|
(setq byp (math-sub (math-mul (math-mul j two-over-x) by)
|
|
bym)
|
|
bym by
|
|
by byp))
|
|
by))))))
|
|
|
|
(defun math-besY0 (x)
|
|
(cond ((Math-lessp (math-abs-approx x) '(float 8 0))
|
|
(let ((y (math-sqr x)))
|
|
(math-add
|
|
(math-div (math-poly-eval y
|
|
'((float (bigpos 733 622 284 2) -7)
|
|
(float (bigneg 757 792 632 8) -5)
|
|
(float (bigpos 129 988 087 1) -2)
|
|
(float (bigneg 036 598 123 5) -1)
|
|
(float (bigpos 065 834 062 7) 0)
|
|
(float (bigneg 389 821 957 2) 0)))
|
|
(math-poly-eval y
|
|
'((float 1 0)
|
|
(float (bigpos 244 030 261 2) -7)
|
|
(float (bigpos 647 472 474) -4)
|
|
(float (bigpos 438 466 189 7) -3)
|
|
(float (bigpos 648 499 452 7) -1)
|
|
(float (bigpos 269 544 076 40) 0))))
|
|
(math-mul '(float (bigpos 772 619 636) -9)
|
|
(math-mul (math-besJ0 x) (math-ln-raw x))))))
|
|
((math-negp (calcFunc-re x))
|
|
(math-add (math-besJ0 (math-neg x) t)
|
|
(math-mul '(cplx 0 2)
|
|
(math-besJ0 (math-neg x)))))
|
|
(t
|
|
(math-besJ0 x t))))
|
|
|
|
(defun math-besY1 (x)
|
|
(cond ((Math-lessp (math-abs-approx x) '(float 8 0))
|
|
(let ((y (math-sqr x)))
|
|
(math-add
|
|
(math-mul
|
|
x
|
|
(math-div (math-poly-eval y
|
|
'((float (bigpos 935 937 511 8) -6)
|
|
(float (bigneg 726 922 237 4) -3)
|
|
(float (bigpos 551 264 349 7) -1)
|
|
(float (bigneg 139 438 153 5) 1)
|
|
(float (bigpos 439 527 127) 4)
|
|
(float (bigneg 943 604 900 4) 3)))
|
|
(math-poly-eval y
|
|
'((float 1 0)
|
|
(float (bigpos 885 632 549 3) -7)
|
|
(float (bigpos 605 042 102) -3)
|
|
(float (bigpos 002 904 245 2) -2)
|
|
(float (bigpos 367 650 733 3) 0)
|
|
(float (bigpos 664 419 244 4) 2)
|
|
(float (bigpos 057 958 249) 5)))))
|
|
(math-mul '(float (bigpos 772 619 636) -9)
|
|
(math-sub (math-mul (math-besJ1 x) (math-ln-raw x))
|
|
(math-div 1 x))))))
|
|
((math-negp (calcFunc-re x))
|
|
(math-neg
|
|
(math-add (math-besJ1 (math-neg x) t)
|
|
(math-mul '(cplx 0 2)
|
|
(math-besJ1 (math-neg x))))))
|
|
(t
|
|
(math-besJ1 x t))))
|
|
|
|
(defun math-poly-eval (x coefs)
|
|
(let ((accum (car coefs)))
|
|
(while (setq coefs (cdr coefs))
|
|
(setq accum (math-add (car coefs) (math-mul accum x))))
|
|
accum))
|
|
|
|
|
|
;;;; Bernoulli and Euler polynomials and numbers.
|
|
|
|
(defun calcFunc-bern (n &optional x)
|
|
(if (and x (not (math-zerop x)))
|
|
(if (and calc-symbolic-mode (math-floatp x))
|
|
(math-inexact-result)
|
|
(math-build-polynomial-expr (math-bernoulli-coefs n) x))
|
|
(or (math-num-natnump n) (math-reject-arg n 'natnump))
|
|
(if (consp n)
|
|
(progn
|
|
(math-inexact-result)
|
|
(math-float (math-bernoulli-number (math-trunc n))))
|
|
(math-bernoulli-number n))))
|
|
|
|
(defun calcFunc-euler (n &optional x)
|
|
(or (math-num-natnump n) (math-reject-arg n 'natnump))
|
|
(if x
|
|
(let* ((n1 (math-add n 1))
|
|
(coefs (math-bernoulli-coefs n1))
|
|
(fac (math-div (math-pow 2 n1) n1))
|
|
(k -1)
|
|
(x1 (math-div (math-add x 1) 2))
|
|
(x2 (math-div x 2)))
|
|
(if (math-numberp x)
|
|
(if (and calc-symbolic-mode (math-floatp x))
|
|
(math-inexact-result)
|
|
(math-mul fac
|
|
(math-sub (math-build-polynomial-expr coefs x1)
|
|
(math-build-polynomial-expr coefs x2))))
|
|
(calcFunc-collect
|
|
(math-reduce-vec
|
|
'math-add
|
|
(cons 'vec
|
|
(mapcar (function
|
|
(lambda (c)
|
|
(setq k (1+ k))
|
|
(math-mul (math-mul fac c)
|
|
(math-sub (math-pow x1 k)
|
|
(math-pow x2 k)))))
|
|
coefs)))
|
|
x)))
|
|
(math-mul (math-pow 2 n)
|
|
(if (consp n)
|
|
(progn
|
|
(math-inexact-result)
|
|
(calcFunc-euler n '(float 5 -1)))
|
|
(calcFunc-euler n '(frac 1 2))))))
|
|
|
|
(defvar math-bernoulli-b-cache '((frac -174611
|
|
(bigpos 0 200 291 698 662 857 802))
|
|
(frac 43867 (bigpos 0 944 170 217 94 109 5))
|
|
(frac -3617 (bigpos 0 880 842 622 670 10))
|
|
(frac 1 (bigpos 600 249 724 74))
|
|
(frac -691 (bigpos 0 368 674 307 1))
|
|
(frac 1 (bigpos 160 900 47))
|
|
(frac -1 (bigpos 600 209 1))
|
|
(frac 1 30240) (frac -1 720)
|
|
(frac 1 12) 1 ))
|
|
|
|
(defvar math-bernoulli-B-cache '((frac -174611 330) (frac 43867 798)
|
|
(frac -3617 510) (frac 7 6) (frac -691 2730)
|
|
(frac 5 66) (frac -1 30) (frac 1 42)
|
|
(frac -1 30) (frac 1 6) 1 ))
|
|
|
|
(defvar math-bernoulli-cache-size 11)
|
|
(defun math-bernoulli-coefs (n)
|
|
(let* ((coefs (list (calcFunc-bern n)))
|
|
(nn (math-trunc n))
|
|
(k nn)
|
|
(term nn)
|
|
coef
|
|
(calc-prefer-frac (or (integerp n) calc-prefer-frac)))
|
|
(while (>= (setq k (1- k)) 0)
|
|
(setq term (math-div term (- nn k))
|
|
coef (math-mul term (math-bernoulli-number k))
|
|
coefs (cons (if (consp n) (math-float coef) coef) coefs)
|
|
term (math-mul term k)))
|
|
(nreverse coefs)))
|
|
|
|
(defun math-bernoulli-number (n)
|
|
(if (= (% n 2) 1)
|
|
(if (= n 1)
|
|
'(frac -1 2)
|
|
0)
|
|
(setq n (/ n 2))
|
|
(while (>= n math-bernoulli-cache-size)
|
|
(let* ((sum 0)
|
|
(nk 1) ; nk = n-k+1
|
|
(fact 1) ; fact = (n-k+1)!
|
|
ofact
|
|
(p math-bernoulli-b-cache)
|
|
(calc-prefer-frac t))
|
|
(math-working "bernoulli B" (* 2 math-bernoulli-cache-size))
|
|
(while p
|
|
(setq nk (+ nk 2)
|
|
ofact fact
|
|
fact (math-mul fact (* nk (1- nk)))
|
|
sum (math-add sum (math-div (car p) fact))
|
|
p (cdr p)))
|
|
(setq ofact (math-mul ofact (1- nk))
|
|
sum (math-sub (math-div '(frac 1 2) ofact) sum)
|
|
math-bernoulli-b-cache (cons sum math-bernoulli-b-cache)
|
|
math-bernoulli-B-cache (cons (math-mul sum ofact)
|
|
math-bernoulli-B-cache)
|
|
math-bernoulli-cache-size (1+ math-bernoulli-cache-size))))
|
|
(nth (- math-bernoulli-cache-size n 1) math-bernoulli-B-cache)))
|
|
|
|
;;; Bn = n! bn
|
|
;;; bn = - sum_k=0^n-1 bk / (n-k+1)!
|
|
|
|
;;; A faster method would be to use "tangent numbers", c.f., Concrete
|
|
;;; Mathematics pg. 273.
|
|
|
|
|
|
;;; Probability distributions.
|
|
|
|
;;; Binomial.
|
|
(defun calcFunc-utpb (x n p)
|
|
(if math-expand-formulas
|
|
(math-normalize (list 'calcFunc-betaI p x (list '+ (list '- n x) 1)))
|
|
(calcFunc-betaI p x (math-add (math-sub n x) 1))))
|
|
(put 'calcFunc-utpb 'math-expandable t)
|
|
|
|
(defun calcFunc-ltpb (x n p)
|
|
(math-sub 1 (calcFunc-utpb x n p)))
|
|
(put 'calcFunc-ltpb 'math-expandable t)
|
|
|
|
;;; Chi-square.
|
|
(defun calcFunc-utpc (chisq v)
|
|
(if math-expand-formulas
|
|
(math-normalize (list 'calcFunc-gammaQ (list '/ v 2) (list '/ chisq 2)))
|
|
(calcFunc-gammaQ (math-div v 2) (math-div chisq 2))))
|
|
(put 'calcFunc-utpc 'math-expandable t)
|
|
|
|
(defun calcFunc-ltpc (chisq v)
|
|
(if math-expand-formulas
|
|
(math-normalize (list 'calcFunc-gammaP (list '/ v 2) (list '/ chisq 2)))
|
|
(calcFunc-gammaP (math-div v 2) (math-div chisq 2))))
|
|
(put 'calcFunc-ltpc 'math-expandable t)
|
|
|
|
;;; F-distribution.
|
|
(defun calcFunc-utpf (f v1 v2)
|
|
(if math-expand-formulas
|
|
(math-normalize (list 'calcFunc-betaI
|
|
(list '/ v2 (list '+ v2 (list '* v1 f)))
|
|
(list '/ v2 2)
|
|
(list '/ v1 2)))
|
|
(calcFunc-betaI (math-div v2 (math-add v2 (math-mul v1 f)))
|
|
(math-div v2 2)
|
|
(math-div v1 2))))
|
|
(put 'calcFunc-utpf 'math-expandable t)
|
|
|
|
(defun calcFunc-ltpf (f v1 v2)
|
|
(math-sub 1 (calcFunc-utpf f v1 v2)))
|
|
(put 'calcFunc-ltpf 'math-expandable t)
|
|
|
|
;;; Normal.
|
|
(defun calcFunc-utpn (x mean sdev)
|
|
(if math-expand-formulas
|
|
(math-normalize
|
|
(list '/
|
|
(list '+ 1
|
|
(list 'calcFunc-erf
|
|
(list '/ (list '- mean x)
|
|
(list '* sdev (list 'calcFunc-sqrt 2)))))
|
|
2))
|
|
(math-mul (math-add '(float 1 0)
|
|
(calcFunc-erf
|
|
(math-div (math-sub mean x)
|
|
(math-mul sdev (math-sqrt-2)))))
|
|
'(float 5 -1))))
|
|
(put 'calcFunc-utpn 'math-expandable t)
|
|
|
|
(defun calcFunc-ltpn (x mean sdev)
|
|
(if math-expand-formulas
|
|
(math-normalize
|
|
(list '/
|
|
(list '+ 1
|
|
(list 'calcFunc-erf
|
|
(list '/ (list '- x mean)
|
|
(list '* sdev (list 'calcFunc-sqrt 2)))))
|
|
2))
|
|
(math-mul (math-add '(float 1 0)
|
|
(calcFunc-erf
|
|
(math-div (math-sub x mean)
|
|
(math-mul sdev (math-sqrt-2)))))
|
|
'(float 5 -1))))
|
|
(put 'calcFunc-ltpn 'math-expandable t)
|
|
|
|
;;; Poisson.
|
|
(defun calcFunc-utpp (n x)
|
|
(if math-expand-formulas
|
|
(math-normalize (list 'calcFunc-gammaP x n))
|
|
(calcFunc-gammaP x n)))
|
|
(put 'calcFunc-utpp 'math-expandable t)
|
|
|
|
(defun calcFunc-ltpp (n x)
|
|
(if math-expand-formulas
|
|
(math-normalize (list 'calcFunc-gammaQ x n))
|
|
(calcFunc-gammaQ x n)))
|
|
(put 'calcFunc-ltpp 'math-expandable t)
|
|
|
|
;;; Student's t. (As defined in Abramowitz & Stegun and Numerical Recipes.)
|
|
(defun calcFunc-utpt (tt v)
|
|
(if math-expand-formulas
|
|
(math-normalize (list 'calcFunc-betaI
|
|
(list '/ v (list '+ v (list '^ tt 2)))
|
|
(list '/ v 2)
|
|
'(float 5 -1)))
|
|
(calcFunc-betaI (math-div v (math-add v (math-sqr tt)))
|
|
(math-div v 2)
|
|
'(float 5 -1))))
|
|
(put 'calcFunc-utpt 'math-expandable t)
|
|
|
|
(defun calcFunc-ltpt (tt v)
|
|
(math-sub 1 (calcFunc-utpt tt v)))
|
|
(put 'calcFunc-ltpt 'math-expandable t)
|
|
|
|
|
|
;;; calc-funcs.el ends here
|