This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Integrate from maint:
[perl5.git] / numeric.c
CommitLineData
98994639
HS
1/* numeric.c
2 *
4bb101f2
JH
3 * Copyright (C) 1993, 1994, 1995, 1996, 1997, 1998, 1999,
4 * 2000, 2001, 2002, 2003, by Larry Wall and others
98994639
HS
5 *
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
8 *
9 */
10
11/*
12 * "That only makes eleven (plus one mislaid) and not fourteen, unless
13 * wizards count differently to other people."
14 */
15
ccfc67b7
JH
16/*
17=head1 Numeric functions
18*/
19
98994639
HS
20#include "EXTERN.h"
21#define PERL_IN_NUMERIC_C
22#include "perl.h"
23
24U32
25Perl_cast_ulong(pTHX_ NV f)
26{
27 if (f < 0.0)
28 return f < I32_MIN ? (U32) I32_MIN : (U32)(I32) f;
29 if (f < U32_MAX_P1) {
30#if CASTFLAGS & 2
31 if (f < U32_MAX_P1_HALF)
32 return (U32) f;
33 f -= U32_MAX_P1_HALF;
34 return ((U32) f) | (1 + U32_MAX >> 1);
35#else
36 return (U32) f;
37#endif
38 }
39 return f > 0 ? U32_MAX : 0 /* NaN */;
40}
41
42I32
43Perl_cast_i32(pTHX_ NV f)
44{
45 if (f < I32_MAX_P1)
46 return f < I32_MIN ? I32_MIN : (I32) f;
47 if (f < U32_MAX_P1) {
48#if CASTFLAGS & 2
49 if (f < U32_MAX_P1_HALF)
50 return (I32)(U32) f;
51 f -= U32_MAX_P1_HALF;
52 return (I32)(((U32) f) | (1 + U32_MAX >> 1));
53#else
54 return (I32)(U32) f;
55#endif
56 }
57 return f > 0 ? (I32)U32_MAX : 0 /* NaN */;
58}
59
60IV
61Perl_cast_iv(pTHX_ NV f)
62{
63 if (f < IV_MAX_P1)
64 return f < IV_MIN ? IV_MIN : (IV) f;
65 if (f < UV_MAX_P1) {
66#if CASTFLAGS & 2
67 /* For future flexibility allowing for sizeof(UV) >= sizeof(IV) */
68 if (f < UV_MAX_P1_HALF)
69 return (IV)(UV) f;
70 f -= UV_MAX_P1_HALF;
71 return (IV)(((UV) f) | (1 + UV_MAX >> 1));
72#else
73 return (IV)(UV) f;
74#endif
75 }
76 return f > 0 ? (IV)UV_MAX : 0 /* NaN */;
77}
78
79UV
80Perl_cast_uv(pTHX_ NV f)
81{
82 if (f < 0.0)
83 return f < IV_MIN ? (UV) IV_MIN : (UV)(IV) f;
84 if (f < UV_MAX_P1) {
85#if CASTFLAGS & 2
86 if (f < UV_MAX_P1_HALF)
87 return (UV) f;
88 f -= UV_MAX_P1_HALF;
89 return ((UV) f) | (1 + UV_MAX >> 1);
90#else
91 return (UV) f;
92#endif
93 }
94 return f > 0 ? UV_MAX : 0 /* NaN */;
95}
96
97#if defined(HUGE_VAL) || (defined(USE_LONG_DOUBLE) && defined(HUGE_VALL))
98/*
99 * This hack is to force load of "huge" support from libm.a
100 * So it is in perl for (say) POSIX to use.
101 * Needed for SunOS with Sun's 'acc' for example.
102 */
103NV
104Perl_huge(void)
105{
106# if defined(USE_LONG_DOUBLE) && defined(HUGE_VALL)
107 return HUGE_VALL;
108# endif
109 return HUGE_VAL;
110}
111#endif
112
53305cf1
NC
113/*
114=for apidoc grok_bin
98994639 115
53305cf1
NC
116converts a string representing a binary number to numeric form.
117
118On entry I<start> and I<*len> give the string to scan, I<*flags> gives
119conversion flags, and I<result> should be NULL or a pointer to an NV.
120The scan stops at the end of the string, or the first invalid character.
121On return I<*len> is set to the length scanned string, and I<*flags> gives
122output flags.
123
124If the value is <= UV_MAX it is returned as a UV, the output flags are clear,
125and nothing is written to I<*result>. If the value is > UV_MAX C<grok_bin>
126returns UV_MAX, sets C<PERL_SCAN_GREATER_THAN_UV_MAX> in the output flags,
127and writes the value to I<*result> (or the value is discarded if I<result>
128is NULL).
129
d1be9408 130The hex number may optionally be prefixed with "0b" or "b" unless
a4c04bdc
NC
131C<PERL_SCAN_DISALLOW_PREFIX> is set in I<*flags> on entry. If
132C<PERL_SCAN_ALLOW_UNDERSCORES> is set in I<*flags> then the binary
53305cf1
NC
133number may use '_' characters to separate digits.
134
135=cut
136 */
137
138UV
139Perl_grok_bin(pTHX_ char *start, STRLEN *len_p, I32 *flags, NV *result) {
140 const char *s = start;
141 STRLEN len = *len_p;
142 UV value = 0;
143 NV value_nv = 0;
144
145 const UV max_div_2 = UV_MAX / 2;
146 bool allow_underscores = *flags & PERL_SCAN_ALLOW_UNDERSCORES;
147 bool overflowed = FALSE;
148
a4c04bdc
NC
149 if (!(*flags & PERL_SCAN_DISALLOW_PREFIX)) {
150 /* strip off leading b or 0b.
151 for compatibility silently suffer "b" and "0b" as valid binary
152 numbers. */
153 if (len >= 1) {
154 if (s[0] == 'b') {
155 s++;
156 len--;
157 }
158 else if (len >= 2 && s[0] == '0' && s[1] == 'b') {
159 s+=2;
160 len-=2;
161 }
162 }
53305cf1
NC
163 }
164
165 for (; len-- && *s; s++) {
166 char bit = *s;
167 if (bit == '0' || bit == '1') {
168 /* Write it in this wonky order with a goto to attempt to get the
169 compiler to make the common case integer-only loop pretty tight.
170 With gcc seems to be much straighter code than old scan_bin. */
171 redo:
172 if (!overflowed) {
173 if (value <= max_div_2) {
174 value = (value << 1) | (bit - '0');
175 continue;
176 }
177 /* Bah. We're just overflowed. */
178 if (ckWARN_d(WARN_OVERFLOW))
9014280d 179 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
53305cf1
NC
180 "Integer overflow in binary number");
181 overflowed = TRUE;
182 value_nv = (NV) value;
183 }
184 value_nv *= 2.0;
98994639 185 /* If an NV has not enough bits in its mantissa to
d1be9408 186 * represent a UV this summing of small low-order numbers
98994639
HS
187 * is a waste of time (because the NV cannot preserve
188 * the low-order bits anyway): we could just remember when
53305cf1 189 * did we overflow and in the end just multiply value_nv by the
98994639 190 * right amount. */
53305cf1
NC
191 value_nv += (NV)(bit - '0');
192 continue;
193 }
194 if (bit == '_' && len && allow_underscores && (bit = s[1])
195 && (bit == '0' || bit == '1'))
98994639
HS
196 {
197 --len;
198 ++s;
53305cf1 199 goto redo;
98994639 200 }
94dd8549 201 if (!(*flags & PERL_SCAN_SILENT_ILLDIGIT) && ckWARN(WARN_DIGIT))
9014280d 202 Perl_warner(aTHX_ packWARN(WARN_DIGIT),
53305cf1
NC
203 "Illegal binary digit '%c' ignored", *s);
204 break;
98994639 205 }
53305cf1
NC
206
207 if ( ( overflowed && value_nv > 4294967295.0)
98994639 208#if UVSIZE > 4
53305cf1 209 || (!overflowed && value > 0xffffffff )
98994639
HS
210#endif
211 ) {
212 if (ckWARN(WARN_PORTABLE))
9014280d 213 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
53305cf1
NC
214 "Binary number > 0b11111111111111111111111111111111 non-portable");
215 }
216 *len_p = s - start;
217 if (!overflowed) {
218 *flags = 0;
219 return value;
98994639 220 }
53305cf1
NC
221 *flags = PERL_SCAN_GREATER_THAN_UV_MAX;
222 if (result)
223 *result = value_nv;
224 return UV_MAX;
98994639
HS
225}
226
53305cf1
NC
227/*
228=for apidoc grok_hex
229
230converts a string representing a hex number to numeric form.
231
232On entry I<start> and I<*len> give the string to scan, I<*flags> gives
233conversion flags, and I<result> should be NULL or a pointer to an NV.
234The scan stops at the end of the string, or the first non-hex-digit character.
235On return I<*len> is set to the length scanned string, and I<*flags> gives
236output flags.
237
238If the value is <= UV_MAX it is returned as a UV, the output flags are clear,
239and nothing is written to I<*result>. If the value is > UV_MAX C<grok_hex>
240returns UV_MAX, sets C<PERL_SCAN_GREATER_THAN_UV_MAX> in the output flags,
241and writes the value to I<*result> (or the value is discarded if I<result>
242is NULL).
243
d1be9408 244The hex number may optionally be prefixed with "0x" or "x" unless
a4c04bdc
NC
245C<PERL_SCAN_DISALLOW_PREFIX> is set in I<*flags> on entry. If
246C<PERL_SCAN_ALLOW_UNDERSCORES> is set in I<*flags> then the hex
53305cf1
NC
247number may use '_' characters to separate digits.
248
249=cut
250 */
251
252UV
253Perl_grok_hex(pTHX_ char *start, STRLEN *len_p, I32 *flags, NV *result) {
254 const char *s = start;
255 STRLEN len = *len_p;
256 UV value = 0;
257 NV value_nv = 0;
258
259 const UV max_div_16 = UV_MAX / 16;
260 bool allow_underscores = *flags & PERL_SCAN_ALLOW_UNDERSCORES;
261 bool overflowed = FALSE;
262 const char *hexdigit;
98994639 263
a4c04bdc
NC
264 if (!(*flags & PERL_SCAN_DISALLOW_PREFIX)) {
265 /* strip off leading x or 0x.
266 for compatibility silently suffer "x" and "0x" as valid hex numbers.
267 */
268 if (len >= 1) {
269 if (s[0] == 'x') {
270 s++;
271 len--;
272 }
273 else if (len >= 2 && s[0] == '0' && s[1] == 'x') {
274 s+=2;
275 len-=2;
276 }
277 }
98994639
HS
278 }
279
280 for (; len-- && *s; s++) {
281 hexdigit = strchr((char *) PL_hexdigit, *s);
53305cf1
NC
282 if (hexdigit) {
283 /* Write it in this wonky order with a goto to attempt to get the
284 compiler to make the common case integer-only loop pretty tight.
285 With gcc seems to be much straighter code than old scan_hex. */
286 redo:
287 if (!overflowed) {
288 if (value <= max_div_16) {
289 value = (value << 4) | ((hexdigit - PL_hexdigit) & 15);
290 continue;
291 }
292 /* Bah. We're just overflowed. */
293 if (ckWARN_d(WARN_OVERFLOW))
9014280d 294 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
53305cf1
NC
295 "Integer overflow in hexadecimal number");
296 overflowed = TRUE;
297 value_nv = (NV) value;
298 }
299 value_nv *= 16.0;
300 /* If an NV has not enough bits in its mantissa to
d1be9408 301 * represent a UV this summing of small low-order numbers
53305cf1
NC
302 * is a waste of time (because the NV cannot preserve
303 * the low-order bits anyway): we could just remember when
304 * did we overflow and in the end just multiply value_nv by the
305 * right amount of 16-tuples. */
306 value_nv += (NV)((hexdigit - PL_hexdigit) & 15);
307 continue;
308 }
309 if (*s == '_' && len && allow_underscores && s[1]
98994639
HS
310 && (hexdigit = strchr((char *) PL_hexdigit, s[1])))
311 {
312 --len;
313 ++s;
53305cf1 314 goto redo;
98994639 315 }
94dd8549 316 if (!(*flags & PERL_SCAN_SILENT_ILLDIGIT) && ckWARN(WARN_DIGIT))
9014280d 317 Perl_warner(aTHX_ packWARN(WARN_DIGIT),
53305cf1
NC
318 "Illegal hexadecimal digit '%c' ignored", *s);
319 break;
320 }
321
322 if ( ( overflowed && value_nv > 4294967295.0)
323#if UVSIZE > 4
324 || (!overflowed && value > 0xffffffff )
325#endif
326 ) {
327 if (ckWARN(WARN_PORTABLE))
9014280d 328 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
53305cf1
NC
329 "Hexadecimal number > 0xffffffff non-portable");
330 }
331 *len_p = s - start;
332 if (!overflowed) {
333 *flags = 0;
334 return value;
335 }
336 *flags = PERL_SCAN_GREATER_THAN_UV_MAX;
337 if (result)
338 *result = value_nv;
339 return UV_MAX;
340}
341
342/*
343=for apidoc grok_oct
344
345
346=cut
347 */
348
349UV
350Perl_grok_oct(pTHX_ char *start, STRLEN *len_p, I32 *flags, NV *result) {
351 const char *s = start;
352 STRLEN len = *len_p;
353 UV value = 0;
354 NV value_nv = 0;
355
356 const UV max_div_8 = UV_MAX / 8;
357 bool allow_underscores = *flags & PERL_SCAN_ALLOW_UNDERSCORES;
358 bool overflowed = FALSE;
359
360 for (; len-- && *s; s++) {
361 /* gcc 2.95 optimiser not smart enough to figure that this subtraction
362 out front allows slicker code. */
363 int digit = *s - '0';
364 if (digit >= 0 && digit <= 7) {
365 /* Write it in this wonky order with a goto to attempt to get the
366 compiler to make the common case integer-only loop pretty tight.
367 */
368 redo:
369 if (!overflowed) {
370 if (value <= max_div_8) {
371 value = (value << 3) | digit;
372 continue;
373 }
374 /* Bah. We're just overflowed. */
375 if (ckWARN_d(WARN_OVERFLOW))
9014280d 376 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
53305cf1
NC
377 "Integer overflow in octal number");
378 overflowed = TRUE;
379 value_nv = (NV) value;
380 }
381 value_nv *= 8.0;
98994639 382 /* If an NV has not enough bits in its mantissa to
d1be9408 383 * represent a UV this summing of small low-order numbers
98994639
HS
384 * is a waste of time (because the NV cannot preserve
385 * the low-order bits anyway): we could just remember when
53305cf1
NC
386 * did we overflow and in the end just multiply value_nv by the
387 * right amount of 8-tuples. */
388 value_nv += (NV)digit;
389 continue;
390 }
391 if (digit == ('_' - '0') && len && allow_underscores
392 && (digit = s[1] - '0') && (digit >= 0 && digit <= 7))
393 {
394 --len;
395 ++s;
396 goto redo;
397 }
398 /* Allow \octal to work the DWIM way (that is, stop scanning
399 * as soon as non-octal characters are seen, complain only iff
400 * someone seems to want to use the digits eight and nine). */
401 if (digit == 8 || digit == 9) {
94dd8549 402 if (!(*flags & PERL_SCAN_SILENT_ILLDIGIT) && ckWARN(WARN_DIGIT))
9014280d 403 Perl_warner(aTHX_ packWARN(WARN_DIGIT),
53305cf1
NC
404 "Illegal octal digit '%c' ignored", *s);
405 }
406 break;
98994639 407 }
53305cf1
NC
408
409 if ( ( overflowed && value_nv > 4294967295.0)
98994639 410#if UVSIZE > 4
53305cf1 411 || (!overflowed && value > 0xffffffff )
98994639
HS
412#endif
413 ) {
414 if (ckWARN(WARN_PORTABLE))
9014280d 415 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
53305cf1
NC
416 "Octal number > 037777777777 non-portable");
417 }
418 *len_p = s - start;
419 if (!overflowed) {
420 *flags = 0;
421 return value;
98994639 422 }
53305cf1
NC
423 *flags = PERL_SCAN_GREATER_THAN_UV_MAX;
424 if (result)
425 *result = value_nv;
426 return UV_MAX;
427}
428
429/*
430=for apidoc scan_bin
431
432For backwards compatibility. Use C<grok_bin> instead.
433
434=for apidoc scan_hex
435
436For backwards compatibility. Use C<grok_hex> instead.
437
438=for apidoc scan_oct
439
440For backwards compatibility. Use C<grok_oct> instead.
441
442=cut
443 */
444
445NV
446Perl_scan_bin(pTHX_ char *start, STRLEN len, STRLEN *retlen)
447{
448 NV rnv;
449 I32 flags = *retlen ? PERL_SCAN_ALLOW_UNDERSCORES : 0;
450 UV ruv = grok_bin (start, &len, &flags, &rnv);
451
452 *retlen = len;
453 return (flags & PERL_SCAN_GREATER_THAN_UV_MAX) ? rnv : (NV)ruv;
454}
455
456NV
457Perl_scan_oct(pTHX_ char *start, STRLEN len, STRLEN *retlen)
458{
459 NV rnv;
460 I32 flags = *retlen ? PERL_SCAN_ALLOW_UNDERSCORES : 0;
461 UV ruv = grok_oct (start, &len, &flags, &rnv);
462
463 *retlen = len;
464 return (flags & PERL_SCAN_GREATER_THAN_UV_MAX) ? rnv : (NV)ruv;
465}
466
467NV
468Perl_scan_hex(pTHX_ char *start, STRLEN len, STRLEN *retlen)
469{
470 NV rnv;
471 I32 flags = *retlen ? PERL_SCAN_ALLOW_UNDERSCORES : 0;
472 UV ruv = grok_hex (start, &len, &flags, &rnv);
473
474 *retlen = len;
475 return (flags & PERL_SCAN_GREATER_THAN_UV_MAX) ? rnv : (NV)ruv;
98994639
HS
476}
477
478/*
479=for apidoc grok_numeric_radix
480
481Scan and skip for a numeric decimal separator (radix).
482
483=cut
484 */
485bool
486Perl_grok_numeric_radix(pTHX_ const char **sp, const char *send)
487{
488#ifdef USE_LOCALE_NUMERIC
489 if (PL_numeric_radix_sv && IN_LOCALE) {
490 STRLEN len;
491 char* radix = SvPV(PL_numeric_radix_sv, len);
492 if (*sp + len <= send && memEQ(*sp, radix, len)) {
493 *sp += len;
494 return TRUE;
495 }
496 }
497 /* always try "." if numeric radix didn't match because
498 * we may have data from different locales mixed */
499#endif
500 if (*sp < send && **sp == '.') {
501 ++*sp;
502 return TRUE;
503 }
504 return FALSE;
505}
506
507/*
508=for apidoc grok_number
509
510Recognise (or not) a number. The type of the number is returned
511(0 if unrecognised), otherwise it is a bit-ORed combination of
512IS_NUMBER_IN_UV, IS_NUMBER_GREATER_THAN_UV_MAX, IS_NUMBER_NOT_INT,
aa8b85de 513IS_NUMBER_NEG, IS_NUMBER_INFINITY, IS_NUMBER_NAN (defined in perl.h).
60939fb8
NC
514
515If the value of the number can fit an in UV, it is returned in the *valuep
516IS_NUMBER_IN_UV will be set to indicate that *valuep is valid, IS_NUMBER_IN_UV
517will never be set unless *valuep is valid, but *valuep may have been assigned
518to during processing even though IS_NUMBER_IN_UV is not set on return.
519If valuep is NULL, IS_NUMBER_IN_UV will be set for the same cases as when
520valuep is non-NULL, but no actual assignment (or SEGV) will occur.
521
522IS_NUMBER_NOT_INT will be set with IS_NUMBER_IN_UV if trailing decimals were
523seen (in which case *valuep gives the true value truncated to an integer), and
524IS_NUMBER_NEG if the number is negative (in which case *valuep holds the
525absolute value). IS_NUMBER_IN_UV is not set if e notation was used or the
526number is larger than a UV.
98994639
HS
527
528=cut
529 */
530int
531Perl_grok_number(pTHX_ const char *pv, STRLEN len, UV *valuep)
532{
60939fb8
NC
533 const char *s = pv;
534 const char *send = pv + len;
535 const UV max_div_10 = UV_MAX / 10;
536 const char max_mod_10 = UV_MAX % 10;
537 int numtype = 0;
538 int sawinf = 0;
aa8b85de 539 int sawnan = 0;
60939fb8
NC
540
541 while (s < send && isSPACE(*s))
542 s++;
543 if (s == send) {
544 return 0;
545 } else if (*s == '-') {
546 s++;
547 numtype = IS_NUMBER_NEG;
548 }
549 else if (*s == '+')
550 s++;
551
552 if (s == send)
553 return 0;
554
555 /* next must be digit or the radix separator or beginning of infinity */
556 if (isDIGIT(*s)) {
557 /* UVs are at least 32 bits, so the first 9 decimal digits cannot
558 overflow. */
559 UV value = *s - '0';
560 /* This construction seems to be more optimiser friendly.
561 (without it gcc does the isDIGIT test and the *s - '0' separately)
562 With it gcc on arm is managing 6 instructions (6 cycles) per digit.
563 In theory the optimiser could deduce how far to unroll the loop
564 before checking for overflow. */
58bb9ec3
NC
565 if (++s < send) {
566 int digit = *s - '0';
60939fb8
NC
567 if (digit >= 0 && digit <= 9) {
568 value = value * 10 + digit;
58bb9ec3
NC
569 if (++s < send) {
570 digit = *s - '0';
60939fb8
NC
571 if (digit >= 0 && digit <= 9) {
572 value = value * 10 + digit;
58bb9ec3
NC
573 if (++s < send) {
574 digit = *s - '0';
60939fb8
NC
575 if (digit >= 0 && digit <= 9) {
576 value = value * 10 + digit;
58bb9ec3
NC
577 if (++s < send) {
578 digit = *s - '0';
60939fb8
NC
579 if (digit >= 0 && digit <= 9) {
580 value = value * 10 + digit;
58bb9ec3
NC
581 if (++s < send) {
582 digit = *s - '0';
60939fb8
NC
583 if (digit >= 0 && digit <= 9) {
584 value = value * 10 + digit;
58bb9ec3
NC
585 if (++s < send) {
586 digit = *s - '0';
60939fb8
NC
587 if (digit >= 0 && digit <= 9) {
588 value = value * 10 + digit;
58bb9ec3
NC
589 if (++s < send) {
590 digit = *s - '0';
60939fb8
NC
591 if (digit >= 0 && digit <= 9) {
592 value = value * 10 + digit;
58bb9ec3
NC
593 if (++s < send) {
594 digit = *s - '0';
60939fb8
NC
595 if (digit >= 0 && digit <= 9) {
596 value = value * 10 + digit;
58bb9ec3 597 if (++s < send) {
60939fb8
NC
598 /* Now got 9 digits, so need to check
599 each time for overflow. */
58bb9ec3 600 digit = *s - '0';
60939fb8
NC
601 while (digit >= 0 && digit <= 9
602 && (value < max_div_10
603 || (value == max_div_10
604 && digit <= max_mod_10))) {
605 value = value * 10 + digit;
58bb9ec3
NC
606 if (++s < send)
607 digit = *s - '0';
60939fb8
NC
608 else
609 break;
610 }
611 if (digit >= 0 && digit <= 9
51bd16da 612 && (s < send)) {
60939fb8
NC
613 /* value overflowed.
614 skip the remaining digits, don't
615 worry about setting *valuep. */
616 do {
617 s++;
618 } while (s < send && isDIGIT(*s));
619 numtype |=
620 IS_NUMBER_GREATER_THAN_UV_MAX;
621 goto skip_value;
622 }
623 }
624 }
98994639 625 }
60939fb8
NC
626 }
627 }
628 }
629 }
630 }
631 }
632 }
633 }
634 }
635 }
636 }
98994639 637 }
60939fb8 638 }
98994639 639 }
60939fb8
NC
640 numtype |= IS_NUMBER_IN_UV;
641 if (valuep)
642 *valuep = value;
643
644 skip_value:
645 if (GROK_NUMERIC_RADIX(&s, send)) {
646 numtype |= IS_NUMBER_NOT_INT;
647 while (s < send && isDIGIT(*s)) /* optional digits after the radix */
648 s++;
98994639 649 }
60939fb8
NC
650 }
651 else if (GROK_NUMERIC_RADIX(&s, send)) {
652 numtype |= IS_NUMBER_NOT_INT | IS_NUMBER_IN_UV; /* valuep assigned below */
653 /* no digits before the radix means we need digits after it */
654 if (s < send && isDIGIT(*s)) {
655 do {
656 s++;
657 } while (s < send && isDIGIT(*s));
658 if (valuep) {
659 /* integer approximation is valid - it's 0. */
660 *valuep = 0;
661 }
98994639 662 }
60939fb8
NC
663 else
664 return 0;
665 } else if (*s == 'I' || *s == 'i') {
666 s++; if (s == send || (*s != 'N' && *s != 'n')) return 0;
667 s++; if (s == send || (*s != 'F' && *s != 'f')) return 0;
668 s++; if (s < send && (*s == 'I' || *s == 'i')) {
669 s++; if (s == send || (*s != 'N' && *s != 'n')) return 0;
670 s++; if (s == send || (*s != 'I' && *s != 'i')) return 0;
671 s++; if (s == send || (*s != 'T' && *s != 't')) return 0;
672 s++; if (s == send || (*s != 'Y' && *s != 'y')) return 0;
673 s++;
98994639 674 }
60939fb8 675 sawinf = 1;
aa8b85de
JH
676 } else if (*s == 'N' || *s == 'n') {
677 /* XXX TODO: There are signaling NaNs and quiet NaNs. */
678 s++; if (s == send || (*s != 'A' && *s != 'a')) return 0;
679 s++; if (s == send || (*s != 'N' && *s != 'n')) return 0;
680 s++;
681 sawnan = 1;
682 } else
98994639 683 return 0;
60939fb8
NC
684
685 if (sawinf) {
686 numtype &= IS_NUMBER_NEG; /* Keep track of sign */
687 numtype |= IS_NUMBER_INFINITY | IS_NUMBER_NOT_INT;
aa8b85de
JH
688 } else if (sawnan) {
689 numtype &= IS_NUMBER_NEG; /* Keep track of sign */
690 numtype |= IS_NUMBER_NAN | IS_NUMBER_NOT_INT;
60939fb8
NC
691 } else if (s < send) {
692 /* we can have an optional exponent part */
693 if (*s == 'e' || *s == 'E') {
694 /* The only flag we keep is sign. Blow away any "it's UV" */
695 numtype &= IS_NUMBER_NEG;
696 numtype |= IS_NUMBER_NOT_INT;
697 s++;
698 if (s < send && (*s == '-' || *s == '+'))
699 s++;
700 if (s < send && isDIGIT(*s)) {
701 do {
702 s++;
703 } while (s < send && isDIGIT(*s));
704 }
705 else
706 return 0;
707 }
708 }
709 while (s < send && isSPACE(*s))
710 s++;
711 if (s >= send)
aa8b85de 712 return numtype;
60939fb8
NC
713 if (len == 10 && memEQ(pv, "0 but true", 10)) {
714 if (valuep)
715 *valuep = 0;
716 return IS_NUMBER_IN_UV;
717 }
718 return 0;
98994639
HS
719}
720
4801ca72 721STATIC NV
98994639
HS
722S_mulexp10(NV value, I32 exponent)
723{
724 NV result = 1.0;
725 NV power = 10.0;
726 bool negative = 0;
727 I32 bit;
728
729 if (exponent == 0)
730 return value;
20f6aaab
AS
731 if (value == 0)
732 return 0;
87032ba1 733
24866caa 734 /* On OpenVMS VAX we by default use the D_FLOAT double format,
67597c89 735 * and that format does not have *easy* capabilities [1] for
24866caa
CB
736 * overflowing doubles 'silently' as IEEE fp does. We also need
737 * to support G_FLOAT on both VAX and Alpha, and though the exponent
738 * range is much larger than D_FLOAT it still doesn't do silent
739 * overflow. Therefore we need to detect early whether we would
740 * overflow (this is the behaviour of the native string-to-float
741 * conversion routines, and therefore of native applications, too).
67597c89 742 *
24866caa
CB
743 * [1] Trying to establish a condition handler to trap floating point
744 * exceptions is not a good idea. */
87032ba1
JH
745
746 /* In UNICOS and in certain Cray models (such as T90) there is no
747 * IEEE fp, and no way at all from C to catch fp overflows gracefully.
748 * There is something you can do if you are willing to use some
749 * inline assembler: the instruction is called DFI-- but that will
750 * disable *all* floating point interrupts, a little bit too large
751 * a hammer. Therefore we need to catch potential overflows before
752 * it's too late. */
353813d9
HS
753
754#if ((defined(VMS) && !defined(__IEEE_FP)) || defined(_UNICOS)) && defined(NV_MAX_10_EXP)
755 STMT_START {
756 NV exp_v = log10(value);
757 if (exponent >= NV_MAX_10_EXP || exponent + exp_v >= NV_MAX_10_EXP)
758 return NV_MAX;
759 if (exponent < 0) {
760 if (-(exponent + exp_v) >= NV_MAX_10_EXP)
761 return 0.0;
762 while (-exponent >= NV_MAX_10_EXP) {
763 /* combination does not overflow, but 10^(-exponent) does */
764 value /= 10;
765 ++exponent;
766 }
767 }
768 } STMT_END;
87032ba1
JH
769#endif
770
353813d9
HS
771 if (exponent < 0) {
772 negative = 1;
773 exponent = -exponent;
774 }
98994639
HS
775 for (bit = 1; exponent; bit <<= 1) {
776 if (exponent & bit) {
777 exponent ^= bit;
778 result *= power;
236f0012
CB
779 /* Floating point exceptions are supposed to be turned off,
780 * but if we're obviously done, don't risk another iteration.
781 */
782 if (exponent == 0) break;
98994639
HS
783 }
784 power *= power;
785 }
786 return negative ? value / result : value * result;
787}
788
789NV
790Perl_my_atof(pTHX_ const char* s)
791{
792 NV x = 0.0;
793#ifdef USE_LOCALE_NUMERIC
794 if (PL_numeric_local && IN_LOCALE) {
795 NV y;
796
797 /* Scan the number twice; once using locale and once without;
798 * choose the larger result (in absolute value). */
a36244b7 799 Perl_atof2(s, x);
98994639 800 SET_NUMERIC_STANDARD();
a36244b7 801 Perl_atof2(s, y);
98994639
HS
802 SET_NUMERIC_LOCAL();
803 if ((y < 0.0 && y < x) || (y > 0.0 && y > x))
804 return y;
805 }
806 else
a36244b7 807 Perl_atof2(s, x);
98994639 808#else
a36244b7 809 Perl_atof2(s, x);
98994639
HS
810#endif
811 return x;
812}
813
814char*
815Perl_my_atof2(pTHX_ const char* orig, NV* value)
816{
20f6aaab 817 NV result[3] = {0.0, 0.0, 0.0};
98994639 818 char* s = (char*)orig;
a36244b7 819#ifdef USE_PERL_ATOF
20f6aaab 820 UV accumulator[2] = {0,0}; /* before/after dp */
a36244b7 821 bool negative = 0;
98994639 822 char* send = s + strlen(orig) - 1;
8194bf88 823 bool seen_digit = 0;
20f6aaab
AS
824 I32 exp_adjust[2] = {0,0};
825 I32 exp_acc[2] = {-1, -1};
826 /* the current exponent adjust for the accumulators */
98994639 827 I32 exponent = 0;
8194bf88 828 I32 seen_dp = 0;
20f6aaab
AS
829 I32 digit = 0;
830 I32 old_digit = 0;
8194bf88
DM
831 I32 sig_digits = 0; /* noof significant digits seen so far */
832
833/* There is no point in processing more significant digits
834 * than the NV can hold. Note that NV_DIG is a lower-bound value,
835 * while we need an upper-bound value. We add 2 to account for this;
836 * since it will have been conservative on both the first and last digit.
837 * For example a 32-bit mantissa with an exponent of 4 would have
838 * exact values in the set
839 * 4
840 * 8
841 * ..
842 * 17179869172
843 * 17179869176
844 * 17179869180
845 *
846 * where for the purposes of calculating NV_DIG we would have to discount
847 * both the first and last digit, since neither can hold all values from
848 * 0..9; but for calculating the value we must examine those two digits.
849 */
850#define MAX_SIG_DIGITS (NV_DIG+2)
851
852/* the max number we can accumulate in a UV, and still safely do 10*N+9 */
853#define MAX_ACCUMULATE ( (UV) ((UV_MAX - 9)/10))
98994639 854
96a05aee
HS
855 /* leading whitespace */
856 while (isSPACE(*s))
857 ++s;
858
98994639
HS
859 /* sign */
860 switch (*s) {
861 case '-':
862 negative = 1;
863 /* fall through */
864 case '+':
865 ++s;
866 }
867
8194bf88
DM
868 /* we accumulate digits into an integer; when this becomes too
869 * large, we add the total to NV and start again */
98994639 870
8194bf88
DM
871 while (1) {
872 if (isDIGIT(*s)) {
873 seen_digit = 1;
20f6aaab 874 old_digit = digit;
8194bf88 875 digit = *s++ - '0';
20f6aaab
AS
876 if (seen_dp)
877 exp_adjust[1]++;
98994639 878
8194bf88
DM
879 /* don't start counting until we see the first significant
880 * digit, eg the 5 in 0.00005... */
881 if (!sig_digits && digit == 0)
882 continue;
883
884 if (++sig_digits > MAX_SIG_DIGITS) {
98994639 885 /* limits of precision reached */
20f6aaab
AS
886 if (digit > 5) {
887 ++accumulator[seen_dp];
888 } else if (digit == 5) {
889 if (old_digit % 2) { /* round to even - Allen */
890 ++accumulator[seen_dp];
891 }
892 }
893 if (seen_dp) {
894 exp_adjust[1]--;
895 } else {
896 exp_adjust[0]++;
897 }
8194bf88 898 /* skip remaining digits */
98994639 899 while (isDIGIT(*s)) {
98994639 900 ++s;
20f6aaab
AS
901 if (! seen_dp) {
902 exp_adjust[0]++;
903 }
98994639
HS
904 }
905 /* warn of loss of precision? */
98994639 906 }
8194bf88 907 else {
20f6aaab 908 if (accumulator[seen_dp] > MAX_ACCUMULATE) {
8194bf88 909 /* add accumulator to result and start again */
20f6aaab
AS
910 result[seen_dp] = S_mulexp10(result[seen_dp],
911 exp_acc[seen_dp])
912 + (NV)accumulator[seen_dp];
913 accumulator[seen_dp] = 0;
914 exp_acc[seen_dp] = 0;
98994639 915 }
20f6aaab
AS
916 accumulator[seen_dp] = accumulator[seen_dp] * 10 + digit;
917 ++exp_acc[seen_dp];
98994639 918 }
8194bf88
DM
919 }
920 else if (!seen_dp && GROK_NUMERIC_RADIX((const char **)&s, send)) {
921 seen_dp = 1;
20f6aaab
AS
922 if (sig_digits > MAX_SIG_DIGITS) {
923 ++s;
924 while (isDIGIT(*s)) {
925 ++s;
926 }
927 break;
928 }
8194bf88
DM
929 }
930 else {
931 break;
98994639
HS
932 }
933 }
934
20f6aaab
AS
935 result[0] = S_mulexp10(result[0], exp_acc[0]) + (NV)accumulator[0];
936 if (seen_dp) {
937 result[1] = S_mulexp10(result[1], exp_acc[1]) + (NV)accumulator[1];
938 }
98994639 939
8194bf88 940 if (seen_digit && (*s == 'e' || *s == 'E')) {
98994639
HS
941 bool expnegative = 0;
942
943 ++s;
944 switch (*s) {
945 case '-':
946 expnegative = 1;
947 /* fall through */
948 case '+':
949 ++s;
950 }
951 while (isDIGIT(*s))
952 exponent = exponent * 10 + (*s++ - '0');
953 if (expnegative)
954 exponent = -exponent;
955 }
956
20f6aaab
AS
957
958
98994639 959 /* now apply the exponent */
20f6aaab
AS
960
961 if (seen_dp) {
962 result[2] = S_mulexp10(result[0],exponent+exp_adjust[0])
963 + S_mulexp10(result[1],exponent-exp_adjust[1]);
964 } else {
965 result[2] = S_mulexp10(result[0],exponent+exp_adjust[0]);
966 }
98994639
HS
967
968 /* now apply the sign */
969 if (negative)
20f6aaab 970 result[2] = -result[2];
a36244b7 971#endif /* USE_PERL_ATOF */
20f6aaab 972 *value = result[2];
98994639
HS
973 return s;
974}
975
55954f19
JH
976#if ! defined(HAS_MODFL) && defined(HAS_AINTL) && defined(HAS_COPYSIGNL)
977long double
978Perl_my_modfl(long double x, long double *ip)
979{
980 *ip = aintl(x);
981 return (x == *ip ? copysignl(0.0L, x) : x - *ip);
982}
983#endif
984
985#if ! defined(HAS_FREXPL) && defined(HAS_ILOGBL) && defined(HAS_SCALBNL)
986long double
987Perl_my_frexpl(long double x, int *e) {
988 *e = x == 0.0L ? 0 : ilogbl(x) + 1;
989 return (scalbnl(x, -*e));
990}
991#endif