Mercurial > emacs
view src/floatfns.c @ 1333:5054c696885d
(dired-lisp-ls): var `short' was erronously unbound.
author | Sebastian Kremer <sk@thp.uni-koeln.de> |
---|---|
date | Sun, 04 Oct 1992 10:36:03 +0000 |
parents | 70ed307d9047 |
children | bef6b6903528 |
line wrap: on
line source
/* Primitive operations on floating point for GNU Emacs Lisp interpreter. Copyright (C) 1988, 1992 Free Software Foundation, Inc. This file is part of GNU Emacs. GNU Emacs is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2, or (at your option) any later version. GNU Emacs is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with GNU Emacs; see the file COPYING. If not, write to the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. */ #include <signal.h> #include "config.h" #include "lisp.h" #include "syssignal.h" Lisp_Object Qarith_error; #ifdef LISP_FLOAT_TYPE #include <math.h> #include <errno.h> extern int errno; /* Avoid traps on VMS from sinh and cosh. All the other functions set errno instead. */ #ifdef VMS #undef cosh #undef sinh #define cosh(x) ((exp(x)+exp(-x))*0.5) #define sinh(x) ((exp(x)-exp(-x))*0.5) #endif /* VMS */ static SIGTYPE float_error (); /* Nonzero while executing in floating point. This tells float_error what to do. */ static int in_float; /* If an argument is out of range for a mathematical function, here is the actual argument value to use in the error message. */ static Lisp_Object float_error_arg; /* Evaluate the floating point expression D, recording NUM as the original argument for error messages. D is normally an assignment expression. Handle errors which may result in signals or may set errno. */ #define IN_FLOAT(D, NUM) \ (in_float = 1, errno = 0, float_error_arg = NUM, (D), \ (errno == ERANGE || errno == EDOM ? float_error () : (SIGTYPE) 0), \ in_float = 0) /* Extract a Lisp number as a `double', or signal an error. */ double extract_float (num) Lisp_Object num; { CHECK_NUMBER_OR_FLOAT (num, 0); if (XTYPE (num) == Lisp_Float) return XFLOAT (num)->data; return (double) XINT (num); } /* Trig functions. */ DEFUN ("acos", Facos, Sacos, 1, 1, 0, "Return the inverse cosine of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = acos (d), num); return make_float (d); } DEFUN ("asin", Fasin, Sasin, 1, 1, 0, "Return the inverse sine of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = asin (d), num); return make_float (d); } DEFUN ("atan", Fatan, Satan, 1, 1, 0, "Return the inverse tangent of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = atan (d), num); return make_float (d); } DEFUN ("cos", Fcos, Scos, 1, 1, 0, "Return the cosine of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = cos (d), num); return make_float (d); } DEFUN ("sin", Fsin, Ssin, 1, 1, 0, "Return the sine of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = sin (d), num); return make_float (d); } DEFUN ("tan", Ftan, Stan, 1, 1, 0, "Return the tangent of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = tan (d), num); return make_float (d); } #if 0 /* Leave these out unless we find there's a reason for them. */ DEFUN ("bessel-j0", Fbessel_j0, Sbessel_j0, 1, 1, 0, "Return the bessel function j0 of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = j0 (d), num); return make_float (d); } DEFUN ("bessel-j1", Fbessel_j1, Sbessel_j1, 1, 1, 0, "Return the bessel function j1 of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = j1 (d), num); return make_float (d); } DEFUN ("bessel-jn", Fbessel_jn, Sbessel_jn, 2, 2, 0, "Return the order N bessel function output jn of ARG.\n\ The first arg (the order) is truncated to an integer.") (num1, num2) register Lisp_Object num1, num2; { int i1 = extract_float (num1); double f2 = extract_float (num2); IN_FLOAT (f2 = jn (i1, f2), num1); return make_float (f2); } DEFUN ("bessel-y0", Fbessel_y0, Sbessel_y0, 1, 1, 0, "Return the bessel function y0 of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = y0 (d), num); return make_float (d); } DEFUN ("bessel-y1", Fbessel_y1, Sbessel_y1, 1, 1, 0, "Return the bessel function y1 of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = y1 (d), num); return make_float (d); } DEFUN ("bessel-yn", Fbessel_yn, Sbessel_yn, 2, 2, 0, "Return the order N bessel function output yn of ARG.\n\ The first arg (the order) is truncated to an integer.") (num1, num2) register Lisp_Object num1, num2; { int i1 = extract_float (num1); double f2 = extract_float (num2); IN_FLOAT (f2 = yn (i1, f2), num1); return make_float (f2); } #endif #if 0 /* Leave these out unless we see they are worth having. */ DEFUN ("erf", Ferf, Serf, 1, 1, 0, "Return the mathematical error function of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = erf (d), num); return make_float (d); } DEFUN ("erfc", Ferfc, Serfc, 1, 1, 0, "Return the complementary error function of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = erfc (d), num); return make_float (d); } DEFUN ("log-gamma", Flog_gamma, Slog_gamma, 1, 1, 0, "Return the log gamma of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = lgamma (d), num); return make_float (d); } DEFUN ("cbrt", Fcbrt, Scbrt, 1, 1, 0, "Return the cube root of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = cbrt (d), num); return make_float (d); } #endif DEFUN ("exp", Fexp, Sexp, 1, 1, 0, "Return the exponential base e of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = exp (d), num); return make_float (d); } DEFUN ("expt", Fexpt, Sexpt, 2, 2, 0, "Return the exponential X ** Y.") (num1, num2) register Lisp_Object num1, num2; { double f1, f2; CHECK_NUMBER_OR_FLOAT (num1, 0); CHECK_NUMBER_OR_FLOAT (num2, 0); if ((XTYPE (num1) == Lisp_Int) && /* common lisp spec */ (XTYPE (num2) == Lisp_Int)) /* don't promote, if both are ints */ { /* this can be improved by pre-calculating */ int acc, x, y; /* some binary powers of x then acumulating */ /* these, therby saving some time. -wsr */ x = XINT (num1); y = XINT (num2); acc = 1; if (y < 0) { for (; y < 0; y++) acc /= x; } else { for (; y > 0; y--) acc *= x; } return XSET (x, Lisp_Int, acc); } f1 = (XTYPE (num1) == Lisp_Float) ? XFLOAT (num1)->data : XINT (num1); f2 = (XTYPE (num2) == Lisp_Float) ? XFLOAT (num2)->data : XINT (num2); IN_FLOAT (f1 = pow (f1, f2), num1); return make_float (f1); } DEFUN ("log", Flog, Slog, 1, 2, 0, "Return the natural logarithm of NUM. If second optional argument BASE is given, return log NUM using that base.") (num, base) register Lisp_Object num; { double d = extract_float (num); if (NILP (base)) IN_FLOAT (d = log (d), num); else { double b = extract_float (base); IN_FLOAT (d = log (num) / log (b), num); } return make_float (d); } DEFUN ("log10", Flog10, Slog10, 1, 1, 0, "Return the logarithm base 10 of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = log10 (d), num); return make_float (d); } DEFUN ("sqrt", Fsqrt, Ssqrt, 1, 1, 0, "Return the square root of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = sqrt (d), num); return make_float (d); } #if 0 /* Not clearly worth adding. */ DEFUN ("acosh", Facosh, Sacosh, 1, 1, 0, "Return the inverse hyperbolic cosine of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = acosh (d), num); return make_float (d); } DEFUN ("asinh", Fasinh, Sasinh, 1, 1, 0, "Return the inverse hyperbolic sine of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = asinh (d), num); return make_float (d); } DEFUN ("atanh", Fatanh, Satanh, 1, 1, 0, "Return the inverse hyperbolic tangent of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = atanh (d), num); return make_float (d); } DEFUN ("cosh", Fcosh, Scosh, 1, 1, 0, "Return the hyperbolic cosine of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = cosh (d), num); return make_float (d); } DEFUN ("sinh", Fsinh, Ssinh, 1, 1, 0, "Return the hyperbolic sine of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = sinh (d), num); return make_float (d); } DEFUN ("tanh", Ftanh, Stanh, 1, 1, 0, "Return the hyperbolic tangent of ARG.") (num) register Lisp_Object num; { double d = extract_float (num); IN_FLOAT (d = tanh (d), num); return make_float (d); } #endif DEFUN ("abs", Fabs, Sabs, 1, 1, 0, "Return the absolute value of ARG.") (num) register Lisp_Object num; { CHECK_NUMBER_OR_FLOAT (num, 0); if (XTYPE (num) == Lisp_Float) IN_FLOAT (num = make_float (fabs (XFLOAT (num)->data)), num); else if (XINT (num) < 0) XSETINT (num, - XFASTINT (num)); return num; } DEFUN ("float", Ffloat, Sfloat, 1, 1, 0, "Return the floating point number equal to ARG.") (num) register Lisp_Object num; { CHECK_NUMBER_OR_FLOAT (num, 0); if (XTYPE (num) == Lisp_Int) return make_float ((double) XINT (num)); else /* give 'em the same float back */ return num; } DEFUN ("logb", Flogb, Slogb, 1, 1, 0, "Returns the integer that is the base 2 log of ARG.\n\ This is the same as the exponent of a float.") (num) Lisp_Object num; { #ifdef USG /* System V apparently doesn't have a `logb' function. */ return Flog (num, make_number (2)); #else Lisp_Object val; double f = extract_float (num); IN_FLOAT (val = logb (f), num); XSET (val, Lisp_Int, val); return val; #endif } /* the rounding functions */ DEFUN ("ceiling", Fceiling, Sceiling, 1, 1, 0, "Return the smallest integer no less than ARG. (Round toward +inf.)") (num) register Lisp_Object num; { CHECK_NUMBER_OR_FLOAT (num, 0); if (XTYPE (num) == Lisp_Float) IN_FLOAT (XSET (num, Lisp_Int, ceil (XFLOAT (num)->data)), num); return num; } DEFUN ("floor", Ffloor, Sfloor, 1, 1, 0, "Return the largest integer no greater than ARG. (Round towards -inf.)") (num) register Lisp_Object num; { CHECK_NUMBER_OR_FLOAT (num, 0); if (XTYPE (num) == Lisp_Float) IN_FLOAT (XSET (num, Lisp_Int, floor (XFLOAT (num)->data)), num); return num; } DEFUN ("round", Fround, Sround, 1, 1, 0, "Return the nearest integer to ARG.") (num) register Lisp_Object num; { CHECK_NUMBER_OR_FLOAT (num, 0); if (XTYPE (num) == Lisp_Float) { #ifdef USG /* Screw the prevailing rounding mode. */ IN_FLOAT (XSET (num, Lisp_Int, floor (XFLOAT (num)->data + 0.5)), num); #else IN_FLOAT (XSET (num, Lisp_Int, rint (XFLOAT (num)->data)), num); #endif } return num; } DEFUN ("truncate", Ftruncate, Struncate, 1, 1, 0, "Truncate a floating point number to an int.\n\ Rounds the value toward zero.") (num) register Lisp_Object num; { CHECK_NUMBER_OR_FLOAT (num, 0); if (XTYPE (num) == Lisp_Float) XSET (num, Lisp_Int, (int) XFLOAT (num)->data); return num; } static SIGTYPE float_error (signo) int signo; { if (! in_float) fatal_error_signal (signo); #ifdef BSD #ifdef BSD4_1 sigrelse (SIGILL); #else /* not BSD4_1 */ sigsetmask (SIGEMPTYMASK); #endif /* not BSD4_1 */ #else /* Must reestablish handler each time it is called. */ signal (SIGILL, float_error); #endif /* BSD */ in_float = 0; Fsignal (Qarith_error, Fcons (float_error_arg, Qnil)); } init_floatfns () { signal (SIGILL, float_error); in_float = 0; } syms_of_floatfns () { defsubr (&Sacos); defsubr (&Sasin); defsubr (&Satan); defsubr (&Scos); defsubr (&Ssin); defsubr (&Stan); #if 0 defsubr (&Sacosh); defsubr (&Sasinh); defsubr (&Satanh); defsubr (&Scosh); defsubr (&Ssinh); defsubr (&Stanh); defsubr (&Sbessel_y0); defsubr (&Sbessel_y1); defsubr (&Sbessel_yn); defsubr (&Sbessel_j0); defsubr (&Sbessel_j1); defsubr (&Sbessel_jn); defsubr (&Serf); defsubr (&Serfc); defsubr (&Slog_gamma); defsubr (&Scbrt); #endif defsubr (&Sexp); defsubr (&Sexpt); defsubr (&Slog); defsubr (&Slog10); defsubr (&Ssqrt); defsubr (&Sabs); defsubr (&Sfloat); defsubr (&Slogb); defsubr (&Sceiling); defsubr (&Sfloor); defsubr (&Sround); defsubr (&Struncate); } #else /* not LISP_FLOAT_TYPE */ init_floatfns () {} syms_of_floatfns () {} #endif /* not LISP_FLOAT_TYPE */