New version of the Plotting section of the manual reflecting recent changes
[maxima.git] / src / numeric.lisp
blobc2a5b680a69137a6c25af838247906e1dbcb5601
1 ;;; -*- Mode: Lisp; Package: Maxima; Syntax: Common-Lisp; Base: 10-*- ;;;;
3 ;;; This package contains a numeric class for use with Maxima. The
4 ;;; purpose is to allow users to write numerical algorithms that
5 ;;; support double-float, (complex double-float) and Maxima bfloat and
6 ;;; complex bfloat arithmetic, without having to write separate
7 ;;; versions for each. Of course, specially written versions for
8 ;;; double-float and (complex double-float) will probably be much
9 ;;; faster, but this allows users to write just one routine that can
10 ;;; handle all of the data types in a more "natural" Lisp style.
12 #+cmu
13 (eval-when (:compile-toplevel :load-toplevel :execute)
14 (setf lisp::*enable-package-locked-errors* nil))
16 (in-package #-gcl #:bigfloat #+gcl "BIGFLOAT")
18 (defun intofp (re)
19 ;; Kind of like Maxima's INTOFP, but we only handle numeric types.
20 ;; We should return a Maxima bigfloat object (list of bigfloat
21 ;; marker, mantissa, and exponent).
22 (cond ((floatp re)
23 (maxima::bcons (maxima::floattofp re)))
24 ((eql re 0)
25 (maxima::bcons '(0 0)))
26 ((integerp re)
27 (maxima::bcons (list (maxima::fpround re) (cl:+ maxima::*m maxima::fpprec))))
28 ((typep re 'ratio)
29 ;; Should we do something better than converting the
30 ;; numerator and denominator to floats and dividing?
31 (maxima::bcons (maxima::fpquotient (cdr (intofp (numerator re)))
32 (cdr (intofp (denominator re))))))
33 ((maxima::$bfloatp re)
34 ;; Call bigfloatp to make sure we round/scale the bigfloat to
35 ;; the correct precision!
36 (maxima::bigfloatp re))
38 (error "Don't know how to convert ~S to a BIGFLOAT" re))))
40 (defclass numeric ()
42 (:documentation "Basic class, like CL's NUMBER type"))
44 (defclass bigfloat (numeric)
45 ;; We store the Maxima internal bigfloat format because we also need
46 ;; the precision in case we have mixed size bigfloat operations.
47 ;; (We could recompute it from the size of the mantissa part, but
48 ;; why bother?
49 ((real :initform (intofp 0)
50 :initarg :real
51 :documentation "A Maxima bigfloat. This contains the full
52 Maxima bigfloat object including the mantissa, the exponent
53 and the bigfloat marker and precision." ))
54 (:documentation "Big float, equivalent to a Maxima bfloat object"))
56 ;; Extract the internal representation of a bigfloat, and adjust the
57 ;; precision to the current value of fpprec.
58 (defmethod real-value ((b bigfloat))
59 (maxima::bigfloatp (slot-value b 'real)))
61 (defclass complex-bigfloat (numeric)
62 ;; Currently, the real and imaginary parts contain a Maxima bigfloat
63 ;; including the bigfloat marker and the mantissa and exponent.
64 ;; Should they be BIGFLOAT objects instead?
65 ((real :initform (intofp 0)
66 :initarg :real
67 :documentation "Real part of a complex bigfloat")
68 (imag :initform (intofp 0)
69 :initarg :imag
70 :documentation "Imaginary part of a complex bigfloat"))
71 (:documentation "Complex bigfloat"))
73 ;; Extract the internal representation of the real part of a
74 ;; complex-bigfloat, and adjust the precision to the current value of
75 ;; fpprec.
76 (defmethod real-value ((b complex-bigfloat))
77 (maxima::bigfloatp (slot-value b 'real)))
79 ;; Extract the internal representation of the imaginary part of a
80 ;; complex-bigfloat, and adjust the precision to the current value of
81 ;; fpprec.
82 (defmethod imag-value ((b complex-bigfloat))
83 (maxima::bigfloatp (slot-value b 'imag)))
85 (defmethod make-load-form ((x bigfloat) &optional environment)
86 (declare (ignore environment))
87 `(make-instance ',(class-of x)
88 :real ',(real-value x)))
90 ;;; BIGFLOAT - External
91 ;;;
92 ;;; BIGFLOAT converts a number to a BIGFLOAT or COMPLEX-BIGFLOAT.
93 ;;; This is intended to convert CL numbers or Maxima (internal)
94 ;;; numbers to a bigfloat object.
95 (defun bigfloat (re &optional im)
96 "Convert RE to a BIGFLOAT. If IM is given, return a COMPLEX-BIGFLOAT"
97 (cond (im
98 (make-instance 'complex-bigfloat
99 :real (intofp re)
100 :imag (intofp im)))
101 ((cl:realp re)
102 (make-instance 'bigfloat :real (intofp re)))
103 ((cl:complexp re)
104 (make-instance 'complex-bigfloat
105 :real (intofp (cl:realpart re))
106 :imag (intofp (cl:imagpart re))))
107 ((maxima::$bfloatp re)
108 (make-instance 'bigfloat :real (intofp re)))
109 ((maxima::complex-number-p re 'maxima::bigfloat-or-number-p)
110 (make-instance 'complex-bigfloat
111 :real (intofp (maxima::$realpart re))
112 :imag (intofp (maxima::$imagpart re))))
113 ((typep re 'bigfloat)
114 ;; Done this way so the new bigfloat updates the precision of
115 ;; the given bigfloat, if necessary.
116 (make-instance 'bigfloat :real (real-value re)))
117 ((typep re 'complex-bigfloat)
118 ;; Done this way so the new bigfloat updates the precision of
119 ;; the given bigfloat, if necessary.
120 (make-instance 'complex-bigfloat
121 :real (real-value re)
122 :imag (imag-value re)))
124 (make-instance 'bigfloat :real (intofp re)))))
127 ;;; MAXIMA::TO - External
129 ;;; Convert a CL number, a BIGFLOAT, or a COMPLEX-BIGFLOAT to
130 ;;; Maxima's internal representation of the number.
131 (defmethod maxima::to ((z cl:float))
134 (defmethod maxima::to ((z cl:rational))
135 (if (typep z 'ratio)
136 (list '(maxima::rat maxima::simp) (numerator z) (denominator z))
139 (defmethod maxima::to ((z cl:complex))
140 (maxima::add (maxima::to (cl:realpart z))
141 (maxima::mul 'maxima::$%i
142 (maxima::to (cl:imagpart z)))))
144 (defmethod maxima::to ((z bigfloat))
145 "Convert BIGFLOAT object to a maxima number"
146 (real-value z))
148 (defmethod maxima::to ((z complex-bigfloat))
149 "Convert COMPLEX-BIGFLOAT object to a maxima number"
150 (maxima::add (real-value z)
151 (maxima::mul 'maxima::$%i
152 (imag-value z))))
154 (defmethod maxima::to ((z t))
157 ;; MAX-EXPONENT roughly computes the log2(|x|). If x is real and x =
158 ;; 2^n*f, with |f| < 1, MAX-EXPONENT returns |n|. For complex
159 ;; numbers, we return one more than the max of the exponent of the
160 ;; real and imaginary parts.
161 (defmethod max-exponent ((x bigfloat))
162 ;; The third element is the exponent of a bigfloat.
163 (cl:abs (third (slot-value x 'real))))
165 (defmethod max-exponent ((x complex-bigfloat))
166 (cl:1+ (cl:max (cl:abs (third (slot-value x 'real)))
167 (cl:abs (third (slot-value x 'imag))))))
169 (defmethod max-exponent ((x cl:float))
170 (if (zerop x)
172 (cl:abs (nth-value 1 (cl:decode-float x)))))
174 (defmethod max-exponent ((x cl:rational))
175 (if (zerop x)
177 (cl:ceiling (cl:log (cl:abs x) 2))))
179 (defmethod max-exponent ((x cl:complex))
180 (cl:1+ (cl:max (max-exponent (cl:realpart x))
181 (max-exponent (cl:imagpart x)))))
183 ;; When computing x^a using exp(a*log(x)), we need extra bits because
184 ;; the integer part of a*log(x) doesn't contribute to the accuracy of
185 ;; the result. The number of extra bits needed is basically the
186 ;; "size" of a plus the number of bits for ceiling(log(x)). We need
187 ;; ceiling(log(x)) extra bits because that's how many bits are taken
188 ;; up by the log(x). The "size" of a is, basically, the exponent of
189 ;; a. If a = 2^n*f where |f| < 1, then the size is abs(n) because
190 ;; that's how many extra bits are added to the integer part of
191 ;; a*log(x). If either |x| or |a| < 1, the size is 0, since no
192 ;; additional bits are taken up.
193 (defun expt-extra-bits (x a)
194 (max 1 (+ (integer-length (max-exponent x))
195 (max-exponent a))))
197 ;;; WITH-EXTRA-PRECISION - Internal
199 ;;; Executes the body BODY with extra precision. The precision is
200 ;;; increased by EXTRA, and the list of variables given in VARLIST have
201 ;;; the precision increased. The precision of the first value of the
202 ;;; body is then reduced back to the normal precision.
203 (defmacro with-extra-precision ((extra (&rest varlist)) &body body)
204 (let ((result (gensym))
205 (old-fpprec (gensym)))
206 `(let ((,result
207 (let ((,old-fpprec maxima::fpprec))
208 (unwind-protect
209 (let ((maxima::fpprec (cl:+ maxima::fpprec ,extra)))
210 (let ,(mapcar #'(lambda (v)
211 ;; Could probably do this in a faster
212 ;; way, but conversion to a maxima
213 ;; form automatically increases the
214 ;; precision of the bigfloat to the
215 ;; new precision. Conversion of that
216 ;; to a bigfloat object preserves the
217 ;; precision.
218 `(,v (bigfloat:to (maxima::to ,v))))
219 varlist)
220 ,@body))
221 (setf maxima::fpprec ,old-fpprec)))))
222 ;; Conversion of the result to a maxima number adjusts the
223 ;; precision appropriately.
224 (bigfloat:to (maxima::to ,result)))))
226 ;;; REALP
228 ;; GCL doesn't have the REAL class! But since a real is a rational or
229 ;; float, we can fake it by defining methods on rationals and floats
230 ;; for gcl.
231 #-gcl
232 (defmethod realp ((x cl:real))
235 #+gcl
236 (progn
237 (defmethod realp ((x cl:rational))
239 (defmethod realp ((x cl:float))
244 (defmethod realp ((x bigfloat))
247 (defmethod realp ((x t))
248 nil)
250 ;;; COMPLEXP
251 (defmethod complexp ((x cl:complex))
254 (defmethod complexp ((x complex-bigfloat))
257 (defmethod complexp ((x t))
258 nil)
260 ;;; NUMBERP
261 (defmethod numberp ((x cl:number))
264 (defmethod numberp ((x bigfloat))
267 (defmethod numberp ((x complex-bigfloat))
270 (defmethod numberp ((x t))
271 nil)
273 (defmethod make-load-form ((x complex-bigfloat) &optional environment)
274 (declare (ignore environment))
275 `(make-instance ',(class-of x)
276 :real ',(real-value x)
277 :imag ',(imag-value x)))
279 ;; The print-object and describe-object methods are mostly for
280 ;; debugging purposes. Maxima itself shouldn't ever see such objects.
281 (defmethod print-object ((x bigfloat) stream)
282 (let ((r (cdr (real-value x))))
283 (multiple-value-bind (sign output-list)
284 (if (cl:minusp (first r))
285 (values "-" (maxima::fpformat (maxima::bcons (list (cl:- (first r)) (second r)))))
286 (values "+" (maxima::fpformat (maxima::bcons r))))
287 (format stream "~A~{~D~}" sign output-list))))
289 (defmethod print-object ((x complex-bigfloat) stream)
290 (format stream "~A~A*%i" (realpart x) (imagpart x)))
292 (defmethod describe-object ((x bigfloat) stream)
293 (let ((r (slot-value x 'real)))
294 (format stream "~&~S is a ~D-bit BIGFLOAT with mantissa ~D and exponent ~D~%"
295 x (third (first r)) (second r) (third r))))
297 (defmethod describe-object ((x complex-bigfloat) stream)
298 (format stream "~S is a COMPLEX-BIGFLOAT~%" x)
299 (describe-object (make-instance 'bigfloat :real (slot-value x 'real)) stream)
300 (describe-object (make-instance 'bigfloat :real (slot-value x 'imag)) stream))
303 (defgeneric add1 (a)
304 (:documentation "Add 1"))
306 (defgeneric sub1 (a)
307 (:documentation "Subtract 1"))
310 (defgeneric two-arg-+ (a b)
311 (:documentation "A + B"))
313 (defgeneric two-arg-- (a b)
314 (:documentation "A - B"))
316 (defgeneric two-arg-* (a b)
317 (:documentation "A * B"))
319 (defgeneric two-arg-/ (a b)
320 (:documentation "A / B"))
322 (defgeneric two-arg-< (a b)
323 (:documentation "A < B"))
325 (defgeneric two-arg-> (a b)
326 (:documentation "A > B"))
328 (defgeneric two-arg-<= (a b)
329 (:documentation "A <= B"))
331 (defgeneric two-arg->= (a b)
332 (:documentation "A >= B"))
334 (defgeneric two-arg-= (a b)
335 (:documentation "A = B?"))
338 (defgeneric unary-minus (a)
339 (:documentation "-A"))
341 (defgeneric unary-divide (a)
342 (:documentation "1 / A"))
345 ;;; Basic arithmetic operations
347 ;;; 1+ and 1-
349 (defmethod add1 ((a number))
350 (cl::1+ a))
352 (defmethod add1 ((a bigfloat))
353 (make-instance 'bigfloat
354 :real (maxima::bcons
355 (maxima::fpplus (cdr (real-value a))
356 (maxima::fpone)))))
358 (defmethod add1 ((a complex-bigfloat))
359 (make-instance 'complex-bigfloat
360 :real (maxima::bcons
361 (maxima::fpplus (cdr (real-value a))
362 (maxima::fpone)))
363 :imag (imag-value a)))
365 (defmethod sub1 ((a number))
366 (cl::1- a))
368 (defmethod sub1 ((a bigfloat))
369 (make-instance 'bigfloat
370 :real (maxima::bcons
371 (maxima::fpdifference (cdr (real-value a))
372 (maxima::fpone)))))
374 (defmethod sub1 ((a complex-bigfloat))
375 (make-instance 'complex-bigfloat
376 :real (maxima::bcons
377 (maxima::fpdifference (cdr (real-value a))
378 (maxima::fpone)))
379 :imag (imag-value a)))
381 (declaim (inline 1+ 1-))
383 (defun 1+ (x)
384 (add1 x))
386 (defun 1- (x)
387 (sub1 x))
389 ;; Add two numbers
390 (defmethod two-arg-+ ((a cl:number) (b cl:number))
391 (cl:+ a b))
393 (defmethod two-arg-+ ((a bigfloat) (b bigfloat))
394 (make-instance 'bigfloat
395 :real (maxima::bcons
396 (maxima::fpplus (cdr (real-value a))
397 (cdr (real-value b))))))
399 (defmethod two-arg-+ ((a complex-bigfloat) (b complex-bigfloat))
400 (make-instance 'complex-bigfloat
401 :real (maxima::bcons
402 (maxima::fpplus (cdr (real-value a))
403 (cdr (real-value b))))
404 :imag (maxima::bcons
405 (maxima::fpplus (cdr (imag-value a))
406 (cdr (imag-value b))))))
408 ;; Handle contagion for two-arg-+
409 (defmethod two-arg-+ ((a bigfloat) (b cl:float))
410 (make-instance 'bigfloat
411 :real (maxima::bcons
412 (maxima::fpplus (cdr (real-value a))
413 (cdr (intofp b))))))
415 (defmethod two-arg-+ ((a bigfloat) (b cl:rational))
416 (make-instance 'bigfloat
417 :real (maxima::bcons
418 (maxima::fpplus (cdr (real-value a))
419 (cdr (intofp b))))))
421 (defmethod two-arg-+ ((a bigfloat) (b cl:complex))
422 (make-instance 'complex-bigfloat
423 :real (maxima::bcons
424 (maxima::fpplus (cdr (real-value a))
425 (cdr (intofp (realpart b)))))
426 :imag (intofp (imagpart b))))
428 (defmethod two-arg-+ ((a cl:number) (b bigfloat))
429 (two-arg-+ b a))
431 (defmethod two-arg-+ ((a complex-bigfloat) (b bigfloat))
432 (make-instance 'complex-bigfloat
433 :real (maxima::bcons
434 (maxima::fpplus (cdr (real-value a))
435 (cdr (real-value b))))
436 :imag (imag-value a)))
438 (defmethod two-arg-+ ((a complex-bigfloat) (b number))
439 (make-instance 'complex-bigfloat
440 :real (maxima::bcons
441 (maxima::fpplus (cdr (real-value a))
442 (cdr (intofp (cl:realpart b)))))
443 :imag (maxima::bcons
444 (maxima::fpplus (cdr (imag-value a))
445 (cdr (intofp (cl:imagpart b)))))))
447 (defmethod two-arg-+ ((a bigfloat) (b complex-bigfloat))
448 (two-arg-+ b a))
450 (defmethod two-arg-+ ((a number) (b complex-bigfloat))
451 (two-arg-+ b a))
453 (defun + (&rest args)
454 (if (null args)
456 (do ((args (cdr args) (cdr args))
457 (res (car args)
458 (two-arg-+ res (car args))))
459 ((null args) res))))
461 ;; Negate a number
462 (defmethod unary-minus ((a number))
463 (cl:- a))
465 (defmethod unary-minus ((a bigfloat))
466 (make-instance 'bigfloat
467 :real (maxima::bcons
468 (maxima::fpminus (cdr (real-value a))))))
470 (defmethod unary-minus ((a complex-bigfloat))
471 (make-instance 'complex-bigfloat
472 :real (maxima::bcons
473 (maxima::fpminus (cdr (real-value a))))
474 :imag (maxima::bcons
475 (maxima::fpminus (cdr (imag-value a))))))
477 ;;; Subtract two numbers
478 (defmethod two-arg-- ((a number) (b number))
479 (cl:- a b))
481 (defmethod two-arg-- ((a bigfloat) (b bigfloat))
482 (make-instance 'bigfloat
483 :real (maxima::bcons
484 (maxima::fpdifference (cdr (real-value a))
485 (cdr (real-value b))))))
487 (defmethod two-arg-- ((a complex-bigfloat) (b complex-bigfloat))
488 (make-instance 'complex-bigfloat
489 :real (maxima::bcons
490 (maxima::fpdifference (cdr (real-value a))
491 (cdr (real-value b))))
492 :imag (maxima::bcons
493 (maxima::fpdifference (cdr (imag-value a))
494 (cdr (imag-value b))))))
496 ;; Handle contagion for two-arg--
497 (defmethod two-arg-- ((a bigfloat) (b cl:float))
498 (make-instance 'bigfloat
499 :real (maxima::bcons
500 (maxima::fpdifference (cdr (real-value a))
501 (cdr (intofp b))))))
503 (defmethod two-arg-- ((a bigfloat) (b cl:rational))
504 (make-instance 'bigfloat
505 :real (maxima::bcons
506 (maxima::fpdifference (cdr (real-value a))
507 (cdr (intofp b))))))
509 (defmethod two-arg-- ((a bigfloat) (b cl:complex))
510 (make-instance 'complex-bigfloat
511 :real (maxima::bcons
512 (maxima::fpdifference (cdr (real-value a))
513 (cdr (intofp (realpart b)))))
514 :imag (maxima::bcons (maxima::fpminus (cdr (intofp (imagpart b)))))))
516 (defmethod two-arg-- ((a cl:float) (b bigfloat))
517 (make-instance 'bigfloat
518 :real (maxima::bcons
519 (maxima::fpdifference (cdr (intofp a))
520 (cdr (real-value b))))))
522 (defmethod two-arg-- ((a cl:rational) (b bigfloat))
523 (make-instance 'bigfloat
524 :real (maxima::bcons
525 (maxima::fpdifference (cdr (intofp a))
526 (cdr (real-value b))))))
528 (defmethod two-arg-- ((a cl:complex) (b bigfloat))
529 (two-arg-- (bigfloat (cl:realpart a) (cl:imagpart a)) b))
531 (defmethod two-arg-- ((a complex-bigfloat) (b bigfloat))
532 (make-instance 'complex-bigfloat
533 :real (maxima::bcons
534 (maxima::fpdifference (cdr (real-value a))
535 (cdr (real-value b))))
536 :imag (imag-value a)))
538 (defmethod two-arg-- ((a complex-bigfloat) (b number))
539 (if (cl:complexp b)
540 (two-arg-- a (bigfloat (cl:realpart b) (cl:imagpart b)))
541 (two-arg-- a (bigfloat b))))
543 (defmethod two-arg-- ((a bigfloat) (b complex-bigfloat))
544 (make-instance 'complex-bigfloat
545 :real (maxima::bcons
546 (maxima::fpdifference (cdr (real-value a))
547 (cdr (real-value b))))
548 :imag (maxima::bcons (maxima::fpminus (cdr (imag-value b))))))
550 (defmethod two-arg-- ((a number) (b complex-bigfloat))
551 (if (cl:complexp a)
552 (two-arg-- (bigfloat (cl:realpart a) (cl:imagpart a))
554 (two-arg-- (bigfloat a) b)))
556 (defun - (number &rest more-numbers)
557 (if more-numbers
558 (do ((nlist more-numbers (cdr nlist))
559 (result number))
560 ((atom nlist) result)
561 (declare (list nlist))
562 (setq result (two-arg-- result (car nlist))))
563 (unary-minus number)))
565 ;;; Multiply two numbers
566 (defmethod two-arg-* ((a number) (b number))
567 (cl:* a b))
569 (defmethod two-arg-* ((a bigfloat) (b bigfloat))
570 (make-instance 'bigfloat
571 :real (maxima::bcons
572 (maxima::fptimes* (cdr (real-value a))
573 (cdr (real-value b))))))
575 (defmethod two-arg-* ((a complex-bigfloat) (b complex-bigfloat))
576 (let ((a-re (cdr (real-value a)))
577 (a-im (cdr (imag-value a)))
578 (b-re (cdr (real-value b)))
579 (b-im (cdr (imag-value b))))
580 (make-instance 'complex-bigfloat
581 :real (maxima::bcons
582 (maxima::fpdifference (maxima::fptimes* a-re b-re)
583 (maxima::fptimes* a-im b-im)))
584 :imag (maxima::bcons
585 (maxima::fpplus (maxima::fptimes* a-re b-im)
586 (maxima::fptimes* a-im b-re))))))
588 ;; Handle contagion for two-arg-*
589 (defmethod two-arg-* ((a bigfloat) (b cl:float))
590 (make-instance 'bigfloat
591 :real (maxima::bcons
592 (maxima::fptimes* (cdr (real-value a))
593 (cdr (intofp b))))))
595 (defmethod two-arg-* ((a bigfloat) (b cl:rational))
596 (make-instance 'bigfloat
597 :real (maxima::bcons
598 (maxima::fptimes* (cdr (real-value a))
599 (cdr (intofp b))))))
601 (defmethod two-arg-* ((a bigfloat) (b cl:complex))
602 (make-instance 'complex-bigfloat
603 :real (maxima::bcons
604 (maxima::fptimes* (cdr (real-value a))
605 (cdr (intofp (realpart b)))))
606 :imag (maxima::bcons
607 (maxima::fptimes* (cdr (real-value a))
608 (cdr (intofp (imagpart b)))))))
610 (defmethod two-arg-* ((a cl:number) (b bigfloat))
611 (two-arg-* b a))
613 (defmethod two-arg-* ((a complex-bigfloat) (b bigfloat))
614 (make-instance 'complex-bigfloat
615 :real (maxima::bcons
616 (maxima::fptimes* (cdr (real-value a))
617 (cdr (real-value b))))
618 :imag (maxima::bcons
619 (maxima::fptimes* (cdr (imag-value a))
620 (cdr (real-value b))))))
622 (defmethod two-arg-* ((a complex-bigfloat) (b number))
623 (if (cl:complexp b)
624 (two-arg-* a (bigfloat (cl:realpart b) (cl:imagpart b)))
625 (two-arg-* a (bigfloat b))))
627 (defmethod two-arg-* ((a bigfloat) (b complex-bigfloat))
628 (two-arg-* b a))
630 (defmethod two-arg-* ((a number) (b complex-bigfloat))
631 (two-arg-* b a))
633 (defun * (&rest args)
634 (if (null args)
636 (do ((args (cdr args) (cdr args))
637 (res (car args)
638 (two-arg-* res (car args))))
639 ((null args) res))))
641 ;;; Reciprocal of a number
642 (defmethod unary-divide ((a number))
643 (cl:/ a))
645 (defmethod unary-divide ((a bigfloat))
646 (make-instance 'bigfloat
647 :real (maxima::bcons
648 (maxima::fpquotient (maxima::fpone)
649 (cdr (real-value a))))))
651 (defmethod unary-divide ((b complex-bigfloat))
652 ;; Could just call two-arg-/, but let's optimize this a little
653 (let ((a-re (maxima::fpone))
654 (b-re (cdr (real-value b)))
655 (b-im (cdr (imag-value b))))
656 (if (maxima::fpgreaterp (maxima::fpabs b-re) (maxima::fpabs b-im))
657 (let* ((r (maxima::fpquotient b-im b-re))
658 (dn (maxima::fpplus b-re (maxima::fptimes* r b-im))))
659 (make-instance 'complex-bigfloat
660 :real (maxima::bcons (maxima::fpquotient a-re dn))
661 :imag (maxima::bcons
662 (maxima::fpquotient (maxima::fpminus r)
663 dn))))
664 (let* ((r (maxima::fpquotient b-re b-im))
665 (dn (maxima::fpplus b-im (maxima::fptimes* r b-re))))
666 (make-instance 'complex-bigfloat
667 :real (maxima::bcons (maxima::fpquotient r dn))
668 :imag (maxima::bcons
669 (maxima::fpquotient (maxima::fpminus a-re)
670 dn)))))))
671 ;;; Divide two numbers
672 (defmethod two-arg-/ ((a number) (b number))
673 (cl:/ a b))
675 (defmethod two-arg-/ ((a bigfloat) (b bigfloat))
676 (make-instance 'bigfloat
677 :real (maxima::bcons
678 (maxima::fpquotient (cdr (real-value a))
679 (cdr (real-value b))))))
681 (defmethod two-arg-/ ((a complex-bigfloat) (b complex-bigfloat))
682 (let ((a-re (cdr (real-value a)))
683 (a-im (cdr (imag-value a)))
684 (b-re (cdr (real-value b)))
685 (b-im (cdr (imag-value b))))
686 (if (maxima::fpgreaterp (maxima::fpabs b-re) (maxima::fpabs b-im))
687 (let* ((r (maxima::fpquotient b-im b-re))
688 (dn (maxima::fpplus b-re (maxima::fptimes* r b-im))))
689 (make-instance 'complex-bigfloat
690 :real (maxima::bcons
691 (maxima::fpquotient
692 (maxima::fpplus a-re
693 (maxima::fptimes* a-im r))
694 dn))
695 :imag (maxima::bcons
696 (maxima::fpquotient
697 (maxima::fpdifference a-im
698 (maxima::fptimes* a-re r))
699 dn))))
700 (let* ((r (maxima::fpquotient b-re b-im))
701 (dn (maxima::fpplus b-im (maxima::fptimes* r b-re))))
702 (make-instance 'complex-bigfloat
703 :real (maxima::bcons
704 (maxima::fpquotient
705 (maxima::fpplus (maxima::fptimes* a-re r)
706 a-im)
707 dn))
708 :imag (maxima::bcons
709 (maxima::fpquotient (maxima::fpdifference
710 (maxima::fptimes* a-im r)
711 a-re)
712 dn)))))))
713 ;; Handle contagion for two-arg-/
714 (defmethod two-arg-/ ((a bigfloat) (b cl:float))
715 (make-instance 'bigfloat
716 :real (maxima::bcons
717 (maxima::fpquotient (cdr (real-value a))
718 (cdr (intofp b))))))
720 (defmethod two-arg-/ ((a bigfloat) (b cl:rational))
721 (make-instance 'bigfloat
722 :real (maxima::bcons
723 (maxima::fpquotient (cdr (real-value a))
724 (cdr (intofp b))))))
726 (defmethod two-arg-/ ((a bigfloat) (b cl:complex))
727 (two-arg-/ (complex a) b))
729 (defmethod two-arg-/ ((a cl:float) (b bigfloat))
730 (make-instance 'bigfloat
731 :real (maxima::bcons
732 (maxima::fpquotient (cdr (intofp a))
733 (cdr (real-value b))))))
735 (defmethod two-arg-/ ((a cl:rational) (b bigfloat))
736 (make-instance 'bigfloat
737 :real (maxima::bcons
738 (maxima::fpquotient (cdr (intofp a))
739 (cdr (real-value b))))))
741 (defmethod two-arg-/ ((a cl:complex) (b bigfloat))
742 (two-arg-/ (bigfloat a) b))
745 (defmethod two-arg-/ ((a complex-bigfloat) (b bigfloat))
746 (make-instance 'complex-bigfloat
747 :real (maxima::bcons
748 (maxima::fpquotient (cdr (real-value a))
749 (cdr (real-value b))))
750 :imag (maxima::bcons
751 (maxima::fpquotient (cdr (imag-value a))
752 (cdr (real-value b))))))
754 (defmethod two-arg-/ ((a complex-bigfloat) (b number))
755 (if (cl:complexp b)
756 (two-arg-/ a (bigfloat (cl:realpart b) (cl:imagpart b)))
757 (two-arg-/ a (bigfloat b))))
759 (defmethod two-arg-/ ((a bigfloat) (b complex-bigfloat))
760 (two-arg-/ (make-instance 'complex-bigfloat :real (real-value a))
763 (defmethod two-arg-/ ((a number) (b complex-bigfloat))
764 (if (cl:complexp a)
765 (two-arg-/ (bigfloat (cl:realpart a) (cl:imagpart a))
767 (two-arg-/ (bigfloat a) b)))
770 (defun / (number &rest more-numbers)
771 (if more-numbers
772 (do ((nlist more-numbers (cdr nlist))
773 (result number))
774 ((atom nlist) result)
775 (declare (list nlist))
776 (setq result (two-arg-/ result (car nlist))))
777 (unary-divide number)))
779 ;;; Compare against zero (zerop, minusp, plusp)
780 (macrolet
781 ((frob (name)
782 (let ((cl-name (intern (symbol-name name) :cl)))
783 `(progn
784 (defmethod ,name ((x cl:float))
785 (,cl-name x))
786 (defmethod ,name ((x cl:rational))
787 (,cl-name x))))))
788 (frob plusp)
789 (frob minusp))
791 (defmethod zerop ((x number))
792 (cl:zerop x))
794 (defmethod zerop ((x bigfloat))
795 (let ((r (cdr (real-value x))))
796 (and (zerop (first r))
797 (zerop (second r)))))
799 (defmethod zerop ((a complex-bigfloat))
800 (and (equal (cdr (real-value a)) '(0 0))
801 (equal (cdr (imag-value a)) '(0 0))))
803 (defmethod plusp ((x bigfloat))
804 (cl:plusp (first (cdr (real-value x)))))
806 (defmethod minusp ((x bigfloat))
807 (cl:minusp (first (cdr (real-value x)))))
811 ;;; Equality
812 (defmethod two-arg-= ((a number) (b number))
813 (cl:= a b))
815 (defmethod two-arg-= ((a bigfloat) (b bigfloat))
816 (zerop (two-arg-- a b)))
818 (defmethod two-arg-= ((a complex-bigfloat) (b complex-bigfloat))
819 (zerop (two-arg-- a b)))
821 ;; Handle contagion for two-arg-=. This needs some work. CL says
822 ;; floats and rationals are compared by converting the float to a
823 ;; rational before converting.
824 (defmethod two-arg-= ((a bigfloat) (b number))
825 (zerop (two-arg-- a b)))
827 (defmethod two-arg-= ((a number) (b bigfloat))
828 (two-arg-= b a))
830 (defmethod two-arg-= ((a complex-bigfloat) (b number))
831 (zerop (two-arg-- a b)))
833 (defmethod two-arg-= ((a number) (b complex-bigfloat))
834 (two-arg-= b a))
836 (defun = (number &rest more-numbers)
837 "Returns T if all of its arguments are numerically equal, NIL otherwise."
838 (declare (optimize (safety 2))
839 #-gcl (dynamic-extent more-numbers))
840 (do ((nlist more-numbers (cdr nlist)))
841 ((atom nlist) t)
842 (declare (list nlist))
843 (if (not (two-arg-= (car nlist) number))
844 (return nil))))
846 (defun /= (number &rest more-numbers)
847 "Returns T if no two of its arguments are numerically equal, NIL otherwise."
848 (declare (optimize (safety 2))
849 #-gcl (dynamic-extent more-numbers))
850 (do* ((head number (car nlist))
851 (nlist more-numbers (cdr nlist)))
852 ((atom nlist) t)
853 (declare (list nlist))
854 (unless (do* ((nl nlist (cdr nl)))
855 ((atom nl) t)
856 (declare (list nl))
857 (if (two-arg-= head (car nl))
858 (return nil)))
859 (return nil))))
861 ;;; Comparison operations
862 (macrolet
863 ((frob (op)
864 (let ((method (intern (concatenate 'string
865 (string '#:two-arg-)
866 (symbol-name op))))
867 (cl-fun (find-symbol (symbol-name op) :cl)))
868 `(progn
869 (defmethod ,method ((a cl:float) (b cl:float))
870 (,cl-fun a b))
871 (defmethod ,method ((a cl:float) (b cl:rational))
872 (,cl-fun a b))
873 (defmethod ,method ((a cl:rational) (b cl:float))
874 (,cl-fun a b))
875 (defmethod ,method ((a cl:rational) (b cl:rational))
876 (,cl-fun a b))
877 (defun ,op (number &rest more-numbers)
878 "Returns T if its arguments are in strictly increasing order, NIL otherwise."
879 (declare (optimize (safety 2))
880 #-gcl (dynamic-extent more-numbers))
881 (do* ((n number (car nlist))
882 (nlist more-numbers (cdr nlist)))
883 ((atom nlist) t)
884 (declare (list nlist))
885 (if (not (,method n (car nlist))) (return nil))))))))
886 (frob <)
887 (frob >)
888 (frob <=)
889 (frob >=))
891 (defmethod two-arg-< ((x bigfloat) (y bigfloat))
892 (maxima::fplessp (cdr (real-value x)) (cdr (real-value y))))
894 (defmethod two-arg-< ((x bigfloat) (y cl:float))
895 (maxima::fplessp (cdr (real-value x)) (cdr (intofp y))))
897 (defmethod two-arg-< ((x bigfloat) (y cl:rational))
898 (maxima::fplessp (cdr (real-value x)) (cdr (intofp y))))
900 (defmethod two-arg-< ((x cl:float) (y bigfloat))
901 (maxima::fplessp (cdr (intofp x)) (cdr (real-value y))))
903 (defmethod two-arg-< ((x cl:rational) (y bigfloat))
904 (maxima::fplessp (cdr (intofp x)) (cdr (real-value y))))
906 (defmethod two-arg-> ((x bigfloat) (y bigfloat))
907 (maxima::fpgreaterp (cdr (real-value x)) (cdr (real-value y))))
909 (defmethod two-arg-> ((x bigfloat) (y cl:float))
910 (maxima::fpgreaterp (cdr (real-value x)) (cdr (intofp y))))
912 (defmethod two-arg-> ((x bigfloat) (y cl:rational))
913 (maxima::fpgreaterp (cdr (real-value x)) (cdr (intofp y))))
915 (defmethod two-arg-> ((x cl:float) (y bigfloat))
916 (maxima::fpgreaterp (cdr (intofp x)) (cdr (real-value y))))
918 (defmethod two-arg-> ((x cl:rational) (y bigfloat))
919 (maxima::fpgreaterp (cdr (intofp x)) (cdr (real-value y))))
921 (defmethod two-arg-<= ((x bigfloat) (y bigfloat))
922 (or (zerop (two-arg-- x y))
923 (maxima::fplessp (cdr (real-value x)) (cdr (real-value y)))))
925 (defmethod two-arg-<= ((x bigfloat) (y cl:float))
926 (or (zerop (two-arg-- x y))
927 (maxima::fplessp (cdr (real-value x)) (cdr (intofp y)))))
929 (defmethod two-arg-<= ((x bigfloat) (y cl:rational))
930 (or (zerop (two-arg-- x y))
931 (maxima::fplessp (cdr (real-value x)) (cdr (intofp y)))))
933 (defmethod two-arg-<= ((x cl:float) (y bigfloat))
934 (or (zerop (two-arg-- x y))
935 (maxima::fplessp (cdr (intofp x)) (cdr (real-value y)))))
937 (defmethod two-arg-<= ((x cl:rational) (y bigfloat))
938 (or (zerop (two-arg-- x y))
939 (maxima::fplessp (cdr (intofp x)) (cdr (real-value y)))))
941 (defmethod two-arg->= ((x bigfloat) (y bigfloat))
942 (or (zerop (two-arg-- x y))
943 (maxima::fpgreaterp (cdr (real-value x)) (cdr (real-value y)))))
945 (defmethod two-arg->= ((x bigfloat) (y cl:float))
946 (or (zerop (two-arg-- x y))
947 (maxima::fpgreaterp (cdr (real-value x)) (cdr (intofp y)))))
949 (defmethod two-arg->= ((x bigfloat) (y cl:rational))
950 (or (zerop (two-arg-- x y))
951 (maxima::fpgreaterp (cdr (real-value x)) (cdr (intofp y)))))
953 (defmethod two-arg->= ((x cl:float) (y bigfloat))
954 (or (zerop (two-arg-- x y))
955 (maxima::fpgreaterp (cdr (intofp x)) (cdr (real-value y)))))
957 (defmethod two-arg->= ((x cl:rational) (y bigfloat))
958 (or (zerop (two-arg-- x y))
959 (maxima::fpgreaterp (cdr (intofp x)) (cdr (real-value y)))))
961 ;; Need to define incf and decf to call our generic +/- methods.
962 (defmacro incf (place &optional (delta 1) &environment env)
963 "The first argument is some location holding a number. This number is
964 incremented by the second argument, DELTA, which defaults to 1."
965 (multiple-value-bind (dummies vals newval setter getter)
966 (get-setf-expansion place env)
967 (let ((d (gensym)))
968 `(let* (,@(mapcar #'list dummies vals)
969 (,d ,delta)
970 (,(car newval) (+ ,getter ,d)))
971 ,setter))))
973 (defmacro decf (place &optional (delta 1) &environment env)
974 "The first argument is some location holding a number. This number is
975 decremented by the second argument, DELTA, which defaults to 1."
976 (multiple-value-bind (dummies vals newval setter getter)
977 (get-setf-expansion place env)
978 (let ((d (gensym)))
979 `(let* (,@(mapcar #'list dummies vals)
980 (,d ,delta)
981 (,(car newval) (- ,getter ,d)))
982 ,setter))))
986 ;;; Special functions for real-valued arguments
987 (macrolet
988 ((frob (name)
989 (let ((cl-name (intern (symbol-name name) :cl)))
990 `(progn
991 (defmethod ,name ((x number))
992 (,cl-name x))))))
993 (frob sqrt)
994 (frob abs)
995 (frob exp)
996 (frob sin)
997 (frob cos)
998 (frob tan)
999 (frob asin)
1000 (frob acos)
1001 (frob sinh)
1002 (frob cosh)
1003 (frob tanh)
1004 (frob asinh)
1005 (frob acosh)
1006 (frob atanh)
1009 (defmethod abs ((x bigfloat))
1010 (make-instance 'bigfloat
1011 :real (maxima::bcons (maxima::fpabs (cdr (real-value x))))))
1013 (defmethod abs ((z complex-bigfloat))
1014 (let ((x (make-instance 'bigfloat :real (real-value z)))
1015 (y (make-instance 'bigfloat :real (imag-value z))))
1016 ;; Bigfloats don't overflow, so we don't need anything special.
1017 (sqrt (+ (* x x) (* y y)))))
1019 (defmethod exp ((x bigfloat))
1020 (make-instance 'bigfloat
1021 :real (maxima::bcons (maxima::fpexp (cdr (real-value x))))))
1023 (defmethod sin ((x bigfloat))
1024 (make-instance 'bigfloat
1025 :real (maxima::bcons (maxima::fpsin (cdr (real-value x)) t))))
1027 (defmethod cos ((x bigfloat))
1028 (make-instance 'bigfloat
1029 :real (maxima::bcons (maxima::fpsin (cdr (real-value x)) nil))))
1031 (defmethod tan ((x bigfloat))
1032 (let ((r (cdr (real-value x))))
1033 (make-instance 'bigfloat
1034 :real (maxima::bcons
1035 (maxima::fpquotient (maxima::fpsin r t)
1036 (maxima::fpsin r nil))))))
1038 (defmethod asin ((x bigfloat))
1039 (bigfloat (maxima::fpasin (real-value x))))
1041 (defmethod acos ((x bigfloat))
1042 (bigfloat (maxima::fpacos (real-value x))))
1045 (defmethod sqrt ((x bigfloat))
1046 (if (minusp x)
1047 (make-instance 'complex-bigfloat
1048 :real (intofp 0)
1049 :imag (maxima::bcons
1050 (maxima::fproot (maxima::bcons (maxima::fpabs (cdr (real-value x)))) 2)))
1051 (make-instance 'bigfloat
1052 :real (maxima::bcons
1053 (maxima::fproot (real-value x) 2)))))
1055 (defmethod sqrt ((z complex-bigfloat))
1056 (multiple-value-bind (rx ry)
1057 (maxima::complex-sqrt (real-value z)
1058 (imag-value z))
1059 (make-instance 'complex-bigfloat
1060 :real (maxima::bcons rx)
1061 :imag (maxima::bcons ry))))
1063 (defmethod one-arg-log ((a number))
1064 (cl:log a))
1066 (defmethod one-arg-log ((a bigfloat))
1067 (if (minusp a)
1068 (make-instance 'complex-bigfloat
1069 :real (maxima::bcons
1070 (maxima::fplog (maxima::fpabs (cdr (real-value a)))))
1071 :imag (maxima::bcons (maxima::fppi)))
1072 (make-instance 'bigfloat
1073 :real (maxima::bcons (maxima::fplog (cdr (real-value a)))))))
1075 (defmethod one-arg-log ((a complex-bigfloat))
1076 (let* ((res (maxima::big-float-log (real-value a)
1077 (imag-value a))))
1078 (bigfloat res)))
1080 (defmethod two-arg-log ((a number) (b number))
1081 (cl:log a b))
1083 (defmethod two-arg-log ((a numeric) (b numeric))
1084 (two-arg-/ (one-arg-log a) (one-arg-log b)))
1086 (defmethod two-arg-log ((a numeric) (b cl:number))
1087 (two-arg-/ (one-arg-log a) (one-arg-log (bigfloat b))))
1089 (defmethod two-arg-log ((a cl:number) (b numeric))
1090 (two-arg-/ (one-arg-log (bigfloat a)) (one-arg-log b)))
1092 (defun log (a &optional b)
1093 (if b
1094 (two-arg-log a b)
1095 (one-arg-log a)))
1097 (defmethod sinh ((x bigfloat))
1098 (let ((r (real-value x)))
1099 (make-instance 'bigfloat :real (maxima::fpsinh r))))
1101 (defmethod cosh ((x bigfloat))
1102 (let ((r (real-value x)))
1103 (make-instance 'bigfloat :real (maxima::$bfloat `((maxima::%cosh) ,r)))))
1105 (defmethod tanh ((x bigfloat))
1106 (let ((r (real-value x)))
1107 (make-instance 'bigfloat :real (maxima::fptanh r))))
1109 (defmethod asinh ((x bigfloat))
1110 (let ((r (real-value x)))
1111 (make-instance 'bigfloat :real (maxima::fpasinh r))))
1113 (defmethod atanh ((x bigfloat))
1114 (let ((r (maxima::big-float-atanh (real-value x))))
1115 (if (maxima::bigfloatp r)
1116 (make-instance 'bigfloat :real r)
1117 (make-instance 'complex-bigfloat
1118 :real (maxima::$realpart r)
1119 :imag (maxima::$imagpart r)))))
1121 (defmethod acosh ((x bigfloat))
1122 (let* ((r (real-value x))
1123 (value (maxima::mevalp `((maxima::%acosh maxima::simp)
1124 ,r))))
1125 (if (maxima::bigfloatp value)
1126 (make-instance 'bigfloat :real value)
1127 (make-instance 'complex-bigfloat
1128 :real (maxima::$realpart value)
1129 :imag (maxima::$imagpart value)))))
1131 ;;; Complex arguments
1133 ;;; Special functions for complex args
1134 (macrolet
1135 ((frob (name &optional big-float-op-p)
1136 (if big-float-op-p
1137 (let ((big-op (intern (concatenate 'string
1138 (string '#:big-float-)
1139 (string name))
1140 '#:maxima)))
1141 `(defmethod ,name ((a complex-bigfloat))
1142 (let ((res (,big-op (real-value a)
1143 (imag-value a))))
1144 (to res))))
1145 (let ((max-op (intern (concatenate 'string "$" (string name)) '#:maxima)))
1146 `(defmethod ,name ((a complex-bigfloat))
1147 ;; We should do something better than calling mevalp
1148 (let* ((arg (maxima::add (real-value a)
1149 (maxima::mul 'maxima::$%i (imag-value a))))
1150 (result (maxima::mevalp `((,',max-op maxima::simp) ,arg))))
1151 (to result)))))))
1152 (frob exp)
1153 (frob sin)
1154 (frob cos)
1155 (frob tan)
1156 (frob asin t)
1157 (frob acos t)
1158 (frob sinh)
1159 (frob cosh)
1160 (frob tanh t)
1161 (frob asinh t)
1162 (frob acosh)
1163 (frob atanh t))
1165 (defmethod one-arg-atan ((a number))
1166 (cl:atan a))
1168 (defmethod one-arg-atan ((a bigfloat))
1169 (make-instance 'bigfloat
1170 :real (maxima::bcons (maxima::fpatan (cdr (real-value a))))))
1172 (defmethod one-arg-atan ((a complex-bigfloat))
1173 ;; We should do something better than calling mevalp
1174 (let* ((arg (maxima::add (real-value a)
1175 (maxima::mul 'maxima::$%i (imag-value a))))
1176 (result (maxima::mevalp `((maxima::%atan maxima::simp) ,arg))))
1177 (make-instance 'complex-bigfloat
1178 :real (maxima::$realpart result)
1179 :imag (maxima::$imagpart result))))
1181 ;; Really want type real, but gcl doesn't like that. Define methods for rational and float
1182 #-gcl
1183 (defmethod two-arg-atan ((a real) (b real))
1184 (cl:atan a b))
1186 #+gcl
1187 (progn
1188 (defmethod two-arg-atan ((a cl:float) (b cl:float))
1189 (cl:atan a b))
1190 (defmethod two-arg-atan ((a cl:rational) (b cl:rational))
1191 (cl:atan a b))
1192 (defmethod two-arg-atan ((a cl:float) (b cl:rational))
1193 (cl:atan a b))
1194 (defmethod two-arg-atan ((a cl:rational) (b cl:float))
1195 (cl:atan a b))
1198 (defmethod two-arg-atan ((a bigfloat) (b bigfloat))
1199 (make-instance 'bigfloat
1200 :real (maxima::bcons
1201 (maxima::fpatan2 (cdr (real-value a))
1202 (cdr (real-value b))))))
1204 (defmethod two-arg-atan ((a bigfloat) (b cl:float))
1205 (make-instance 'bigfloat
1206 :real (maxima::bcons (maxima::fpatan2 (cdr (real-value a))
1207 (cdr (intofp b))))))
1209 (defmethod two-arg-atan ((a bigfloat) (b cl:rational))
1210 (make-instance 'bigfloat
1211 :real (maxima::bcons (maxima::fpatan2 (cdr (real-value a))
1212 (cdr (intofp b))))))
1214 (defmethod two-arg-atan ((a cl:float) (b bigfloat))
1215 (make-instance 'bigfloat
1216 :real (maxima::bcons (maxima::fpatan2 (cdr (intofp a))
1217 (cdr (real-value b))))))
1219 (defmethod two-arg-atan ((a cl:rational) (b bigfloat))
1220 (make-instance 'bigfloat
1221 :real (maxima::bcons (maxima::fpatan2 (cdr (intofp a))
1222 (cdr (real-value b))))))
1224 (defun atan (a &optional b)
1225 (if b
1226 (two-arg-atan a b)
1227 (one-arg-atan a)))
1229 (defmethod scale-float ((a cl:float) (n integer))
1230 (cl:scale-float a n))
1232 (defmethod scale-float ((a bigfloat) (n integer))
1233 (if (cl:zerop (second (real-value a)))
1234 (make-instance 'bigfloat :real (maxima::bcons (list 0 0)))
1235 (destructuring-bind (marker mantissa exp)
1236 (real-value a)
1237 (declare (ignore marker))
1238 (make-instance 'bigfloat :real (maxima::bcons (list mantissa (+ exp n)))))))
1240 (macrolet
1241 ((frob (name)
1242 (let ((cl-name (intern (string name) '#:cl)))
1243 `(defmethod ,name ((a number))
1244 (,cl-name a)))))
1245 (frob realpart)
1246 (frob imagpart)
1247 (frob conjugate)
1248 (frob phase))
1250 (macrolet
1251 ((frob (name)
1252 (let ((cl-name (intern (string name) '#:cl)))
1253 `(defmethod ,name ((a number) &optional (divisor 1))
1254 (,cl-name a divisor)))))
1255 (frob floor)
1256 (frob ffloor)
1257 (frob ceiling)
1258 (frob fceiling)
1259 (frob truncate)
1260 (frob ftruncate)
1261 (frob round)
1262 (frob fround))
1265 (defmethod realpart ((a bigfloat))
1266 (make-instance 'bigfloat :real (real-value a)))
1268 (defmethod realpart ((a complex-bigfloat))
1269 (make-instance 'bigfloat :real (real-value a)))
1271 (defmethod imagpart ((a bigfloat))
1272 (make-instance 'bigfloat :real (intofp 0)))
1274 (defmethod imagpart ((a complex-bigfloat))
1275 (make-instance 'bigfloat :real (imag-value a)))
1277 (defmethod conjugate ((a bigfloat))
1278 (make-instance 'bigfloat :real (real-value a)))
1280 (defmethod conjugate ((a complex-bigfloat))
1281 (make-instance 'complex-bigfloat
1282 :real (real-value a)
1283 :imag (maxima::bcons (maxima::fpminus (cdr (imag-value a))))))
1285 (defmethod cis ((a cl:float))
1286 (cl:cis a))
1288 (defmethod cis ((a cl:rational))
1289 (cl:cis a))
1291 (defmethod cis ((a bigfloat))
1292 (make-instance 'complex-bigfloat
1293 :real (maxima::bcons (maxima::fpsin (cdr (real-value a)) nil))
1294 :imag (maxima::bcons (maxima::fpsin (cdr (real-value a)) t))))
1296 (defmethod phase ((a bigfloat))
1297 (let ((r (cdr (real-value a))))
1298 (if (cl:>= (car r) 0)
1299 (make-instance 'bigfloat :real (maxima::bcons (list 0 0)))
1300 (make-instance 'bigfloat :real (maxima::bcons (maxima::fppi))))))
1302 (defmethod phase ((a complex-bigfloat))
1303 (make-instance 'bigfloat
1304 :real (maxima::bcons (maxima::fpatan2 (cdr (imag-value a))
1305 (cdr (real-value a))))))
1307 (defun max (number &rest more-numbers)
1308 "Returns the greatest of its arguments."
1309 (declare (optimize (safety 2)) (type (or real bigfloat) number)
1310 #-gcl (dynamic-extent more-numbers))
1311 (dolist (real more-numbers)
1312 (when (> real number)
1313 (setq number real)))
1314 number)
1316 (defun min (number &rest more-numbers)
1317 "Returns the least of its arguments."
1318 (declare (optimize (safety 2)) (type (or real bigfloat) number)
1319 #-gcl (dynamic-extent more-numbers))
1320 (do ((nlist more-numbers (cdr nlist))
1321 (result (the (or real bigfloat) number)))
1322 ((null nlist) (return result))
1323 (declare (list nlist))
1324 (if (< (car nlist) result)
1325 (setq result (car nlist)))))
1327 ;; We really want a real type, but gcl doesn't like it, so use number
1328 ;; instead.
1329 #-gcl
1330 (defmethod one-arg-complex ((a real))
1331 (cl:complex a))
1333 #+gcl
1334 (progn
1335 (defmethod one-arg-complex ((a cl:float))
1336 (cl:complex a))
1337 (defmethod one-arg-complex ((a cl:rational))
1338 (cl:complex a))
1341 (defmethod one-arg-complex ((a bigfloat))
1342 (make-instance 'complex-bigfloat
1343 :real (real-value a)
1344 :imag (intofp 0)))
1346 #-gcl
1347 (defmethod two-arg-complex ((a real) (b real))
1348 (cl:complex a b))
1350 #+gcl
1351 (progn
1352 (defmethod two-arg-complex ((a cl:float) (b cl:float))
1353 (cl:complex a b))
1354 (defmethod two-arg-complex ((a cl:rational) (b cl:rational))
1355 (cl:complex a b))
1356 (defmethod two-arg-complex ((a cl:float) (b cl:rational))
1357 (cl:complex a b))
1358 (defmethod two-arg-complex ((a cl:rational) (b cl:float))
1359 (cl:complex a b))
1362 (defmethod two-arg-complex ((a bigfloat) (b bigfloat))
1363 (make-instance 'complex-bigfloat
1364 :real (real-value a)
1365 :imag (real-value b)))
1367 (defmethod two-arg-complex ((a cl:float) (b bigfloat))
1368 (make-instance 'complex-bigfloat
1369 :real (intofp a)
1370 :imag (real-value b)))
1372 (defmethod two-arg-complex ((a cl:rational) (b bigfloat))
1373 (make-instance 'complex-bigfloat
1374 :real (intofp a)
1375 :imag (real-value b)))
1377 (defmethod two-arg-complex ((a bigfloat) (b cl:float))
1378 (make-instance 'complex-bigfloat
1379 :real (real-value a)
1380 :imag (intofp b)))
1382 (defmethod two-arg-complex ((a bigfloat) (b cl:rational))
1383 (make-instance 'complex-bigfloat
1384 :real (real-value a)
1385 :imag (intofp b)))
1387 (defun complex (a &optional b)
1388 (if b
1389 (two-arg-complex a b)
1390 (one-arg-complex a)))
1392 (defmethod unary-floor ((a bigfloat))
1393 ;; fpentier truncates to zero, so adjust for negative numbers
1394 (let ((trunc (maxima::fpentier (real-value a))))
1395 (cond ((minusp a)
1396 ;; If the truncated value is the same as the original,
1397 ;; there's nothing to do because A was an integer.
1398 ;; Otherwise, we need to subtract 1 to make it the floor.
1399 (if (= trunc a)
1400 trunc
1401 (1- trunc)))
1403 trunc))))
1405 (defmethod unary-ffloor ((a bigfloat))
1406 ;; We can probably do better than converting to an integer and
1407 ;; converting back to a float.
1408 (make-instance 'bigfloat :real (intofp (unary-floor a))))
1410 (defmethod floor ((a bigfloat) &optional (divisor 1))
1411 (if (= divisor 1)
1412 (let ((int (unary-floor a)))
1413 (values int (- a int)))
1414 (let ((q (unary-floor (/ a divisor))))
1415 (values q (- a (* q divisor))))))
1417 (defmethod ffloor ((a bigfloat) &optional (divisor 1))
1418 (if (= divisor 1)
1419 (let ((int (unary-ffloor a)))
1420 (values int (- a int)))
1421 (let ((q (unary-ffloor (/ a divisor))))
1422 (values q (- a (* q divisor))))))
1424 (defmethod unary-truncate ((a bigfloat))
1425 (maxima::fpentier (real-value a)))
1427 (defmethod unary-ftruncate ((a bigfloat))
1428 ;; We can probably do better than converting to an integer and
1429 ;; converting back to a float.
1430 (make-instance 'bigfloat :real (intofp (unary-truncate a))))
1432 (defmethod truncate ((a bigfloat) &optional (divisor 1))
1433 (if (eql divisor 1)
1434 (let ((int (unary-truncate a)))
1435 (values int (- a int)))
1436 (let ((q (unary-truncate (/ a divisor))))
1437 (values q (- a (* q divisor))))))
1439 (defmethod ftruncate ((a bigfloat) &optional (divisor 1))
1440 (if (eql divisor 1)
1441 (let ((int (unary-ftruncate a)))
1442 (values int (- a int)))
1443 (let ((q (unary-ftruncate (/ a divisor))))
1444 (values q (- a (* q divisor))))))
1446 (defmethod unary-ceiling ((a bigfloat))
1447 ;; fpentier truncates to zero, so adjust for positive numbers.
1448 (if (minusp a)
1449 (maxima::fpentier (real-value a))
1450 (maxima::fpentier (real-value (+ a 1)))))
1452 (defmethod unary-fceiling ((a bigfloat))
1453 ;; We can probably do better than converting to an integer and
1454 ;; converting back to a float.
1455 (make-instance 'bigfloat :real (intofp (unary-ceiling a))))
1457 (defmethod ceiling ((a bigfloat) &optional (divisor 1))
1458 (if (eql divisor 1)
1459 (let ((int (unary-ceiling a)))
1460 (values int (- a int)))
1461 (let ((q (unary-ceiling (/ a divisor))))
1462 (values q (- a (* q divisor))))))
1464 (defmethod fceiling ((a bigfloat) &optional (divisor 1))
1465 (if (eql divisor 1)
1466 (let ((int (unary-fceiling a)))
1467 (values int (- a int)))
1468 (let ((q (unary-fceiling (/ a divisor))))
1469 (values q (- a (* q divisor))))))
1471 ;; Stolen from CMUCL.
1472 (defmethod round ((a bigfloat) &optional (divisor 1))
1473 (multiple-value-bind (tru rem)
1474 (truncate a divisor)
1475 (if (zerop rem)
1476 (values tru rem)
1477 (let ((thresh (/ (abs divisor) 2)))
1478 (cond ((or (> rem thresh)
1479 (and (= rem thresh) (oddp tru)))
1480 (if (minusp divisor)
1481 (values (- tru 1) (+ rem divisor))
1482 (values (+ tru 1) (- rem divisor))))
1483 ((let ((-thresh (- thresh)))
1484 (or (< rem -thresh)
1485 (and (= rem -thresh) (oddp tru))))
1486 (if (minusp divisor)
1487 (values (+ tru 1) (- rem divisor))
1488 (values (- tru 1) (+ rem divisor))))
1489 (t (values tru rem)))))))
1491 (defmethod fround ((number bigfloat) &optional (divisor 1))
1492 "Same as ROUND, but returns first value as a float."
1493 (multiple-value-bind (res rem)
1494 (round number divisor)
1495 (values (bigfloat res) rem)))
1497 (defmethod expt ((a number) (b number))
1498 (cl:expt a b))
1500 ;; This needs more work
1501 (defmethod expt ((a numeric) (b numeric))
1502 (if (zerop b)
1503 ;; CLHS says if the power is 0, the answer is 1 of the appropriate type.
1504 (if (or (typep a 'complex-bigfloat)
1505 (typep b 'complex-bigfloat))
1506 (complex (bigfloat 1))
1507 (bigfloat 1))
1508 (cond ((and (zerop a) (plusp (realpart b)))
1509 (* a b))
1510 ((and (realp b) (= b (truncate b)))
1511 ;; Use the numeric^number method because it can be much
1512 ;; more accurate when b is an integer.
1513 (expt a (truncate b)))
1515 (with-extra-precision ((expt-extra-bits a b)
1516 (a b))
1517 (exp (* b (log a))))))))
1519 (defmethod expt ((a cl:number) (b numeric))
1520 (if (zerop b)
1521 ;; CLHS says if the power is 0, the answer is 1 of the appropriate type.
1522 (if (or (typep a 'cl:complex)
1523 (typep b 'complex-bigfloat))
1524 (complex (bigfloat 1))
1525 (bigfloat 1))
1526 (cond ((and (zerop a) (plusp (realpart b)))
1527 (* a b))
1528 ((and (realp b) (= b (truncate b)))
1529 (with-extra-precision ((expt-extra-bits a b)
1530 (a b))
1531 (intofp (expt a (truncate b)))))
1533 (with-extra-precision ((expt-extra-bits a b)
1534 (a b))
1535 (exp (* b (log (bigfloat a)))))))))
1537 (defmethod expt ((a numeric) (b cl:number))
1538 (if (zerop b)
1539 ;; CLHS says if the power is 0, the answer is 1 of the appropriate type.
1540 (if (or (typep a 'complex-bigfloat)
1541 (typep b 'cl:complex))
1542 (complex (bigfloat 1))
1543 (bigfloat 1))
1544 (if (and (zerop a) (plusp (realpart b)))
1545 (* a b)
1546 ;; Handle a few special cases using multiplication.
1547 (cond ((= b 1)
1549 ((= b -1)
1550 (/ a))
1551 ((= b 2)
1552 (* a a))
1553 ((= b -2)
1554 (/ (* a a)))
1555 ((= b 3) (* a a a))
1556 ((= b -3) (/ (* a a a)))
1557 ((= b 4)
1558 (let ((a2 (* a a)))
1559 (* a2 a2)))
1560 ((= b -4)
1561 (let ((a2 (* a a)))
1562 (/ (* a2 a2))))
1564 (with-extra-precision ((expt-extra-bits a b)
1565 (a b))
1566 (exp (* (bigfloat b) (log a)))))))))
1568 ;; Handle a^b a little more carefully because the result is known to
1569 ;; be real when a is real and b is an integer.
1570 (defmethod expt ((a bigfloat) (b integer))
1571 (cond ((zerop b)
1572 (bigfloat 1))
1573 ((and (zerop a) (plusp b))
1574 ;; 0^b, for positive b
1575 (* a b))
1576 ;; Handle a few special cases using multiplication.
1577 ((eql b 1) a)
1578 ((eql b -1) (/ a))
1579 ((eql b 2) (* a a))
1580 ((eql b -2) (/ (* a a)))
1581 ((eql b 3) (* a a a))
1582 ((eql b -3) (/ (* a a a)))
1583 ((eql b 4)
1584 (let ((a2 (* a a)))
1585 (* a2 a2)))
1586 ((eql b -4)
1587 (let ((a2 (* a a)))
1588 (/ (* a2 a2))))
1589 ((minusp a)
1590 ;; a^b = exp(b*log(|a|) + %i*%pi*b)
1591 ;; = exp(b*log(|a|))*exp(%i*%pi*b)
1592 ;; = (-1)^b*exp(b*log(|a|))
1593 (with-extra-precision ((expt-extra-bits a b)
1594 (a b))
1595 (* (exp (* b (log (abs a))))
1596 (if (oddp b) -1 1))))
1598 (with-extra-precision ((expt-extra-bits a b)
1599 (a b))
1600 (exp (* b (log a)))))))
1602 ;;; TO - External
1604 ;;; TO takes a maxima number and converts it. Floats remain
1605 ;;; floats, maxima cl:rationals are converted to CL cl:rationals. Maxima
1606 ;;; bigfloats are convert to BIGFLOATS. Maxima complex numbers are
1607 ;;; converted to CL complex numbers or to COMPLEX-BIGFLOAT's.
1608 (defun to (maxima-num &optional imag)
1609 (let ((result (ignore-errors (%to maxima-num imag))))
1610 (or result
1611 (maxima::merror (intl:gettext "BIGFLOAT: unable to convert ~M to a CL or BIGFLOAT number.")
1612 (if imag
1613 (maxima::add maxima-num (maxima::mul imag 'maxima::$%i))
1614 maxima-num)))))
1616 ;;; MAYBE-TO - External
1618 ;;; Like TO, but if the maxima number can't be converted to a CL
1619 ;;; number or BIGFLOAT, just return the maxima number.
1620 (defun maybe-to (maxima-num &optional imag)
1621 (let ((result (ignore-errors (%to maxima-num imag))))
1622 (or result
1623 (if imag
1624 (maxima::add maxima-num imag)
1625 maxima-num))))
1627 (defun %to (maxima-num &optional imag)
1628 (cond (imag
1629 ;; Clisp has a "feature" that (complex rat float) does not
1630 ;; make the both components of the complex number a float.
1631 ;; Sometimes this is nice, but other times it's annoying
1632 ;; because it is non-ANSI behavior. For our code, we really
1633 ;; want both components to be a float.
1634 #-clisp
1635 (complex (to maxima-num) (to imag))
1636 #+clisp
1637 (let ((re (to maxima-num))
1638 (im (to imag)))
1639 (cond ((and (rationalp re) (floatp im))
1640 (setf re (float re im)))
1641 ((and (rational im) (floatp re))
1642 (setf im (float im re))))
1643 (complex re im)))
1645 (cond ((cl:numberp maxima-num)
1646 maxima-num)
1647 ((eq maxima-num 'maxima::$%i)
1648 ;; Convert %i to an exact complex cl:rational.
1649 #c(0 1))
1650 ((consp maxima-num)
1651 ;; Some kind of maxima number
1652 (cond ((maxima::ratnump maxima-num)
1653 ;; Maxima cl:rational ((mrat ...) num den)
1654 (/ (second maxima-num) (third maxima-num)))
1655 ((maxima::$bfloatp maxima-num)
1656 (bigfloat maxima-num))
1657 ((maxima::complex-number-p maxima-num #'(lambda (x)
1658 (or (cl:realp x)
1659 (maxima::$bfloatp x)
1660 (and (consp x)
1661 (eq (caar x) 'maxima::rat)))))
1662 ;; We have some kind of complex number whose
1663 ;; parts are a cl:real, a bfloat, or a Maxima
1664 ;; cl:rational.
1665 (let ((re (maxima::$realpart maxima-num))
1666 (im (maxima::$imagpart maxima-num)))
1667 (to re im)))))
1668 ((or (typep maxima-num 'bigfloat)
1669 (typep maxima-num 'complex-bigfloat))
1670 maxima-num)
1672 (error "BIGFLOAT: unable to convert to a CL or BIGFLOAT number."))))))
1674 ;;; EPSILON - External
1676 ;;; Return the float epsilon value for the given float type.
1677 (defmethod epsilon ((x cl:float))
1678 (etypecase x
1679 (short-float short-float-epsilon)
1680 (single-float single-float-epsilon)
1681 (double-float double-float-epsilon)
1682 (long-float long-float-epsilon)))
1684 (defmethod epsilon ((x cl:complex))
1685 (epsilon (cl:realpart x)))
1687 (defmethod epsilon ((x bigfloat))
1688 ;; epsilon is just above 2^(-fpprec).
1689 (make-instance 'bigfloat
1690 :real (maxima::bcons (list (1+ (ash 1 (1- maxima::fpprec)))
1691 (- (1- maxima::fpprec))))))
1693 (defmethod epsilon ((x complex-bigfloat))
1694 (epsilon (realpart x)))
1698 ;; Compiler macros to convert + to multiple calls to two-arg-+. Same
1699 ;; for -, *, and /.
1700 (define-compiler-macro + (&whole form &rest args)
1701 (declare (ignore form))
1702 (if (null args)
1704 (do ((args (cdr args) (cdr args))
1705 (res (car args)
1706 `(two-arg-+ ,res ,(car args))))
1707 ((null args) res))))
1709 (define-compiler-macro - (&whole form number &rest more-numbers)
1710 (declare (ignore form))
1711 (if more-numbers
1712 (do ((nlist more-numbers (cdr nlist))
1713 (result number))
1714 ((atom nlist) result)
1715 (declare (list nlist))
1716 (setq result `(two-arg-- ,result ,(car nlist))))
1717 `(unary-minus ,number)))
1719 (define-compiler-macro * (&whole form &rest args)
1720 (declare (ignore form))
1721 (if (null args)
1723 (do ((args (cdr args) (cdr args))
1724 (res (car args)
1725 `(two-arg-* ,res ,(car args))))
1726 ((null args) res))))
1728 (define-compiler-macro / (number &rest more-numbers)
1729 (if more-numbers
1730 (do ((nlist more-numbers (cdr nlist))
1731 (result number))
1732 ((atom nlist) result)
1733 (declare (list nlist))
1734 (setq result `(two-arg-/ ,result ,(car nlist))))
1735 `(unary-divide ,number)))
1737 (define-compiler-macro /= (&whole form number &rest more-numbers)
1738 ;; Convert (/= x y) to (not (two-arg-= x y)). Should we try to
1739 ;; handle a few more cases?
1740 (if (cdr more-numbers)
1741 form
1742 `(not (two-arg-= ,number ,(car more-numbers)))))
1744 ;; Compiler macros to convert <, >, <=, and >= into multiple calls of
1745 ;; the corresponding two-arg-<foo> function.
1746 (macrolet
1747 ((frob (op)
1748 (let ((method (intern (concatenate 'string
1749 (string '#:two-arg-)
1750 (symbol-name op)))))
1751 `(define-compiler-macro ,op (number &rest more-numbers)
1752 (do* ((n number (car nlist))
1753 (nlist more-numbers (cdr nlist))
1754 (res nil))
1755 ((atom nlist)
1756 `(and ,@(nreverse res)))
1757 (push `(,',method ,n ,(car nlist)) res))))))
1758 (frob <)
1759 (frob >)
1760 (frob <=)
1761 (frob >=))
1763 (defmethod integer-decode-float ((x cl:float))
1764 (cl:integer-decode-float x))
1766 (defmethod integer-decode-float ((x bigfloat))
1767 (let ((r (real-value x)))
1768 (values (abs (second r))
1769 (- (third r) (third (first r)))
1770 (signum (second r)))))
1772 (defmethod decode-float ((x cl:float))
1773 (cl:decode-float x))
1775 (defmethod decode-float ((x bigfloat))
1776 (let ((r (real-value x)))
1777 (values (make-instance 'bigfloat
1778 :real (maxima::bcons (list (abs (second r)) 0)))
1779 (third r)
1780 (bigfloat (signum (second r))))))
1782 ;; GCL doesn't have a REAL class!
1783 #+gcl
1784 (progn
1785 (defmethod float ((x cl:float) (y cl:float))
1786 (cl:float x y))
1788 (defmethod float ((x cl:rational) (y cl:float))
1789 (cl:float x y))
1791 (defmethod float ((x cl:float) (y bigfloat))
1792 (bigfloat x))
1794 (defmethod float ((x cl:rational) (y bigfloat))
1795 (bigfloat x))
1798 #-gcl
1799 (progn
1800 (defmethod float ((x real) (y cl:float))
1801 (cl:float x y))
1803 (defmethod float ((x real) (y bigfloat))
1804 (bigfloat x))
1807 ;; Like Maxima's fp2flo, but for single-float numbers.
1808 (defun fp2single (l)
1809 (let ((precision (caddar l))
1810 (mantissa (cadr l))
1811 (exponent (caddr l))
1812 (fpprec (float-digits 1f0))
1813 (maxima::*m 0))
1814 ;; Round the mantissa to the number of bits of precision of the
1815 ;; machine, and then convert it to a floating point fraction. We
1816 ;; have 0.5 <= mantissa < 1
1817 (setq mantissa (maxima::quotient (maxima::fpround mantissa)
1818 (expt 2f0 fpprec)))
1819 ;; Multiply the mantissa by the exponent portion. I'm not sure
1820 ;; why the exponent computation is so complicated.
1822 ;; GCL doesn't signal overflow from scale-float if the number
1823 ;; would overflow. We have to do it this way. 0.5 <= mantissa <
1824 ;; 1. The largest double-float is .999999 * 2^128. So if the
1825 ;; exponent is 128 or higher, we have an overflow.
1826 (let ((e (+ exponent (- precision) maxima::*m fpprec)))
1827 (if (>= (abs e) 129)
1828 (maxima::merror (intl:gettext "FP2SINGLE: floating point overflow converting ~:M to float.") l)
1829 (cl:scale-float mantissa e)))))
1832 (defmethod float ((x bigfloat) (y cl:float))
1833 (if (typep y 'maxima::flonum)
1834 (maxima::fp2flo (real-value x))
1835 (fp2single (real-value x))))
1837 (defmethod random ((x cl:float) &optional (state cl:*random-state*))
1838 (cl:random x state))
1839 (defmethod random ((x integer) &optional (state cl:*random-state*))
1840 (cl:random x state))
1842 (defmethod random ((x bigfloat) &optional (state cl:*random-state*))
1843 ;; Generate an integer with fpprec bits, and convert to a bigfloat
1844 ;; by making the exponent 0. Then multiply by the arg to get the
1845 ;; correct range.
1846 (if (plusp x)
1847 (let ((int (cl:random (ash 1 maxima::fpprec) state)))
1848 (* x (bigfloat (maxima::bcons (list int 0)))))
1849 (error "Argument is not a positive bigfloat: ~A~%" x)))
1851 (defmethod signum ((x number))
1852 (cl:signum x))
1854 (defmethod signum ((x bigfloat))
1855 (cond ((minusp x)
1856 (bigfloat -1))
1857 ((plusp x)
1858 (bigfloat 1))
1860 x)))
1862 (defmethod signum ((x complex-bigfloat))
1863 (/ x (abs x)))
1865 (defmethod float-sign ((x cl:float))
1866 (cl:float-sign x))
1868 (defmethod float-sign ((x bigfloat))
1869 (if (minusp x)
1870 (bigfloat -1)
1871 (bigfloat 1)))
1873 (defmethod float-digits ((x cl:float))
1874 (cl:float-digits x))
1876 (defmethod float-digits ((x bigfloat))
1877 ;; Should we just return fpprec or should we get the actual number
1878 ;; of bits in the bigfloat number? We choose the latter in case the
1879 ;; number and fpprec don't match.
1880 (let ((r (slot-value x 'real)))
1881 (third (first r))))
1883 #-gcl
1884 (defmethod rational ((x real))
1885 (cl:rational x))
1887 #+gcl
1888 (progn
1889 (defmethod rational ((x cl:float))
1890 (cl:rational x))
1891 (defmethod rational ((x cl:rational))
1892 (cl:rational x))
1895 (defmethod rational ((x bigfloat))
1896 (destructuring-bind ((marker simp prec) mantissa exp)
1897 (real-value x)
1898 (declare (ignore marker simp))
1899 (* mantissa (expt 2 (- exp prec)))))
1901 #-gcl
1902 (defmethod rationalize ((x real))
1903 (cl:rationalize x))
1905 #+gcl
1906 (progn
1907 (defmethod rationalize ((x cl:float))
1908 (cl:rationalize x))
1909 (defmethod rationalize ((x cl:rational))
1910 (cl:rationalize x))
1914 ;;; This routine taken from CMUCL, which, in turn is a routine from
1915 ;;; CLISP, which is GPL.
1917 ;;; I (rtoy) have modified it from CMUCL so that it only handles bigfloats.
1919 ;;; RATIONALIZE -- Public
1921 ;;; The algorithm here is the method described in CLISP. Bruno Haible has
1922 ;;; graciously given permission to use this algorithm. He says, "You can use
1923 ;;; it, if you present the following explanation of the algorithm."
1925 ;;; Algorithm (recursively presented):
1926 ;;; If x is a rational number, return x.
1927 ;;; If x = 0.0, return 0.
1928 ;;; If x < 0.0, return (- (rationalize (- x))).
1929 ;;; If x > 0.0:
1930 ;;; Call (integer-decode-float x). It returns a m,e,s=1 (mantissa,
1931 ;;; exponent, sign).
1932 ;;; If m = 0 or e >= 0: return x = m*2^e.
1933 ;;; Search a rational number between a = (m-1/2)*2^e and b = (m+1/2)*2^e
1934 ;;; with smallest possible numerator and denominator.
1935 ;;; Note 1: If m is a power of 2, we ought to take a = (m-1/4)*2^e.
1936 ;;; But in this case the result will be x itself anyway, regardless of
1937 ;;; the choice of a. Therefore we can simply ignore this case.
1938 ;;; Note 2: At first, we need to consider the closed interval [a,b].
1939 ;;; but since a and b have the denominator 2^(|e|+1) whereas x itself
1940 ;;; has a denominator <= 2^|e|, we can restrict the search to the open
1941 ;;; interval (a,b).
1942 ;;; So, for given a and b (0 < a < b) we are searching a rational number
1943 ;;; y with a <= y <= b.
1944 ;;; Recursive algorithm fraction_between(a,b):
1945 ;;; c := (ceiling a)
1946 ;;; if c < b
1947 ;;; then return c ; because a <= c < b, c integer
1948 ;;; else
1949 ;;; ; a is not integer (otherwise we would have had c = a < b)
1950 ;;; k := c-1 ; k = floor(a), k < a < b <= k+1
1951 ;;; return y = k + 1/fraction_between(1/(b-k), 1/(a-k))
1952 ;;; ; note 1 <= 1/(b-k) < 1/(a-k)
1954 ;;; You can see that we are actually computing a continued fraction expansion.
1956 ;;; Algorithm (iterative):
1957 ;;; If x is rational, return x.
1958 ;;; Call (integer-decode-float x). It returns a m,e,s (mantissa,
1959 ;;; exponent, sign).
1960 ;;; If m = 0 or e >= 0, return m*2^e*s. (This includes the case x = 0.0.)
1961 ;;; Create rational numbers a := (2*m-1)*2^(e-1) and b := (2*m+1)*2^(e-1)
1962 ;;; (positive and already in lowest terms because the denominator is a
1963 ;;; power of two and the numerator is odd).
1964 ;;; Start a continued fraction expansion
1965 ;;; p[-1] := 0, p[0] := 1, q[-1] := 1, q[0] := 0, i := 0.
1966 ;;; Loop
1967 ;;; c := (ceiling a)
1968 ;;; if c >= b
1969 ;;; then k := c-1, partial_quotient(k), (a,b) := (1/(b-k),1/(a-k)),
1970 ;;; goto Loop
1971 ;;; finally partial_quotient(c).
1972 ;;; Here partial_quotient(c) denotes the iteration
1973 ;;; i := i+1, p[i] := c*p[i-1]+p[i-2], q[i] := c*q[i-1]+q[i-2].
1974 ;;; At the end, return s * (p[i]/q[i]).
1975 ;;; This rational number is already in lowest terms because
1976 ;;; p[i]*q[i-1]-p[i-1]*q[i] = (-1)^i.
1978 (defmethod rationalize ((x bigfloat))
1979 (multiple-value-bind (frac expo sign)
1980 (integer-decode-float x)
1981 (cond ((or (zerop frac) (>= expo 0))
1982 (if (minusp sign)
1983 (- (ash frac expo))
1984 (ash frac expo)))
1986 ;; expo < 0 and (2*m-1) and (2*m+1) are coprime to 2^(1-e),
1987 ;; so build the fraction up immediately, without having to do
1988 ;; a gcd.
1989 (let ((a (/ (- (* 2 frac) 1) (ash 1 (- 1 expo))))
1990 (b (/ (+ (* 2 frac) 1) (ash 1 (- 1 expo))))
1991 (p0 0)
1992 (q0 1)
1993 (p1 1)
1994 (q1 0))
1995 (do ((c (ceiling a) (ceiling a)))
1996 ((< c b)
1997 (let ((top (+ (* c p1) p0))
1998 (bot (+ (* c q1) q0)))
1999 (/ (if (minusp sign)
2000 (- top)
2001 top)
2002 bot)))
2003 (let* ((k (- c 1))
2004 (p2 (+ (* k p1) p0))
2005 (q2 (+ (* k q1) q0)))
2006 (psetf a (/ (- b k))
2007 b (/ (- a k)))
2008 (setf p0 p1
2009 q0 q1
2010 p1 p2
2011 q1 q2))))))))
2013 (defun coerce (obj type)
2014 (flet ((coerce-error ()
2015 (error "Cannot coerce ~A to type ~S" obj type)))
2016 (cond ((typep obj type)
2017 obj)
2018 ((subtypep type 'bigfloat)
2019 ;; (coerce foo 'bigfloat). Foo has to be a real
2020 (cond ((typep obj 'real)
2021 (bigfloat obj))
2023 (coerce-error))))
2024 ((subtypep type 'complex-bigfloat)
2025 ;; (coerce foo 'complex-bigfloat). Foo has to be a real or complex
2026 (cond ((typep obj 'real)
2027 (bigfloat obj 0))
2028 ((typep obj 'cl:complex)
2029 (bigfloat obj))
2030 ((typep obj 'bigfloat)
2031 (bigfloat obj 0))
2033 (coerce-error))))
2034 ((typep obj 'bigfloat)
2035 ;; (coerce bigfloat foo)
2036 (cond ((subtypep type 'cl:float)
2037 (float obj (cl:coerce 0 type)))
2038 ((subtypep type '(cl:complex long-float))
2039 (cl:complex (float (realpart obj) 1l0)
2040 (float (imagpart obj) 1l0)))
2041 ((subtypep type '(cl:complex double-float))
2042 (cl:complex (float (realpart obj) 1d0)
2043 (float (imagpart obj) 1d0)))
2044 ((subtypep type '(cl:complex single-float))
2045 (cl:complex (float (realpart obj) 1f0)
2046 (float (imagpart obj) 1f0)))
2047 ((subtypep type 'cl:complex)
2048 ;; What should we do here? Return a
2049 ;; complex-bigfloat? A complex double-float?
2050 ;; complex single-float? I arbitrarily select
2051 ;; complex maxima:flonum for now.
2052 (cl:complex (float (realpart obj) 1.0)
2053 (float (imagpart obj) 1.0)))
2055 (coerce-error))))
2056 ((typep obj 'complex-bigfloat)
2057 ;; (coerce complex-bigfloat foo)
2058 (cond ((subtypep type 'complex-bigfloat)
2059 obj)
2060 ((subtypep type '(cl:complex long-float))
2061 (cl:complex (float (realpart obj) 1l0)
2062 (float (imagpart obj) 1l0)))
2063 ((subtypep type '(cl:complex double-float))
2064 (cl:complex (float (realpart obj) 1d0)
2065 (float (imagpart obj) 1d0)))
2066 ((subtypep type '(cl:complex single-float))
2067 (cl:complex (float (realpart obj) 1f0)
2068 (float (imagpart obj) 1f0)))
2070 (coerce-error))))
2072 (cl:coerce obj type)))))
2074 ;;; %PI - External
2076 ;;; Return a value of pi with the same precision as the argument.
2077 ;;; For rationals, we return a single-float approximation.
2078 (defmethod %pi ((x cl:rational))
2079 (cl:coerce cl:pi 'single-float))
2081 (defmethod %pi ((x cl:float))
2082 (cl:float cl:pi x))
2084 (defmethod %pi ((x bigfloat))
2085 (to (maxima::bcons (maxima::fppi))))
2087 (defmethod %pi ((x cl:complex))
2088 (cl:float cl:pi (realpart x)))
2090 (defmethod %pi ((x complex-bigfloat))
2091 (to (maxima::bcons (maxima::fppi))))
2093 ;;; %e - External
2095 ;;; Return a value of e with the same precision as the argument.
2096 ;;; For rationals, we return a single-float approximation.
2097 (defmethod %e ((x cl:rational))
2098 (cl:coerce maxima::%e-val 'single-float))
2100 (defmethod %e ((x cl:float))
2101 (cl:float maxima::%e-val x))
2103 (defmethod %e ((x bigfloat))
2104 (to (maxima::bcons (maxima::fpe))))
2106 (defmethod %e ((x cl:complex))
2107 (cl:float maxima::%e-val (realpart x)))
2109 (defmethod %e ((x complex-bigfloat))
2110 (to (maxima::bcons (maxima::fpe))))
2112 ;;;; Useful routines
2114 ;;; Evaluation of continued fractions
2116 (defvar *debug-cf-eval*
2118 "When true, enable some debugging prints when evaluating a
2119 continued fraction.")
2121 ;; Max number of iterations allowed when evaluating the continued
2122 ;; fraction. When this is reached, we assume that the continued
2123 ;; fraction did not converge.
2124 (defvar *max-cf-iterations*
2125 10000
2126 "Max number of iterations allowed when evaluating the continued
2127 fraction. When this is reached, we assume that the continued
2128 fraction did not converge.")
2130 ;;; LENTZ - External
2132 ;;; Lentz's algorithm for evaluating continued fractions.
2134 ;;; Let the continued fraction be:
2136 ;;; a1 a2 a3
2137 ;;; b0 + ---- ---- ----
2138 ;;; b1 + b2 + b3 +
2141 ;;; Then LENTZ expects two functions, each taking a single fixnum
2142 ;;; index. The first returns the b term and the second returns the a
2143 ;;; terms as above for a give n.
2144 (defun lentz (bf af)
2145 (let ((tiny-value-count 0))
2146 (flet ((value-or-tiny (v)
2147 ;; If v is zero, return a "tiny" number.
2148 (if (zerop v)
2149 (progn
2150 (incf tiny-value-count)
2151 (etypecase v
2152 ((or double-float cl:complex)
2153 (sqrt least-positive-normalized-double-float))
2154 ((or bigfloat complex-bigfloat)
2155 ;; What is a "tiny" bigfloat? Bigfloats have
2156 ;; unbounded exponents, so we need something
2157 ;; small, but not zero. Arbitrarily choose an
2158 ;; exponent of 50 times the precision.
2159 (expt 10 (- (* 50 maxima::$fpprec))))))
2160 v)))
2161 (let* ((f (value-or-tiny (funcall bf 0)))
2162 (c f)
2163 (d 0)
2164 (eps (epsilon f)))
2165 (loop
2166 for j from 1 upto *max-cf-iterations*
2167 for an = (funcall af j)
2168 for bn = (funcall bf j)
2169 do (progn
2170 (setf d (value-or-tiny (+ bn (* an d))))
2171 (setf c (value-or-tiny (+ bn (/ an c))))
2172 (when *debug-cf-eval*
2173 (format t "~&j = ~d~%" j)
2174 (format t " an = ~s~%" an)
2175 (format t " bn = ~s~%" bn)
2176 (format t " c = ~s~%" c)
2177 (format t " d = ~s~%" d))
2178 (let ((delta (/ c d)))
2179 (setf d (/ d))
2180 (setf f (* f delta))
2181 (when *debug-cf-eval*
2182 (format t " dl= ~S (|dl - 1| = ~S)~%" delta (abs (1- delta)))
2183 (format t " f = ~S~%" f))
2184 (when (<= (abs (- delta 1)) eps)
2185 (return-from lentz (values f j tiny-value-count)))))
2186 finally
2187 (error 'simple-error
2188 :format-control "~<Continued fraction failed to converge after ~D iterations.~% Delta = ~S~>"
2189 :format-arguments (list *max-cf-iterations* (/ c d))))))))
2191 ;;; SUM-POWER-SERIES - External
2193 ;;; SUM-POWER-SERIES sums the given power series, adding terms until
2194 ;;; the next term would not change the sum.
2196 ;;; The series to be summed is
2198 ;;; S = 1 + sum(c[k]*x^k, k, 1, inf)
2199 ;;; = 1 + sum(prod(f[n]*x, n, 1, k), k, 1, inf)
2201 ;;; where f[n] = c[n]/c[n-1].
2203 (defun sum-power-series (x f)
2204 (let ((eps (epsilon x)))
2205 (do* ((k 1 (+ 1 k))
2206 (sum 1 (+ sum term))
2207 (term (* x (funcall f 1))
2208 (* term x (funcall f k))))
2209 ((< (abs term) (* eps (abs sum)))
2210 sum)
2211 #+nil
2212 (format t "~4d: ~S ~S ~S~%" k sum term (funcall f k)))))
2214 ;; Format bigfloats using ~E format. This is suitable as a ~// format.
2216 ;; NOTE: This is a modified version of FORMAT-EXPONENTIAL from CMUCL to
2217 ;; support printing of bfloats.
2219 (defun format-e (stream number colonp atp
2220 &optional w d e k
2221 overflowchar padchar exponentchar)
2222 (typecase number
2223 (bigfloat
2224 (maxima::bfloat-format-e stream (real-value number) colonp atp
2225 w d e (or k 1)
2226 overflowchar
2227 (or padchar #\space)
2228 (or exponentchar #\b)))
2229 (complex-bigfloat
2230 ;; FIXME: Do something better than this since this doesn't honor
2231 ;; any of the parameters.
2232 (princ number stream))
2233 (otherwise
2234 ;; We were given some other kind of object. Just use CL's normal
2235 ;; ~E printer to print it.
2236 (let ((f
2237 (with-output-to-string (s)
2238 ;; Construct a suitable ~E format string from the given
2239 ;; parameters. First, handle w,d,e,k.
2240 (write-string "~V,V,V,V," s)
2241 (if overflowchar
2242 (format s "'~C," overflowchar)
2243 (write-string "," s))
2244 (if padchar
2245 (format s "'~C," padchar)
2246 (write-string "," s))
2247 (when exponentchar
2248 (format s "'~C" exponentchar))
2249 (when colonp
2250 (write-char #\: s))
2251 (when atp
2252 (write-char #\@ s))
2253 (write-char #\E s))))
2254 (format stream f w d e k number)))))
2257 (defmacro assert-equal (expected form)
2258 (let ((result (gensym))
2259 (e (gensym)))
2260 `(let ((,e ,expected)
2261 (,result ,form))
2262 (unless (equal ,e ,result)
2263 (format *debug-io* "Assertion failed: Expected ~S but got ~S~%" ,e ,result)))))
2265 (assert-equal " 0.990E+00" (format nil
2266 "~11,3,2,0,'*,,'E/bigfloat::format-e/"
2267 (bigfloat:bigfloat 99/100)))
2268 (assert-equal " 0.999E+00" (format nil
2269 "~11,3,2,0,'*,,'E/bigfloat::format-e/"
2270 (bigfloat:bigfloat 999/1000)))
2271 ;; Actually " 0.100E+01", but format-e doesn't round the output.
2272 (assert-equal " 0.999E+00" (format nil
2273 "~11,3,2,0,'*,,'E/bigfloat::format-e/"
2274 (bigfloat:bigfloat 9999/10000)))
2275 (assert-equal " 0.999E-04" (format nil
2276 "~11,3,2,0,'*,,'E/bigfloat::format-e/"
2277 (bigfloat:bigfloat 0000999/10000000)))
2278 ;; Actually " 0.100E-03", but format-e doesn't round the output.
2279 (assert-equal " 0.999E-0e" (format nil
2280 "~11,3,2,0,'*,,'E/bigfloat::format-e/"
2281 (bigfloat:bigfloat 00009999/100000000)))
2282 (assert-equal " 9.999E-05" (format nil
2283 "~11,3,2,,'*,,'E/bigfloat::format-e/"
2284 (bigfloat:bigfloat 00009999/100000000)))
2285 ;; Actually " 1.000E-04", but format-e doesn't round the output.
2286 (assert-equal " 9.999E-05" (format nil
2287 "~11,3,2,,'*,,'E/bigfloat::format-e/"
2288 (bigfloat:bigfloat 000099999/1000000000)))
2289 ;; All of these currently fail.
2290 (assert-equal ".00123d+6" (format nil
2291 "~9,,,-2/bigfloat::format-e/"
2292 (bigfloat:bigfloat 1.2345689d3)))
2293 (assert-equal "-.0012d+6" (format nil
2294 "~9,,,-2/bigfloat::format-e/"
2295 (bigfloat:bigfloat -1.2345689d3)))
2296 (assert-equal ".00123d+0" (format nil
2297 "~9,,,-2/bigfloat::format-e/"
2298 (bigfloat:bigfloat 1.2345689d-3)))
2299 (assert-equal "-.0012d+0" (format nil
2300 "~9,,,-2/bigfloat::format-e/"
2301 (bigfloat:bigfloat -1.2345689d-3)))
2303 ;; These fail because too many digits are printed and because the
2304 ;; scale factor isn't properly applied.
2305 (assert-equal ".00000003d+8" (format nil
2306 "~9,4,,-7E"
2307 (bigfloat:bigfloat pi)))
2308 (assert-equal ".000003d+6" (format nil
2309 "~9,4,,-5E"
2310 (bigfloat:bigfloat pi)))
2311 (assert-equal "3141600.d-6" (format nil
2312 "~5,4,,7E"
2313 (bigfloat:bigfloat pi)))
2314 (assert-equal " 314.16d-2" (format nil
2315 "~11,4,,3E"
2316 (bigfloat:bigfloat pi)))
2317 (assert-equal " 31416.d-4" (format nil
2318 "~11,4,,5E"
2319 (bigfloat:bigfloat pi)))
2320 (assert-equal " 0.3142d+1" (format nil
2321 "~11,4,,0E"
2322 (bigfloat:bigfloat pi)))
2323 (assert-equal ".03142d+2" (format nil
2324 "~9,,,-1E"
2325 (bigfloat:bigfloat pi)))
2326 (assert-equal "0.003141592653589793d+3" (format nil
2327 "~,,,-2E"
2328 (bigfloat:bigfloat pi)))
2329 (assert-equal "31.41592653589793d-1" (format nil
2330 "~,,,2E"
2331 (bigfloat:bigfloat pi)))
2332 ;; Fails because exponent is printed as "b0" instead of "b+0"
2333 (assert-equal "3.141592653589793b+0" (format nil "~E" (bigfloat:bigfloat pi)))
2336 ;; These fail because too many digits are printed and because the
2337 ;; scale factor isn't properly applied.
2338 (assert-equal ".03142d+2" (format nil "~9,5,,-1E" (bigfloat:bigfloat pi)))
2339 (assert-equal " 0.03142d+2" (format nil "~11,5,,-1E" (bigfloat:bigfloat pi)))
2340 (assert-equal "| 3141593.d-06|" (format nil "|~13,6,2,7E|" (bigfloat:bigfloat pi)))
2341 (assert-equal "0.314d+01" (format nil "~9,3,2,0,'%E" (bigfloat:bigfloat pi)))
2342 (assert-equal "+.003d+03" (format nil "~9,3,2,-2,'%@E" (bigfloat:bigfloat pi)))
2343 (assert-equal "+0.003d+03" (format nil "~10,3,2,-2,'%@E" (bigfloat:bigfloat pi)))
2344 (assert-equal "=====+0.003d+03" (format nil "~15,3,2,-2,'%,'=@E" (bigfloat:bigfloat pi)))
2345 (assert-equal "0.003d+03" (format nil "~9,3,2,-2,'%E" (bigfloat:bigfloat pi)))
2346 (assert-equal "%%%%%%%%" (format nil "~8,3,2,-2,'%@E" (bigfloat:bigfloat pi)))
2348 ;; Works
2349 (assert-equal "0.0f+0" (format nil "~e" 0))
2351 ;; Fails because exponent is printed as "b0" instead of "b+0'
2352 (assert-equal "0.0b+0" (format nil "~e" (bigfloat:bigfloat 0d0)))
2353 ;; Fails because exponent is printed as "b0 " instead of "b+0000"
2354 (assert-equal "0.0b+0000" (format nil "~9,,4e" (bigfloat:bigfloat 0d0)))
2355 ;; Fails because exponent is printed as "b4" isntead of "b+4"
2356 (assert-equal "1.2345678901234567b+4" (format nil "~E"
2357 (bigfloat:bigfloat 1.234567890123456789d4)))
2359 ;; Fails because exponent is printed as "b36" instead of "b+36"
2360 (assert-equal "1.32922799578492b+36" (format nil "~20E"
2361 (bigfloat:bigfloat (expt 2d0 120))))
2362 ;; Fails because too many digits are printed and the exponent doesn't include "+".
2363 (assert-equal " 1.32922800b+36" (format nil "~21,8E"
2364 (bigfloat:bigfloat (expt 2d0 120))))
2368 ;; Format bigfloats using ~F format. This is suitable as a ~// format.
2370 ;; NOTE: This is a modified version of FORMAT-FIXED from CMUCL to
2371 ;; support printing of bfloats.
2373 (defun format-f (stream number colonp atp
2374 &optional w d k overflowchar padchar)
2375 (typecase number
2376 (bigfloat
2377 (maxima::bfloat-format-f stream (real-value number) colonp atp
2378 w d (or k 0)
2379 overflowchar
2380 (or padchar #\space)))
2381 (complex-bigfloat
2382 ;; FIXME: Do something better than this since this doesn't honor
2383 ;; any of the parameters.
2384 (princ number stream))
2385 (otherwise
2386 ;; We were given some other kind of object. Just use CL's normal
2387 ;; ~F printer to print it.
2388 (let ((f
2389 (with-output-to-string (s)
2390 ;; Construct a suitable ~F format string from the given
2391 ;; parameters. First handle w,d,k.
2392 (write-string "~V,V,V," s)
2393 (if overflowchar
2394 (format s "'~C," overflowchar)
2395 (write-string "," s))
2396 (if (char= padchar #\space)
2397 (write-string "," s)
2398 (format s "'~C," padchar))
2399 (when colonp
2400 (write-char #\: s))
2401 (when atp
2402 (write-char #\@ s))
2403 (write-char #\F s))))
2404 (format stream f w d k number)))))
2406 ;; Format bigfloats using ~G format. This is suitable as a ~// format.
2408 ;; NOTE: This is a modified version of FORMAT-GENERAL from CMUCL to
2409 ;; support printing of bfloats.
2411 (defun format-g (stream number colonp atp
2412 &optional w d e k overflowchar padchar marker)
2413 (typecase number
2414 (bigfloat
2415 (maxima::bfloat-format-g stream (real-value number) colonp atp
2416 w d e (or k 1)
2417 overflowchar
2418 (or padchar #\space)
2419 (or marker #\b)))
2420 (complex-bigfloat
2421 ;; FIXME: Do something better than this since this doesn't honor
2422 ;; any of the parameters.
2423 (princ number stream))
2424 (otherwise
2425 ;; We were given some other kind of object. Just use CL's normal
2426 ;; ~G printer to print it.
2427 (let ((f
2428 (with-output-to-string (s)
2429 ;; Construct a suitable ~E format string from the given
2430 ;; parameters. First, handle w,d,e,k.
2431 (write-string "~V,V,V,V," s)
2432 (if overflowchar
2433 (format s "'~C," overflowchar)
2434 (write-string "," s))
2435 (if padchar
2436 (format s "'~C," padchar)
2437 (write-string "," s))
2438 (when marker
2439 (format s "'~C" marker))
2440 (when colonp
2441 (write-char #\: s))
2442 (when atp
2443 (write-char #\@ s))
2444 (write-char #\G s))))
2445 (format stream f w d e k number)))))