X-Git-Url: https://perl5.git.perl.org/perl5.git/blobdiff_plain/8194bf885900463241631dc78b51b65378478827..e3c71926d33e006cc01a7516d80dda2a9888bb0e:/numeric.c diff --git a/numeric.c b/numeric.c index 969901e..2749e5e 100644 --- a/numeric.c +++ b/numeric.c @@ -1,6 +1,7 @@ /* numeric.c * - * Copyright (c) 2001-2002, Larry Wall + * Copyright (C) 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001, + * 2002, 2003, 2004, 2005, 2006, 2007, 2008 by Larry Wall and others * * You may distribute under the terms of either the GNU General Public * License or the Artistic License, as specified in the README file. @@ -8,12 +9,20 @@ */ /* - * "That only makes eleven (plus one mislaid) and not fourteen, unless - * wizards count differently to other people." + * "That only makes eleven (plus one mislaid) and not fourteen, + * unless wizards count differently to other people." --Beorn + * + * [p.115 of _The Hobbit_: "Queer Lodgings"] */ /* =head1 Numeric functions + +This file contains all the stuff needed by perl for manipulating numeric +values, including such things as replacements for the OS's atof() function + +=cut + */ #include "EXTERN.h" @@ -23,6 +32,7 @@ U32 Perl_cast_ulong(pTHX_ NV f) { + PERL_UNUSED_CONTEXT; if (f < 0.0) return f < I32_MIN ? (U32) I32_MIN : (U32)(I32) f; if (f < U32_MAX_P1) { @@ -41,6 +51,7 @@ Perl_cast_ulong(pTHX_ NV f) I32 Perl_cast_i32(pTHX_ NV f) { + PERL_UNUSED_CONTEXT; if (f < I32_MAX_P1) return f < I32_MIN ? I32_MIN : (I32) f; if (f < U32_MAX_P1) { @@ -59,6 +70,7 @@ Perl_cast_i32(pTHX_ NV f) IV Perl_cast_iv(pTHX_ NV f) { + PERL_UNUSED_CONTEXT; if (f < IV_MAX_P1) return f < IV_MIN ? IV_MIN : (IV) f; if (f < UV_MAX_P1) { @@ -78,6 +90,7 @@ Perl_cast_iv(pTHX_ NV f) UV Perl_cast_uv(pTHX_ NV f) { + PERL_UNUSED_CONTEXT; if (f < 0.0) return f < IV_MIN ? (UV) IV_MIN : (UV)(IV) f; if (f < UV_MAX_P1) { @@ -93,22 +106,6 @@ Perl_cast_uv(pTHX_ NV f) return f > 0 ? UV_MAX : 0 /* NaN */; } -#if defined(HUGE_VAL) || (defined(USE_LONG_DOUBLE) && defined(HUGE_VALL)) -/* - * This hack is to force load of "huge" support from libm.a - * So it is in perl for (say) POSIX to use. - * Needed for SunOS with Sun's 'acc' for example. - */ -NV -Perl_huge(void) -{ -# if defined(USE_LONG_DOUBLE) && defined(HUGE_VALL) - return HUGE_VALL; -# endif - return HUGE_VAL; -} -#endif - /* =for apidoc grok_bin @@ -117,52 +114,61 @@ converts a string representing a binary number to numeric form. On entry I and I<*len> give the string to scan, I<*flags> gives conversion flags, and I should be NULL or a pointer to an NV. The scan stops at the end of the string, or the first invalid character. -On return I<*len> is set to the length scanned string, and I<*flags> gives -output flags. +Unless C is set in I<*flags>, encountering an +invalid character will also trigger a warning. +On return I<*len> is set to the length of the scanned string, +and I<*flags> gives output flags. -If the value is <= UV_MAX it is returned as a UV, the output flags are clear, +If the value is <= C it is returned as a UV, the output flags are clear, and nothing is written to I<*result>. If the value is > UV_MAX C returns UV_MAX, sets C in the output flags, and writes the value to I<*result> (or the value is discarded if I is NULL). -The hex number may optionally be prefixed with "0b" or "b" unless +The binary number may optionally be prefixed with "0b" or "b" unless C is set in I<*flags> on entry. If C is set in I<*flags> then the binary number may use '_' characters to separate digits. =cut + +Not documented yet because experimental is C= 1) { - if (s[0] == 'b') { + if (s[0] == 'b' || s[0] == 'B') { s++; len--; } - else if (len >= 2 && s[0] == '0' && s[1] == 'b') { + else if (len >= 2 && s[0] == '0' && (s[1] == 'b' || s[1] == 'B')) { s+=2; len-=2; } } } - for (; len-- && *s; s++) { - char bit = *s; + for (; len-- && (bit = *s); s++) { if (bit == '0' || bit == '1') { /* Write it in this wonky order with a goto to attempt to get the compiler to make the common case integer-only loop pretty tight. @@ -174,9 +180,8 @@ Perl_grok_bin(pTHX_ char *start, STRLEN *len_p, I32 *flags, NV *result) { continue; } /* Bah. We're just overflowed. */ - if (ckWARN_d(WARN_OVERFLOW)) - Perl_warner(aTHX_ packWARN(WARN_OVERFLOW), - "Integer overflow in binary number"); + Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW), + "Integer overflow in binary number"); overflowed = TRUE; value_nv = (NV) value; } @@ -197,20 +202,20 @@ Perl_grok_bin(pTHX_ char *start, STRLEN *len_p, I32 *flags, NV *result) { ++s; goto redo; } - if (ckWARN(WARN_DIGIT)) - Perl_warner(aTHX_ packWARN(WARN_DIGIT), - "Illegal binary digit '%c' ignored", *s); + if (!(*flags & PERL_SCAN_SILENT_ILLDIGIT)) + Perl_ck_warner(aTHX_ packWARN(WARN_DIGIT), + "Illegal binary digit '%c' ignored", *s); break; } if ( ( overflowed && value_nv > 4294967295.0) #if UVSIZE > 4 - || (!overflowed && value > 0xffffffff ) + || (!overflowed && value > 0xffffffff + && ! (*flags & PERL_SCAN_SILENT_NON_PORTABLE)) #endif ) { - if (ckWARN(WARN_PORTABLE)) - Perl_warner(aTHX_ packWARN(WARN_PORTABLE), - "Binary number > 0b11111111111111111111111111111111 non-portable"); + Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE), + "Binary number > 0b11111111111111111111111111111111 non-portable"); } *len_p = s - start; if (!overflowed) { @@ -230,9 +235,11 @@ converts a string representing a hex number to numeric form. On entry I and I<*len> give the string to scan, I<*flags> gives conversion flags, and I should be NULL or a pointer to an NV. -The scan stops at the end of the string, or the first non-hex-digit character. -On return I<*len> is set to the length scanned string, and I<*flags> gives -output flags. +The scan stops at the end of the string, or the first invalid character. +Unless C is set in I<*flags>, encountering an +invalid character will also trigger a warning. +On return I<*len> is set to the length of the scanned string, +and I<*flags> gives output flags. If the value is <= UV_MAX it is returned as a UV, the output flags are clear, and nothing is written to I<*result>. If the value is > UV_MAX C @@ -246,30 +253,36 @@ C is set in I<*flags> then the hex number may use '_' characters to separate digits. =cut + +Not documented yet because experimental is C= 1) { - if (s[0] == 'x') { + if (s[0] == 'x' || s[0] == 'X') { s++; len--; } - else if (len >= 2 && s[0] == '0' && s[1] == 'x') { + else if (len >= 2 && s[0] == '0' && (s[1] == 'x' || s[1] == 'X')) { s+=2; len-=2; } @@ -277,7 +290,7 @@ Perl_grok_hex(pTHX_ char *start, STRLEN *len_p, I32 *flags, NV *result) { } for (; len-- && *s; s++) { - hexdigit = strchr((char *) PL_hexdigit, *s); + const char *hexdigit = strchr(PL_hexdigit, *s); if (hexdigit) { /* Write it in this wonky order with a goto to attempt to get the compiler to make the common case integer-only loop pretty tight. @@ -289,9 +302,8 @@ Perl_grok_hex(pTHX_ char *start, STRLEN *len_p, I32 *flags, NV *result) { continue; } /* Bah. We're just overflowed. */ - if (ckWARN_d(WARN_OVERFLOW)) - Perl_warner(aTHX_ packWARN(WARN_OVERFLOW), - "Integer overflow in hexadecimal number"); + Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW), + "Integer overflow in hexadecimal number"); overflowed = TRUE; value_nv = (NV) value; } @@ -306,26 +318,26 @@ Perl_grok_hex(pTHX_ char *start, STRLEN *len_p, I32 *flags, NV *result) { continue; } if (*s == '_' && len && allow_underscores && s[1] - && (hexdigit = strchr((char *) PL_hexdigit, s[1]))) + && (hexdigit = strchr(PL_hexdigit, s[1]))) { --len; ++s; goto redo; } - if (ckWARN(WARN_DIGIT)) - Perl_warner(aTHX_ packWARN(WARN_DIGIT), + if (!(*flags & PERL_SCAN_SILENT_ILLDIGIT)) + Perl_ck_warner(aTHX_ packWARN(WARN_DIGIT), "Illegal hexadecimal digit '%c' ignored", *s); break; } if ( ( overflowed && value_nv > 4294967295.0) #if UVSIZE > 4 - || (!overflowed && value > 0xffffffff ) + || (!overflowed && value > 0xffffffff + && ! (*flags & PERL_SCAN_SILENT_NON_PORTABLE)) #endif ) { - if (ckWARN(WARN_PORTABLE)) - Perl_warner(aTHX_ packWARN(WARN_PORTABLE), - "Hexadecimal number > 0xffffffff non-portable"); + Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE), + "Hexadecimal number > 0xffffffff non-portable"); } *len_p = s - start; if (!overflowed) { @@ -341,21 +353,45 @@ Perl_grok_hex(pTHX_ char *start, STRLEN *len_p, I32 *flags, NV *result) { /* =for apidoc grok_oct +converts a string representing an octal number to numeric form. + +On entry I and I<*len> give the string to scan, I<*flags> gives +conversion flags, and I should be NULL or a pointer to an NV. +The scan stops at the end of the string, or the first invalid character. +Unless C is set in I<*flags>, encountering an +8 or 9 will also trigger a warning. +On return I<*len> is set to the length of the scanned string, +and I<*flags> gives output flags. + +If the value is <= UV_MAX it is returned as a UV, the output flags are clear, +and nothing is written to I<*result>. If the value is > UV_MAX C +returns UV_MAX, sets C in the output flags, +and writes the value to I<*result> (or the value is discarded if I +is NULL). + +If C is set in I<*flags> then the octal +number may use '_' characters to separate digits. =cut + +Not documented yet because experimental is C 4294967295.0) #if UVSIZE > 4 - || (!overflowed && value > 0xffffffff ) + || (!overflowed && value > 0xffffffff + && ! (*flags & PERL_SCAN_SILENT_NON_PORTABLE)) #endif ) { - if (ckWARN(WARN_PORTABLE)) - Perl_warner(aTHX_ packWARN(WARN_PORTABLE), - "Octal number > 037777777777 non-portable"); + Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE), + "Octal number > 037777777777 non-portable"); } *len_p = s - start; if (!overflowed) { @@ -442,33 +477,39 @@ For backwards compatibility. Use C instead. */ NV -Perl_scan_bin(pTHX_ char *start, STRLEN len, STRLEN *retlen) +Perl_scan_bin(pTHX_ const char *start, STRLEN len, STRLEN *retlen) { NV rnv; I32 flags = *retlen ? PERL_SCAN_ALLOW_UNDERSCORES : 0; - UV ruv = grok_bin (start, &len, &flags, &rnv); + const UV ruv = grok_bin (start, &len, &flags, &rnv); + + PERL_ARGS_ASSERT_SCAN_BIN; *retlen = len; return (flags & PERL_SCAN_GREATER_THAN_UV_MAX) ? rnv : (NV)ruv; } NV -Perl_scan_oct(pTHX_ char *start, STRLEN len, STRLEN *retlen) +Perl_scan_oct(pTHX_ const char *start, STRLEN len, STRLEN *retlen) { NV rnv; I32 flags = *retlen ? PERL_SCAN_ALLOW_UNDERSCORES : 0; - UV ruv = grok_oct (start, &len, &flags, &rnv); + const UV ruv = grok_oct (start, &len, &flags, &rnv); + + PERL_ARGS_ASSERT_SCAN_OCT; *retlen = len; return (flags & PERL_SCAN_GREATER_THAN_UV_MAX) ? rnv : (NV)ruv; } NV -Perl_scan_hex(pTHX_ char *start, STRLEN len, STRLEN *retlen) +Perl_scan_hex(pTHX_ const char *start, STRLEN len, STRLEN *retlen) { NV rnv; I32 flags = *retlen ? PERL_SCAN_ALLOW_UNDERSCORES : 0; - UV ruv = grok_hex (start, &len, &flags, &rnv); + const UV ruv = grok_hex (start, &len, &flags, &rnv); + + PERL_ARGS_ASSERT_SCAN_HEX; *retlen = len; return (flags & PERL_SCAN_GREATER_THAN_UV_MAX) ? rnv : (NV)ruv; @@ -485,9 +526,13 @@ bool Perl_grok_numeric_radix(pTHX_ const char **sp, const char *send) { #ifdef USE_LOCALE_NUMERIC + dVAR; + + PERL_ARGS_ASSERT_GROK_NUMERIC_RADIX; + if (PL_numeric_radix_sv && IN_LOCALE) { STRLEN len; - char* radix = SvPV(PL_numeric_radix_sv, len); + const char * const radix = SvPV(PL_numeric_radix_sv, len); if (*sp + len <= send && memEQ(*sp, radix, len)) { *sp += len; return TRUE; @@ -496,6 +541,9 @@ Perl_grok_numeric_radix(pTHX_ const char **sp, const char *send) /* always try "." if numeric radix didn't match because * we may have data from different locales mixed */ #endif + + PERL_ARGS_ASSERT_GROK_NUMERIC_RADIX; + if (*sp < send && **sp == '.') { ++*sp; return TRUE; @@ -530,13 +578,15 @@ int Perl_grok_number(pTHX_ const char *pv, STRLEN len, UV *valuep) { const char *s = pv; - const char *send = pv + len; + const char * const send = pv + len; const UV max_div_10 = UV_MAX / 10; const char max_mod_10 = UV_MAX % 10; int numtype = 0; int sawinf = 0; int sawnan = 0; + PERL_ARGS_ASSERT_GROK_NUMBER; + while (s < send && isSPACE(*s)) s++; if (s == send) { @@ -717,7 +767,7 @@ Perl_grok_number(pTHX_ const char *pv, STRLEN len, UV *valuep) return 0; } -NV +STATIC NV S_mulexp10(NV value, I32 exponent) { NV result = 1.0; @@ -727,6 +777,8 @@ S_mulexp10(NV value, I32 exponent) if (exponent == 0) return value; + if (value == 0) + return (NV)0; /* On OpenVMS VAX we by default use the D_FLOAT double format, * and that format does not have *easy* capabilities [1] for @@ -750,7 +802,7 @@ S_mulexp10(NV value, I32 exponent) #if ((defined(VMS) && !defined(__IEEE_FP)) || defined(_UNICOS)) && defined(NV_MAX_10_EXP) STMT_START { - NV exp_v = log10(value); + const NV exp_v = log10(value); if (exponent >= NV_MAX_10_EXP || exponent + exp_v >= NV_MAX_10_EXP) return NV_MAX; if (exponent < 0) { @@ -788,6 +840,10 @@ Perl_my_atof(pTHX_ const char* s) { NV x = 0.0; #ifdef USE_LOCALE_NUMERIC + dVAR; + + PERL_ARGS_ASSERT_MY_ATOF; + if (PL_numeric_local && IN_LOCALE) { NV y; @@ -811,20 +867,24 @@ Perl_my_atof(pTHX_ const char* s) char* Perl_my_atof2(pTHX_ const char* orig, NV* value) { - NV result = 0.0; - char* s = (char*)orig; + NV result[3] = {0.0, 0.0, 0.0}; + const char* s = orig; #ifdef USE_PERL_ATOF - UV accumulator = 0; + UV accumulator[2] = {0,0}; /* before/after dp */ bool negative = 0; - char* send = s + strlen(orig) - 1; + const char* send = s + strlen(orig) - 1; bool seen_digit = 0; - I32 exp_adjust = 0; - I32 exp_acc = 0; /* the current exponent adjust for the accumulator */ + I32 exp_adjust[2] = {0,0}; + I32 exp_acc[2] = {-1, -1}; + /* the current exponent adjust for the accumulators */ I32 exponent = 0; I32 seen_dp = 0; - I32 digit; + I32 digit = 0; + I32 old_digit = 0; I32 sig_digits = 0; /* noof significant digits seen so far */ + PERL_ARGS_ASSERT_MY_ATOF2; + /* There is no point in processing more significant digits * than the NV can hold. Note that NV_DIG is a lower-bound value, * while we need an upper-bound value. We add 2 to account for this; @@ -842,7 +902,14 @@ Perl_my_atof2(pTHX_ const char* orig, NV* value) * both the first and last digit, since neither can hold all values from * 0..9; but for calculating the value we must examine those two digits. */ -#define MAX_SIG_DIGITS (NV_DIG+2) +#ifdef MAX_SIG_DIG_PLUS + /* It is not necessarily the case that adding 2 to NV_DIG gets all the + possible digits in a NV, especially if NVs are not IEEE compliant + (e.g., long doubles on IRIX) - Allen */ +# define MAX_SIG_DIGITS (NV_DIG+MAX_SIG_DIG_PLUS) +#else +# define MAX_SIG_DIGITS (NV_DIG+2) +#endif /* the max number we can accumulate in a UV, and still safely do 10*N+9 */ #define MAX_ACCUMULATE ( (UV) ((UV_MAX - 9)/10)) @@ -860,14 +927,31 @@ Perl_my_atof2(pTHX_ const char* orig, NV* value) ++s; } + /* punt to strtod for NaN/Inf; if no support for it there, tough luck */ + +#ifdef HAS_STRTOD + if (*s == 'n' || *s == 'N' || *s == 'i' || *s == 'I') { + const char *p = negative ? s - 1 : s; + char *endp; + NV rslt; + rslt = strtod(p, &endp); + if (endp != p) { + *value = rslt; + return (char *)endp; + } + } +#endif + /* we accumulate digits into an integer; when this becomes too * large, we add the total to NV and start again */ while (1) { if (isDIGIT(*s)) { seen_digit = 1; + old_digit = digit; digit = *s++ - '0'; - exp_adjust -= seen_dp; + if (seen_dp) + exp_adjust[1]++; /* don't start counting until we see the first significant * digit, eg the 5 in 0.00005... */ @@ -876,36 +960,58 @@ Perl_my_atof2(pTHX_ const char* orig, NV* value) if (++sig_digits > MAX_SIG_DIGITS) { /* limits of precision reached */ - if (digit >= 5) - ++accumulator; - ++exp_adjust; + if (digit > 5) { + ++accumulator[seen_dp]; + } else if (digit == 5) { + if (old_digit % 2) { /* round to even - Allen */ + ++accumulator[seen_dp]; + } + } + if (seen_dp) { + exp_adjust[1]--; + } else { + exp_adjust[0]++; + } /* skip remaining digits */ while (isDIGIT(*s)) { ++s; - exp_adjust += 1 - seen_dp; + if (! seen_dp) { + exp_adjust[0]++; + } } /* warn of loss of precision? */ } else { - if (accumulator > MAX_ACCUMULATE) { + if (accumulator[seen_dp] > MAX_ACCUMULATE) { /* add accumulator to result and start again */ - result = S_mulexp10(result, exp_acc) + (NV)accumulator; - accumulator = 0; - exp_acc = 0; + result[seen_dp] = S_mulexp10(result[seen_dp], + exp_acc[seen_dp]) + + (NV)accumulator[seen_dp]; + accumulator[seen_dp] = 0; + exp_acc[seen_dp] = 0; } - accumulator = accumulator * 10 + digit; - ++exp_acc; + accumulator[seen_dp] = accumulator[seen_dp] * 10 + digit; + ++exp_acc[seen_dp]; } } - else if (!seen_dp && GROK_NUMERIC_RADIX((const char **)&s, send)) { + else if (!seen_dp && GROK_NUMERIC_RADIX(&s, send)) { seen_dp = 1; + if (sig_digits > MAX_SIG_DIGITS) { + do { + ++s; + } while (isDIGIT(*s)); + break; + } } else { break; } } - result = S_mulexp10(result, exp_acc) + (NV)accumulator; + result[0] = S_mulexp10(result[0], exp_acc[0]) + (NV)accumulator[0]; + if (seen_dp) { + result[1] = S_mulexp10(result[1], exp_acc[1]) + (NV)accumulator[1]; + } if (seen_digit && (*s == 'e' || *s == 'E')) { bool expnegative = 0; @@ -924,15 +1030,80 @@ Perl_my_atof2(pTHX_ const char* orig, NV* value) exponent = -exponent; } + + /* now apply the exponent */ - exponent += exp_adjust; - result = S_mulexp10(result, exponent); + + if (seen_dp) { + result[2] = S_mulexp10(result[0],exponent+exp_adjust[0]) + + S_mulexp10(result[1],exponent-exp_adjust[1]); + } else { + result[2] = S_mulexp10(result[0],exponent+exp_adjust[0]); + } /* now apply the sign */ if (negative) - result = -result; + result[2] = -result[2]; #endif /* USE_PERL_ATOF */ - *value = result; - return s; + *value = result[2]; + return (char *)s; } +#if ! defined(HAS_MODFL) && defined(HAS_AINTL) && defined(HAS_COPYSIGNL) +long double +Perl_my_modfl(long double x, long double *ip) +{ + *ip = aintl(x); + return (x == *ip ? copysignl(0.0L, x) : x - *ip); +} +#endif + +#if ! defined(HAS_FREXPL) && defined(HAS_ILOGBL) && defined(HAS_SCALBNL) +long double +Perl_my_frexpl(long double x, int *e) { + *e = x == 0.0L ? 0 : ilogbl(x) + 1; + return (scalbnl(x, -*e)); +} +#endif + +/* +=for apidoc Perl_signbit + +Return a non-zero integer if the sign bit on an NV is set, and 0 if +it is not. + +If Configure detects this system has a signbit() that will work with +our NVs, then we just use it via the #define in perl.h. Otherwise, +fall back on this implementation. As a first pass, this gets everything +right except -0.0. Alas, catching -0.0 is the main use for this function, +so this is not too helpful yet. Still, at least we have the scaffolding +in place to support other systems, should that prove useful. + + +Configure notes: This function is called 'Perl_signbit' instead of a +plain 'signbit' because it is easy to imagine a system having a signbit() +function or macro that doesn't happen to work with our particular choice +of NVs. We shouldn't just re-#define signbit as Perl_signbit and expect +the standard system headers to be happy. Also, this is a no-context +function (no pTHX_) because Perl_signbit() is usually re-#defined in +perl.h as a simple macro call to the system's signbit(). +Users should just always call Perl_signbit(). + +=cut +*/ +#if !defined(HAS_SIGNBIT) +int +Perl_signbit(NV x) { + return (x < 0.0) ? 1 : 0; +} +#endif + +/* + * Local variables: + * c-indentation-style: bsd + * c-basic-offset: 4 + * indent-tabs-mode: t + * End: + * + * ex: set ts=8 sts=4 sw=4 noet: + */