Mercurial > emacs
annotate lisp/emacs-lisp/float.el @ 11088:e88e9a016cfa
(syms_of_window): Fix missing \n\.
author | Karl Heuer <kwzh@gnu.org> |
---|---|
date | Wed, 22 Mar 1995 17:44:18 +0000 |
parents | 507f64624555 |
children | 83f275dcd93a |
rev | line source |
---|---|
660
08eb386dd0f3
*** empty log message ***
Eric S. Raymond <esr@snark.thyrsus.com>
parents:
584
diff
changeset
|
1 ;;; float.el --- floating point arithmetic package. |
08eb386dd0f3
*** empty log message ***
Eric S. Raymond <esr@snark.thyrsus.com>
parents:
584
diff
changeset
|
2 |
845 | 3 ;; Copyright (C) 1986 Free Software Foundation, Inc. |
4 | |
807
4f28bd14272c
*** empty log message ***
Eric S. Raymond <esr@snark.thyrsus.com>
parents:
660
diff
changeset
|
5 ;; Author: Bill Rosenblatt |
4f28bd14272c
*** empty log message ***
Eric S. Raymond <esr@snark.thyrsus.com>
parents:
660
diff
changeset
|
6 ;; Maintainer: FSF |
811
e694e0879463
*** empty log message ***
Eric S. Raymond <esr@snark.thyrsus.com>
parents:
807
diff
changeset
|
7 ;; Keywords: extensions |
807
4f28bd14272c
*** empty log message ***
Eric S. Raymond <esr@snark.thyrsus.com>
parents:
660
diff
changeset
|
8 |
59 | 9 ;; This file is part of GNU Emacs. |
10 | |
11 ;; GNU Emacs is free software; you can redistribute it and/or modify | |
12 ;; it under the terms of the GNU General Public License as published by | |
807
4f28bd14272c
*** empty log message ***
Eric S. Raymond <esr@snark.thyrsus.com>
parents:
660
diff
changeset
|
13 ;; the Free Software Foundation; either version 2, or (at your option) |
59 | 14 ;; any later version. |
15 | |
16 ;; GNU Emacs is distributed in the hope that it will be useful, | |
17 ;; but WITHOUT ANY WARRANTY; without even the implied warranty of | |
18 ;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
19 ;; GNU General Public License for more details. | |
20 | |
21 ;; You should have received a copy of the GNU General Public License | |
22 ;; along with GNU Emacs; see the file COPYING. If not, write to | |
23 ;; the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. | |
24 | |
807
4f28bd14272c
*** empty log message ***
Eric S. Raymond <esr@snark.thyrsus.com>
parents:
660
diff
changeset
|
25 ;;; Commentary: |
4f28bd14272c
*** empty log message ***
Eric S. Raymond <esr@snark.thyrsus.com>
parents:
660
diff
changeset
|
26 |
59 | 27 ;; Floating point numbers are represented by dot-pairs (mant . exp) |
28 ;; where mant is the 24-bit signed integral mantissa and exp is the | |
29 ;; base 2 exponent. | |
30 ;; | |
31 ;; Emacs LISP supports a 24-bit signed integer data type, which has a | |
32 ;; range of -(2**23) to +(2**23)-1, or -8388608 to 8388607 decimal. | |
33 ;; This gives six significant decimal digit accuracy. Exponents can | |
34 ;; be anything in the range -(2**23) to +(2**23)-1. | |
35 ;; | |
36 ;; User interface: | |
37 ;; function f converts from integer to floating point | |
38 ;; function string-to-float converts from string to floating point | |
39 ;; function fint converts a floating point to integer (with truncation) | |
40 ;; function float-to-string converts from floating point to string | |
41 ;; | |
42 ;; Caveats: | |
43 ;; - Exponents outside of the range of +/-100 or so will cause certain | |
44 ;; functions (especially conversion routines) to take forever. | |
45 ;; - Very little checking is done for fixed point overflow/underflow. | |
46 ;; - No checking is done for over/underflow of the exponent | |
47 ;; (hardly necessary when exponent can be 2**23). | |
48 ;; | |
49 ;; | |
50 ;; Bill Rosenblatt | |
51 ;; June 20, 1986 | |
52 ;; | |
53 | |
807
4f28bd14272c
*** empty log message ***
Eric S. Raymond <esr@snark.thyrsus.com>
parents:
660
diff
changeset
|
54 ;;; Code: |
4f28bd14272c
*** empty log message ***
Eric S. Raymond <esr@snark.thyrsus.com>
parents:
660
diff
changeset
|
55 |
59 | 56 ;; fundamental implementation constants |
57 (defconst exp-base 2 | |
58 "Base of exponent in this floating point representation.") | |
59 | |
60 (defconst mantissa-bits 24 | |
61 "Number of significant bits in this floating point representation.") | |
62 | |
63 (defconst decimal-digits 6 | |
64 "Number of decimal digits expected to be accurate.") | |
65 | |
66 (defconst expt-digits 2 | |
67 "Maximum permitted digits in a scientific notation exponent.") | |
68 | |
69 ;; other constants | |
70 (defconst maxbit (1- mantissa-bits) | |
71 "Number of highest bit") | |
72 | |
73 (defconst mantissa-maxval (1- (ash 1 maxbit)) | |
3591
507f64624555
Apply typo patches from Paul Eggert.
Jim Blandy <jimb@redhat.com>
parents:
845
diff
changeset
|
74 "Maximum permissible value of mantissa") |
59 | 75 |
76 (defconst mantissa-minval (ash 1 maxbit) | |
3591
507f64624555
Apply typo patches from Paul Eggert.
Jim Blandy <jimb@redhat.com>
parents:
845
diff
changeset
|
77 "Minimum permissible value of mantissa") |
59 | 78 |
79 (defconst floating-point-regexp | |
80 "^[ \t]*\\(-?\\)\\([0-9]*\\)\ | |
81 \\(\\.\\([0-9]*\\)\\|\\)\ | |
82 \\(\\(\\([Ee]\\)\\(-?\\)\\([0-9][0-9]*\\)\\)\\|\\)[ \t]*$" | |
83 "Regular expression to match floating point numbers. Extract matches: | |
84 1 - minus sign | |
85 2 - integer part | |
86 4 - fractional part | |
87 8 - minus sign for power of ten | |
88 9 - power of ten | |
89 ") | |
90 | |
91 (defconst high-bit-mask (ash 1 maxbit) | |
92 "Masks all bits except the high-order (sign) bit.") | |
93 | |
94 (defconst second-bit-mask (ash 1 (1- maxbit)) | |
95 "Masks all bits except the highest-order magnitude bit") | |
96 | |
97 ;; various useful floating point constants | |
98 (setq _f0 '(0 . 1)) | |
99 | |
100 (setq _f1/2 '(4194304 . -23)) | |
101 | |
102 (setq _f1 '(4194304 . -22)) | |
103 | |
104 (setq _f10 '(5242880 . -19)) | |
105 | |
106 ;; support for decimal conversion routines | |
107 (setq powers-of-10 (make-vector (1+ decimal-digits) _f1)) | |
108 (aset powers-of-10 1 _f10) | |
109 (aset powers-of-10 2 '(6553600 . -16)) | |
110 (aset powers-of-10 3 '(8192000 . -13)) | |
111 (aset powers-of-10 4 '(5120000 . -9)) | |
112 (aset powers-of-10 5 '(6400000 . -6)) | |
113 (aset powers-of-10 6 '(8000000 . -3)) | |
114 | |
115 (setq all-decimal-digs-minval (aref powers-of-10 (1- decimal-digits)) | |
116 highest-power-of-10 (aref powers-of-10 decimal-digits)) | |
117 | |
118 (defun fashl (fnum) ; floating-point arithmetic shift left | |
119 (cons (ash (car fnum) 1) (1- (cdr fnum)))) | |
120 | |
121 (defun fashr (fnum) ; floating point arithmetic shift right | |
122 (cons (ash (car fnum) -1) (1+ (cdr fnum)))) | |
123 | |
124 (defun normalize (fnum) | |
125 (if (> (car fnum) 0) ; make sure next-to-highest bit is set | |
126 (while (zerop (logand (car fnum) second-bit-mask)) | |
127 (setq fnum (fashl fnum))) | |
128 (if (< (car fnum) 0) ; make sure highest bit is set | |
129 (while (zerop (logand (car fnum) high-bit-mask)) | |
130 (setq fnum (fashl fnum))) | |
131 (setq fnum _f0))) ; "standard 0" | |
132 fnum) | |
133 | |
134 (defun abs (n) ; integer absolute value | |
135 (if (>= n 0) n (- n))) | |
136 | |
137 (defun fabs (fnum) ; re-normalize after taking abs value | |
138 (normalize (cons (abs (car fnum)) (cdr fnum)))) | |
139 | |
140 (defun xor (a b) ; logical exclusive or | |
141 (and (or a b) (not (and a b)))) | |
142 | |
143 (defun same-sign (a b) ; two f-p numbers have same sign? | |
144 (not (xor (natnump (car a)) (natnump (car b))))) | |
145 | |
146 (defun extract-match (str i) ; used after string-match | |
147 (condition-case () | |
148 (substring str (match-beginning i) (match-end i)) | |
149 (error ""))) | |
150 | |
151 ;; support for the multiplication function | |
152 (setq halfword-bits (/ mantissa-bits 2) ; bits in a halfword | |
153 masklo (1- (ash 1 halfword-bits)) ; isolate the lower halfword | |
154 maskhi (lognot masklo) ; isolate the upper halfword | |
155 round-limit (ash 1 (/ halfword-bits 2))) | |
156 | |
157 (defun hihalf (n) ; return high halfword, shifted down | |
158 (ash (logand n maskhi) (- halfword-bits))) | |
159 | |
160 (defun lohalf (n) ; return low halfword | |
161 (logand n masklo)) | |
162 | |
163 ;; Visible functions | |
164 | |
165 ;; Arithmetic functions | |
166 (defun f+ (a1 a2) | |
167 "Returns the sum of two floating point numbers." | |
168 (let ((f1 (fmax a1 a2)) | |
169 (f2 (fmin a1 a2))) | |
170 (if (same-sign a1 a2) | |
171 (setq f1 (fashr f1) ; shift right to avoid overflow | |
172 f2 (fashr f2))) | |
173 (normalize | |
174 (cons (+ (car f1) (ash (car f2) (- (cdr f2) (cdr f1)))) | |
175 (cdr f1))))) | |
176 | |
177 (defun f- (a1 &optional a2) ; unary or binary minus | |
178 "Returns the difference of two floating point numbers." | |
179 (if a2 | |
180 (f+ a1 (f- a2)) | |
181 (normalize (cons (- (car a1)) (cdr a1))))) | |
182 | |
183 (defun f* (a1 a2) ; multiply in halfword chunks | |
184 "Returns the product of two floating point numbers." | |
185 (let* ((i1 (car (fabs a1))) | |
186 (i2 (car (fabs a2))) | |
187 (sign (not (same-sign a1 a2))) | |
188 (prodlo (+ (hihalf (* (lohalf i1) (lohalf i2))) | |
189 (lohalf (* (hihalf i1) (lohalf i2))) | |
190 (lohalf (* (lohalf i1) (hihalf i2))))) | |
191 (prodhi (+ (* (hihalf i1) (hihalf i2)) | |
192 (hihalf (* (hihalf i1) (lohalf i2))) | |
193 (hihalf (* (lohalf i1) (hihalf i2))) | |
194 (hihalf prodlo)))) | |
195 (if (> (lohalf prodlo) round-limit) | |
196 (setq prodhi (1+ prodhi))) ; round off truncated bits | |
197 (normalize | |
198 (cons (if sign (- prodhi) prodhi) | |
199 (+ (cdr (fabs a1)) (cdr (fabs a2)) mantissa-bits))))) | |
200 | |
201 (defun f/ (a1 a2) ; SLOW subtract-and-shift algorithm | |
202 "Returns the quotient of two floating point numbers." | |
203 (if (zerop (car a2)) ; if divide by 0 | |
204 (signal 'arith-error (list "attempt to divide by zero" a1 a2)) | |
205 (let ((bits (1- maxbit)) | |
206 (quotient 0) | |
207 (dividend (car (fabs a1))) | |
208 (divisor (car (fabs a2))) | |
209 (sign (not (same-sign a1 a2)))) | |
210 (while (natnump bits) | |
211 (if (< (- dividend divisor) 0) | |
212 (setq quotient (ash quotient 1)) | |
213 (setq quotient (1+ (ash quotient 1)) | |
214 dividend (- dividend divisor))) | |
215 (setq dividend (ash dividend 1) | |
216 bits (1- bits))) | |
217 (normalize | |
218 (cons (if sign (- quotient) quotient) | |
219 (- (cdr (fabs a1)) (cdr (fabs a2)) (1- maxbit))))))) | |
220 | |
221 (defun f% (a1 a2) | |
222 "Returns the remainder of first floating point number divided by second." | |
223 (f- a1 (f* (ftrunc (f/ a1 a2)) a2))) | |
224 | |
225 | |
226 ;; Comparison functions | |
227 (defun f= (a1 a2) | |
228 "Returns t if two floating point numbers are equal, nil otherwise." | |
229 (equal a1 a2)) | |
230 | |
231 (defun f> (a1 a2) | |
232 "Returns t if first floating point number is greater than second, | |
233 nil otherwise." | |
234 (cond ((and (natnump (car a1)) (< (car a2) 0)) | |
235 t) ; a1 nonnegative, a2 negative | |
236 ((and (> (car a1) 0) (<= (car a2) 0)) | |
237 t) ; a1 positive, a2 nonpositive | |
238 ((and (<= (car a1) 0) (natnump (car a2))) | |
239 nil) ; a1 nonpos, a2 nonneg | |
240 ((/= (cdr a1) (cdr a2)) ; same signs. exponents differ | |
241 (> (cdr a1) (cdr a2))) ; compare the mantissas. | |
242 (t | |
243 (> (car a1) (car a2))))) ; same exponents. | |
244 | |
245 (defun f>= (a1 a2) | |
246 "Returns t if first floating point number is greater than or equal to | |
247 second, nil otherwise." | |
248 (or (f> a1 a2) (f= a1 a2))) | |
249 | |
250 (defun f< (a1 a2) | |
251 "Returns t if first floating point number is less than second, | |
252 nil otherwise." | |
253 (not (f>= a1 a2))) | |
254 | |
255 (defun f<= (a1 a2) | |
256 "Returns t if first floating point number is less than or equal to | |
257 second, nil otherwise." | |
258 (not (f> a1 a2))) | |
259 | |
260 (defun f/= (a1 a2) | |
261 "Returns t if first floating point number is not equal to second, | |
262 nil otherwise." | |
263 (not (f= a1 a2))) | |
264 | |
265 (defun fmin (a1 a2) | |
266 "Returns the minimum of two floating point numbers." | |
267 (if (f< a1 a2) a1 a2)) | |
268 | |
269 (defun fmax (a1 a2) | |
270 "Returns the maximum of two floating point numbers." | |
271 (if (f> a1 a2) a1 a2)) | |
272 | |
273 (defun fzerop (fnum) | |
274 "Returns t if the floating point number is zero, nil otherwise." | |
275 (= (car fnum) 0)) | |
276 | |
277 (defun floatp (fnum) | |
278 "Returns t if the arg is a floating point number, nil otherwise." | |
279 (and (consp fnum) (integerp (car fnum)) (integerp (cdr fnum)))) | |
280 | |
281 ;; Conversion routines | |
282 (defun f (int) | |
283 "Convert the integer argument to floating point, like a C cast operator." | |
284 (normalize (cons int '0))) | |
285 | |
286 (defun int-to-hex-string (int) | |
287 "Convert the integer argument to a C-style hexadecimal string." | |
288 (let ((shiftval -20) | |
289 (str "0x") | |
290 (hex-chars "0123456789ABCDEF")) | |
291 (while (<= shiftval 0) | |
292 (setq str (concat str (char-to-string | |
293 (aref hex-chars | |
294 (logand (lsh int shiftval) 15)))) | |
295 shiftval (+ shiftval 4))) | |
296 str)) | |
297 | |
298 (defun ftrunc (fnum) ; truncate fractional part | |
299 "Truncate the fractional part of a floating point number." | |
300 (cond ((natnump (cdr fnum)) ; it's all integer, return number as is | |
301 fnum) | |
302 ((<= (cdr fnum) (- maxbit)) ; it's all fractional, return 0 | |
303 '(0 . 1)) | |
304 (t ; otherwise mask out fractional bits | |
305 (let ((mant (car fnum)) (exp (cdr fnum))) | |
306 (normalize | |
307 (cons (if (natnump mant) ; if negative, use absolute value | |
308 (ash (ash mant exp) (- exp)) | |
309 (- (ash (ash (- mant) exp) (- exp)))) | |
310 exp)))))) | |
311 | |
312 (defun fint (fnum) ; truncate and convert to integer | |
313 "Convert the floating point number to integer, with truncation, | |
314 like a C cast operator." | |
315 (let* ((tf (ftrunc fnum)) (tint (car tf)) (texp (cdr tf))) | |
316 (cond ((>= texp mantissa-bits) ; too high, return "maxint" | |
317 mantissa-maxval) | |
318 ((<= texp (- mantissa-bits)) ; too low, return "minint" | |
319 mantissa-minval) | |
320 (t ; in range | |
321 (ash tint texp))))) ; shift so that exponent is 0 | |
322 | |
323 (defun float-to-string (fnum &optional sci) | |
324 "Convert the floating point number to a decimal string. | |
325 Optional second argument non-nil means use scientific notation." | |
326 (let* ((value (fabs fnum)) (sign (< (car fnum) 0)) | |
327 (power 0) (result 0) (str "") | |
328 (temp 0) (pow10 _f1)) | |
329 | |
330 (if (f= fnum _f0) | |
331 "0" | |
332 (if (f>= value _f1) ; find largest power of 10 <= value | |
333 (progn ; value >= 1, power is positive | |
334 (while (f<= (setq temp (f* pow10 highest-power-of-10)) value) | |
335 (setq pow10 temp | |
336 power (+ power decimal-digits))) | |
337 (while (f<= (setq temp (f* pow10 _f10)) value) | |
338 (setq pow10 temp | |
339 power (1+ power)))) | |
340 (progn ; value < 1, power is negative | |
341 (while (f> (setq temp (f/ pow10 highest-power-of-10)) value) | |
342 (setq pow10 temp | |
343 power (- power decimal-digits))) | |
344 (while (f> pow10 value) | |
345 (setq pow10 (f/ pow10 _f10) | |
346 power (1- power))))) | |
347 ; get value in range 100000 to 999999 | |
348 (setq value (f* (f/ value pow10) all-decimal-digs-minval) | |
349 result (ftrunc value)) | |
350 (let (int) | |
351 (if (f> (f- value result) _f1/2) ; round up if remainder > 0.5 | |
352 (setq int (1+ (fint result))) | |
353 (setq int (fint result))) | |
354 (setq str (int-to-string int)) | |
355 (if (>= int 1000000) | |
356 (setq power (1+ power)))) | |
357 | |
358 (if sci ; scientific notation | |
359 (setq str (concat (substring str 0 1) "." (substring str 1) | |
360 "E" (int-to-string power))) | |
361 | |
362 ; regular decimal string | |
363 (cond ((>= power (1- decimal-digits)) | |
364 ; large power, append zeroes | |
365 (let ((zeroes (- power decimal-digits))) | |
366 (while (natnump zeroes) | |
367 (setq str (concat str "0") | |
368 zeroes (1- zeroes))))) | |
369 | |
370 ; negative power, prepend decimal | |
371 ((< power 0) ; point and zeroes | |
372 (let ((zeroes (- (- power) 2))) | |
373 (while (natnump zeroes) | |
374 (setq str (concat "0" str) | |
375 zeroes (1- zeroes))) | |
376 (setq str (concat "0." str)))) | |
377 | |
378 (t ; in range, insert decimal point | |
379 (setq str (concat | |
380 (substring str 0 (1+ power)) | |
381 "." | |
382 (substring str (1+ power))))))) | |
383 | |
384 (if sign ; if negative, prepend minus sign | |
385 (concat "-" str) | |
386 str)))) | |
387 | |
388 | |
389 ;; string to float conversion. | |
390 ;; accepts scientific notation, but ignores anything after the first two | |
391 ;; digits of the exponent. | |
392 (defun string-to-float (str) | |
393 "Convert the string to a floating point number. | |
202 | 394 Accepts a decimal string in scientific notation, with exponent preceded |
395 by either E or e. Only the six most significant digits of the integer | |
396 and fractional parts are used; only the first two digits of the exponent | |
397 are used. Negative signs preceding both the decimal number and the exponent | |
59 | 398 are recognized." |
399 | |
400 (if (string-match floating-point-regexp str 0) | |
401 (let (power) | |
402 (f* | |
403 ; calculate the mantissa | |
404 (let* ((int-subst (extract-match str 2)) | |
405 (fract-subst (extract-match str 4)) | |
406 (digit-string (concat int-subst fract-subst)) | |
407 (mant-sign (equal (extract-match str 1) "-")) | |
408 (leading-0s 0) (round-up nil)) | |
409 | |
410 ; get rid of leading 0's | |
411 (setq power (- (length int-subst) decimal-digits)) | |
412 (while (and (< leading-0s (length digit-string)) | |
413 (= (aref digit-string leading-0s) ?0)) | |
414 (setq leading-0s (1+ leading-0s))) | |
415 (setq power (- power leading-0s) | |
416 digit-string (substring digit-string leading-0s)) | |
417 | |
418 ; if more than 6 digits, round off | |
419 (if (> (length digit-string) decimal-digits) | |
420 (setq round-up (>= (aref digit-string decimal-digits) ?5) | |
421 digit-string (substring digit-string 0 decimal-digits)) | |
422 (setq power (+ power (- decimal-digits (length digit-string))))) | |
423 | |
424 ; round up and add minus sign, if necessary | |
425 (f (* (+ (string-to-int digit-string) | |
426 (if round-up 1 0)) | |
427 (if mant-sign -1 1)))) | |
428 | |
429 ; calculate the exponent (power of ten) | |
430 (let* ((expt-subst (extract-match str 9)) | |
431 (expt-sign (equal (extract-match str 8) "-")) | |
432 (expt 0) (chunks 0) (tens 0) (exponent _f1) | |
433 (func 'f*)) | |
434 | |
435 (setq expt (+ (* (string-to-int | |
436 (substring expt-subst 0 | |
437 (min expt-digits (length expt-subst)))) | |
438 (if expt-sign -1 1)) | |
439 power)) | |
440 (if (< expt 0) ; if power of 10 negative | |
441 (setq expt (- expt) ; take abs val of exponent | |
442 func 'f/)) ; and set up to divide, not multiply | |
443 | |
444 (setq chunks (/ expt decimal-digits) | |
445 tens (% expt decimal-digits)) | |
446 ; divide or multiply by "chunks" of 10**6 | |
447 (while (> chunks 0) | |
448 (setq exponent (funcall func exponent highest-power-of-10) | |
449 chunks (1- chunks))) | |
450 ; divide or multiply by remaining power of ten | |
451 (funcall func exponent (aref powers-of-10 tens))))) | |
452 | |
453 _f0)) ; if invalid, return 0 | |
584 | 454 |
455 (provide 'float) | |
456 | |
660
08eb386dd0f3
*** empty log message ***
Eric S. Raymond <esr@snark.thyrsus.com>
parents:
584
diff
changeset
|
457 ;;; float.el ends here |