From a0e438cd3c28d2eaf4efa18243d5b6edafa14d88 Mon Sep 17 00:00:00 2001 From: mrkn Date: Sun, 2 Dec 2018 05:21:54 +0000 Subject: Import bigdecimal-1.4.0.pre-20181130a * https://github.com/ruby/bigdecimal/compare/74d25ef..v1.4.0.pre.20181130a git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@66124 b2dd03c8-39d4-4d8f-98ff-823fe69b080e --- ext/bigdecimal/bigdecimal.c | 426 ++++++++++++++++++------------ ext/bigdecimal/bigdecimal.def | 3 + ext/bigdecimal/bigdecimal.gemspec | 10 +- ext/bigdecimal/bigdecimal.h | 2 +- ext/bigdecimal/extconf.rb | 38 ++- ext/bigdecimal/lib/bigdecimal/jacobian.rb | 3 + ext/bigdecimal/lib/bigdecimal/util.rb | 30 ++- ext/bigdecimal/sample/linear.rb | 21 +- ext/bigdecimal/sample/nlsolve.rb | 10 +- ext/bigdecimal/util/extconf.rb | 40 +++ ext/bigdecimal/util/util.c | 9 + 11 files changed, 391 insertions(+), 201 deletions(-) create mode 100644 ext/bigdecimal/bigdecimal.def create mode 100644 ext/bigdecimal/util/extconf.rb create mode 100644 ext/bigdecimal/util/util.c (limited to 'ext/bigdecimal') diff --git a/ext/bigdecimal/bigdecimal.c b/ext/bigdecimal/bigdecimal.c index d1aa56c121..d2701cfb8e 100644 --- a/ext/bigdecimal/bigdecimal.c +++ b/ext/bigdecimal/bigdecimal.c @@ -135,24 +135,6 @@ rb_rational_den(VALUE rat) */ #define DoSomeOne(x,y,f) rb_num_coerce_bin(x,y,f) -/* - * Returns the BigDecimal version number. - */ -static VALUE -BigDecimal_version(VALUE self) -{ - /* - * 1.0.0: Ruby 1.8.0 - * 1.0.1: Ruby 1.8.1 - * 1.1.0: Ruby 1.9.3 - */ -#ifndef RUBY_BIGDECIMAL_VERSION -# error RUBY_BIGDECIMAL_VERSION is not defined -#endif - rb_warning("BigDecimal.ver is deprecated; use BigDecimal::VERSION instead."); - return rb_str_new2(RUBY_BIGDECIMAL_VERSION); -} - /* * VP routines used in BigDecimal part */ @@ -664,9 +646,10 @@ VP_EXPORT Real * VpNewRbClass(size_t mx, const char *str, VALUE klass) { VALUE obj = TypedData_Wrap_Struct(klass, &BigDecimal_data_type, 0); - Real *pv = VpAlloc(mx,str); + Real *pv = VpAlloc(mx, str, 1); RTYPEDDATA_DATA(obj) = pv; pv->obj = obj; + RB_OBJ_FREEZE(obj); return pv; } @@ -2165,15 +2148,10 @@ BigDecimal_exponent(VALUE self) return INT2NUM(e); } -/* Returns debugging information about the value as a string of comma-separated - * values in angle brackets with a leading #: +/* Returns a string representation of self. * * BigDecimal("1234.5678").inspect * #=> "0.12345678e4" - * - * The first part is the address, the second is the value as a string, and - * the final part ss(mm) is the current number of significant digits and the - * maximum number of significant digits, respectively. */ static VALUE BigDecimal_inspect(VALUE self) @@ -2335,7 +2313,7 @@ BigDecimal_power(int argc, VALUE*argv, VALUE self) n = NIL_P(prec) ? (ssize_t)(x->Prec*VpBaseFig()) : NUM2SSIZET(prec); if (VpIsNaN(x)) { - y = VpCreateRbObject(n, "0#"); + y = VpCreateRbObject(n, "0"); RB_GC_GUARD(y->obj); VpSetNaN(y); return ToValue(y); @@ -2459,7 +2437,7 @@ BigDecimal_power(int argc, VALUE*argv, VALUE self) } } else { - y = VpCreateRbObject(n, "0#"); + y = VpCreateRbObject(n, "0"); if (BIGDECIMAL_NEGATIVE_P(x)) { if (is_integer(vexp)) { if (is_even(vexp)) { @@ -2492,7 +2470,7 @@ BigDecimal_power(int argc, VALUE*argv, VALUE self) } else if (RTEST(rb_funcall(abs_value, '<', 1, INT2FIX(1)))) { if (is_negative(vexp)) { - y = VpCreateRbObject(n, "0#"); + y = VpCreateRbObject(n, "0"); if (is_even(vexp)) { VpSetInf(y, VpGetSign(x)); } @@ -2510,7 +2488,7 @@ BigDecimal_power(int argc, VALUE*argv, VALUE self) } else { if (is_positive(vexp)) { - y = VpCreateRbObject(n, "0#"); + y = VpCreateRbObject(n, "0"); if (is_even(vexp)) { VpSetInf(y, VpGetSign(x)); } @@ -2560,52 +2538,8 @@ BigDecimal_power_op(VALUE self, VALUE exp) return BigDecimal_power(1, &exp, self); } -static VALUE -BigDecimal_s_allocate(VALUE klass) -{ - return VpNewRbClass(0, NULL, klass)->obj; -} - static Real *BigDecimal_new(int argc, VALUE *argv); -/* call-seq: - * new(initial, digits) - * - * Create a new BigDecimal object. - * - * initial:: The initial value, as an Integer, a Float, a Rational, - * a BigDecimal, or a String. - * - * If it is a String, spaces are ignored and unrecognized characters - * terminate the value. - * - * digits:: The number of significant digits, as an Integer. If omitted or 0, - * the number of significant digits is determined from the initial - * value. - * - * The actual number of significant digits used in computation is usually - * larger than the specified number. - * - * ==== Exceptions - * - * TypeError:: If the +initial+ type is neither Integer, Float, - * Rational, nor BigDecimal, this exception is raised. - * - * TypeError:: If the +digits+ is not an Integer, this exception is raised. - * - * ArgumentError:: If +initial+ is a Float, and the +digits+ is larger than - * Float::DIG + 1, this exception is raised. - * - * ArgumentError:: If the +initial+ is a Float or Rational, and the +digits+ - * value is omitted, this exception is raised. - */ -static VALUE -BigDecimal_s_new(int argc, VALUE *argv, VALUE self) -{ - rb_warning("BigDecimal.new is deprecated; use Kernel.BigDecimal method instead."); - return rb_call_super(argc, argv); -} - static VALUE BigDecimal_initialize(int argc, VALUE *argv, VALUE self) { @@ -2700,10 +2634,40 @@ BigDecimal_new(int argc, VALUE *argv) break; } StringValueCStr(iniValue); - return VpAlloc(mf, RSTRING_PTR(iniValue)); + return VpAlloc(mf, RSTRING_PTR(iniValue), 1); } -/* See also BigDecimal.new */ +/* call-seq: + * BigDecimal(initial, digits) + * + * Create a new BigDecimal object. + * + * initial:: The initial value, as an Integer, a Float, a Rational, + * a BigDecimal, or a String. + * + * If it is a String, spaces are ignored and unrecognized characters + * terminate the value. + * + * digits:: The number of significant digits, as an Integer. If omitted or 0, + * the number of significant digits is determined from the initial + * value. + * + * The actual number of significant digits used in computation is usually + * larger than the specified number. + * + * ==== Exceptions + * + * TypeError:: If the +initial+ type is neither Integer, Float, + * Rational, nor BigDecimal, this exception is raised. + * + * TypeError:: If the +digits+ is not an Integer, this exception is raised. + * + * ArgumentError:: If +initial+ is a Float, and the +digits+ is larger than + * Float::DIG + 1, this exception is raised. + * + * ArgumentError:: If the +initial+ is a Float or Rational, and the +digits+ + * value is omitted, this exception is raised. + */ static VALUE BigDecimal_global_new(int argc, VALUE *argv, VALUE self) { @@ -2715,6 +2679,7 @@ BigDecimal_global_new(int argc, VALUE *argv, VALUE self) GUARD_OBJ(pv, BigDecimal_new(argc, argv)); if (ToValue(pv)) pv = VpCopy(NULL, pv); RTYPEDDATA_DATA(obj) = pv; + RB_OBJ_FREEZE(obj); return pv->obj = obj; } @@ -3138,6 +3103,21 @@ get_vp_value: return y; } +VALUE +rmpd_util_str_to_d(VALUE str) +{ + ENTER(1); + char const *c_str; + Real *pv; + VALUE obj; + + c_str = StringValueCStr(str); + GUARD_OBJ(pv, VpAlloc(0, c_str, 0)); + obj = TypedData_Wrap_Struct(rb_cBigDecimal, &BigDecimal_data_type, pv); + RB_OBJ_FREEZE(obj); + return obj; +} + /* Document-class: BigDecimal * BigDecimal provides arbitrary-precision floating point decimal arithmetic. * @@ -3277,18 +3257,17 @@ Init_bigdecimal(void) /* Class and method registration */ rb_cBigDecimal = rb_define_class("BigDecimal", rb_cNumeric); - rb_define_alloc_func(rb_cBigDecimal, BigDecimal_s_allocate); /* Global function */ rb_define_global_function("BigDecimal", BigDecimal_global_new, -1); /* Class methods */ - rb_define_singleton_method(rb_cBigDecimal, "new", BigDecimal_s_new, -1); + rb_undef_method(CLASS_OF(rb_cBigDecimal), "allocate"); + rb_undef_method(CLASS_OF(rb_cBigDecimal), "new"); rb_define_singleton_method(rb_cBigDecimal, "mode", BigDecimal_mode, -1); rb_define_singleton_method(rb_cBigDecimal, "limit", BigDecimal_limit, -1); rb_define_singleton_method(rb_cBigDecimal, "double_fig", BigDecimal_double_fig, 0); rb_define_singleton_method(rb_cBigDecimal, "_load", BigDecimal_load, 1); - rb_define_singleton_method(rb_cBigDecimal, "ver", BigDecimal_version, 0); rb_define_singleton_method(rb_cBigDecimal, "save_exception_mode", BigDecimal_save_exception_mode, 0); rb_define_singleton_method(rb_cBigDecimal, "save_rounding_mode", BigDecimal_save_rounding_mode, 0); @@ -3745,25 +3724,19 @@ One(void) VP_EXPORT double VpGetDoubleNaN(void) /* Returns the value of NaN */ { - static double fNaN = 0.0; - if (fNaN == 0.0) fNaN = Zero()/Zero(); - return fNaN; + return nan(""); } VP_EXPORT double VpGetDoublePosInf(void) /* Returns the value of +Infinity */ { - static double fInf = 0.0; - if (fInf == 0.0) fInf = One()/Zero(); - return fInf; + return HUGE_VAL; } VP_EXPORT double VpGetDoubleNegInf(void) /* Returns the value of -Infinity */ { - static double fInf = 0.0; - if (fInf == 0.0) fInf = -(One()/Zero()); - return fInf; + return -HUGE_VAL; } VP_EXPORT double @@ -3964,8 +3937,8 @@ VpInit(BDIGIT BaseVal) VpGetDoubleNegZero(); /* Allocates Vp constants. */ - VpConstOne = VpAlloc(1UL, "1"); - VpPt5 = VpAlloc(1UL, ".5"); + VpConstOne = VpAlloc(1UL, "1", 1); + VpPt5 = VpAlloc(1UL, ".5", 1); #ifdef BIGDECIMAL_DEBUG gnAlloc = 0; @@ -4029,6 +4002,52 @@ overflow: return VpException(VP_EXCEPTION_OVERFLOW, "Exponent overflow", 0); } +Real * +rmpd_parse_special_string(const char *str) +{ + static const struct { + const char *str; + size_t len; + int sign; + } table[] = { + { SZ_INF, sizeof(SZ_INF) - 1, VP_SIGN_POSITIVE_INFINITE }, + { SZ_PINF, sizeof(SZ_PINF) - 1, VP_SIGN_POSITIVE_INFINITE }, + { SZ_NINF, sizeof(SZ_NINF) - 1, VP_SIGN_NEGATIVE_INFINITE }, + { SZ_NaN, sizeof(SZ_NaN) - 1, VP_SIGN_NaN } + }; + static const size_t table_length = sizeof(table) / sizeof(table[0]); + size_t i; + + for (i = 0; i < table_length; ++i) { + const char *p; + if (strncmp(str, table[i].str, table[i].len) != 0) { + continue; + } + + p = str + table[i].len; + while (*p && ISSPACE(*p)) ++p; + if (*p == '\0') { + Real *vp = VpAllocReal(1); + vp->MaxPrec = 1; + switch (table[i].sign) { + default: + UNREACHABLE; break; + case VP_SIGN_POSITIVE_INFINITE: + VpSetPosInf(vp); + return vp; + case VP_SIGN_NEGATIVE_INFINITE: + VpSetNegInf(vp); + return vp; + case VP_SIGN_NaN: + VpSetNaN(vp); + return vp; + } + } + } + + return NULL; +} + /* * Allocates variable. * [Input] @@ -4043,10 +4062,10 @@ overflow: * NULL be returned if memory allocation is failed,or any error. */ VP_EXPORT Real * -VpAlloc(size_t mx, const char *szVal) +VpAlloc(size_t mx, const char *szVal, int strict_p) { const char *orig_szVal = szVal; - size_t i, ni, ipn, ipf, nf, ipe, ne, dot_seen, exp_seen, nalloc; + size_t i, j, ni, ipf, nf, ipe, ne, dot_seen, exp_seen, nalloc; char v, *psz; int sign=1; Real *vp = NULL; @@ -4057,7 +4076,10 @@ VpAlloc(size_t mx, const char *szVal) if (mx == 0) ++mx; if (szVal) { + /* Skipping leading spaces */ while (ISSPACE(*szVal)) szVal++; + + /* Processing the leading one `#` */ if (*szVal != '#') { if (mf) { mf = (mf + BASE_FIG - 1) / BASE_FIG + 2; /* Needs 1 more for div */ @@ -4071,6 +4093,7 @@ VpAlloc(size_t mx, const char *szVal) } } else { + return_zero: /* necessary to be able to store */ /* at least mx digits. */ /* szVal==NULL ==> allocate zero value. */ @@ -4081,105 +4104,166 @@ VpAlloc(size_t mx, const char *szVal) return vp; } - /* Skip all '_' after digit: 2006-6-30 */ - ni = 0; + /* Check on Inf & NaN */ + if ((vp = rmpd_parse_special_string(szVal)) != NULL) { + return vp; + } + + /* Scanning digits */ + + /* A buffer for keeping scanned digits */ buf = rb_str_tmp_new(strlen(szVal) + 1); psz = RSTRING_PTR(buf); - i = 0; - ipn = 0; - while ((psz[i] = szVal[ipn]) != 0) { - if (ISSPACE(psz[i])) { - psz[i] = 0; + + /* cursor: i for psz, and j for szVal */ + i = j = 0; + + /* Scanning: sign part */ + v = psz[i] = szVal[j]; + if ((v == '-') || (v == '+')) { + sign = -(v == '-'); + ++i; + ++j; + } + + /* Scanning: integer part */ + ni = 0; /* number of digits in the integer part */ + while ((v = psz[i] = szVal[j]) != '\0') { + if (!strict_p && ISSPACE(v)) { + v = psz[i] = '\0'; break; } - if (ISDIGIT(psz[i])) ++ni; - if (psz[i] == '_') { + if (v == '_') { if (ni > 0) { - ipn++; - continue; + v = szVal[j+1]; + if (v == '\0' || ISSPACE(v) || ISDIGIT(v)) { + ++j; + continue; + } + if (!strict_p) { + v = psz[i] = '\0'; + break; + } } - psz[i] = 0; + goto invalid_value; + } + if (!ISDIGIT(v)) { break; } + ++ni; ++i; - ++ipn; + ++j; } - szVal = psz; - /* Check on Inf & NaN */ - if (StrCmp(szVal, SZ_PINF) == 0 || StrCmp(szVal, SZ_INF) == 0 ) { - vp = VpAllocReal(1); - vp->MaxPrec = 1; /* set max precision */ - VpSetPosInf(vp); - return vp; - } - if (StrCmp(szVal, SZ_NINF) == 0) { - vp = VpAllocReal(1); - vp->MaxPrec = 1; /* set max precision */ - VpSetNegInf(vp); - return vp; - } - if (StrCmp(szVal, SZ_NaN) == 0) { - vp = VpAllocReal(1); - vp->MaxPrec = 1; /* set max precision */ - VpSetNaN(vp); - return vp; - } - - /* check on number szVal[] */ - ipn = i = 0; - if (szVal[i] == '-') { sign=-1; ++i; } - else if (szVal[i] == '+') ++i; - /* Skip digits */ - ni = 0; /* digits in mantissa */ - while ((v = szVal[i]) != 0) { - if (!ISDIGIT(v)) break; - ++i; - ++ni; - } - nf = 0; - ipf = 0; - ipe = 0; - ne = 0; + /* Scanning: fractional part */ + nf = 0; /* number of digits in the fractional part */ + ne = 0; /* number of digits in the exponential part */ + ipf = 0; /* index of the beginning of the fractional part */ + ipe = 0; /* index of the beginning of the exponential part */ dot_seen = 0; exp_seen = 0; - if (v) { - /* other than digit nor \0 */ - if (szVal[i] == '.') { /* xxx. */ + + if (v != '\0') { + /* Scanning fractional part */ + if ((psz[i] = szVal[j]) == '.') { dot_seen = 1; ++i; + ++j; ipf = i; - while ((v = szVal[i]) != 0) { /* get fraction part. */ + while ((v = psz[i] = szVal[j]) != '\0') { + if (!strict_p && ISSPACE(v)) { + v = psz[i] = '\0'; + break; + } + if (v == '_') { + if (nf > 0 && ISDIGIT(szVal[j+1])) { + ++j; + continue; + } + if (!strict_p) { + v = psz[i] = '\0'; + if (nf == 0) { + dot_seen = 0; + } + break; + } + goto invalid_value; + } if (!ISDIGIT(v)) break; ++i; + ++j; ++nf; } } - ipe = 0; /* Exponent */ - - switch (szVal[i]) { - case '\0': - break; - case 'e': case 'E': - case 'd': case 'D': - exp_seen = 1; - ++i; - ipe = i; - v = szVal[i]; - if ((v == '-') || (v == '+')) ++i; - while ((v=szVal[i]) != 0) { - if (!ISDIGIT(v)) break; + + /* Scanning exponential part */ + if (v != '\0') { + switch ((psz[i] = szVal[j])) { + case '\0': + break; + case 'e': case 'E': + case 'd': case 'D': + exp_seen = 1; ++i; - ++ne; - } - break; - default: - break; + ++j; + ipe = i; + v = psz[i] = szVal[j]; + if ((v == '-') || (v == '+')) { + ++i; + ++j; + } + while ((v = psz[i] = szVal[j]) != '\0') { + if (!strict_p && ISSPACE(v)) { + v = psz[i] = '\0'; + break; + } + if (v == '_') { + if (ne > 0 && ISDIGIT(szVal[j+1])) { + ++j; + continue; + } + if (!strict_p) { + v = psz[i] = '\0'; + if (ne == 0) { + exp_seen = 0; + } + break; + } + goto invalid_value; + } + if (!ISDIGIT(v)) break; + ++i; + ++j; + ++ne; + } + break; + default: + break; + } + } + + if (v != '\0') { + /* Scanning trailing spaces */ + while (ISSPACE(szVal[j])) ++j; + + /* Invalid character */ + if (szVal[j]) { + goto invalid_value; + } } } + + psz[i] = '\0'; + if (((ni == 0 || dot_seen) && nf == 0) || (exp_seen && ne == 0)) { - VALUE str = rb_str_new2(orig_szVal); - rb_raise(rb_eArgError, "invalid value for BigDecimal(): \"%"PRIsVALUE"\"", str); + VALUE str; + invalid_value: + if (!strict_p) { + goto return_zero; + } + + str = rb_str_new2(orig_szVal); + rb_raise(rb_eArgError, "invalid value for BigDecimal(): \"%"PRIsVALUE"\"", str); } nalloc = (ni + nf + BASE_FIG - 1) / BASE_FIG + 1; /* set effective allocation */ @@ -4191,7 +4275,7 @@ VpAlloc(size_t mx, const char *szVal) /* xmalloc() alway returns(or throw interruption) */ vp->MaxPrec = mx; /* set max precision */ VpSetZero(vp, sign); - VpCtoV(vp, &szVal[ipn], ni, &szVal[ipf], nf, &szVal[ipe], ne); + VpCtoV(vp, psz, ni, psz + ipf, nf, psz + ipe, ne); rb_str_resize(buf, 0); return vp; } @@ -4754,7 +4838,7 @@ VpMult(Real *c, Real *a, Real *b) if (MxIndC < MxIndAB) { /* The Max. prec. of c < Prec(a)+Prec(b) */ w = c; - c = VpAlloc((size_t)((MxIndAB + 1) * BASE_FIG), "#0"); + c = VpAlloc((size_t)((MxIndAB + 1) * BASE_FIG), "#0", 1); MxIndC = MxIndAB; } @@ -5922,8 +6006,8 @@ VpSqrt(Real *y, Real *x) if (x->MaxPrec > (size_t)n) n = (ssize_t)x->MaxPrec; /* allocate temporally variables */ - f = VpAlloc(y->MaxPrec * (BASE_FIG + 2), "#1"); - r = VpAlloc((n + n) * (BASE_FIG + 2), "#1"); + f = VpAlloc(y->MaxPrec * (BASE_FIG + 2), "#1", 1); + r = VpAlloc((n + n) * (BASE_FIG + 2), "#1", 1); nr = 0; y_prec = y->MaxPrec; @@ -6375,8 +6459,8 @@ VpPower(Real *y, Real *x, SIGNED_VALUE n) /* Allocate working variables */ - w1 = VpAlloc((y->MaxPrec + 2) * BASE_FIG, "#0"); - w2 = VpAlloc((w1->MaxPrec * 2 + 1) * BASE_FIG, "#0"); + w1 = VpAlloc((y->MaxPrec + 2) * BASE_FIG, "#0", 1); + w2 = VpAlloc((w1->MaxPrec * 2 + 1) * BASE_FIG, "#0", 1); /* calculation start */ VpAsgn(y, x, 1); diff --git a/ext/bigdecimal/bigdecimal.def b/ext/bigdecimal/bigdecimal.def new file mode 100644 index 0000000000..615bf72e20 --- /dev/null +++ b/ext/bigdecimal/bigdecimal.def @@ -0,0 +1,3 @@ +EXPORTS +rmpd_util_str_to_d +Init_bigdecimal diff --git a/ext/bigdecimal/bigdecimal.gemspec b/ext/bigdecimal/bigdecimal.gemspec index c8c90870ea..bc0cbac6b5 100644 --- a/ext/bigdecimal/bigdecimal.gemspec +++ b/ext/bigdecimal/bigdecimal.gemspec @@ -1,6 +1,6 @@ # coding: utf-8 -bigdecimal_version = '1.3.4' +bigdecimal_version = '1.4.0.pre.20181121a' Gem::Specification.new do |s| s.name = "bigdecimal" @@ -14,13 +14,15 @@ Gem::Specification.new do |s| s.license = "ruby" s.require_paths = %w[lib] - s.extensions = %w[ext/bigdecimal/extconf.rb] + s.extensions = %w[ext/bigdecimal/extconf.rb ext/bigdecimal/util/extconf.rb] s.files = %w[ bigdecimal.gemspec ext/bigdecimal/bigdecimal.c ext/bigdecimal/bigdecimal.h ext/bigdecimal/depend ext/bigdecimal/extconf.rb + ext/bigdecimal/util/extconf.rb + ext/bigdecimal/util/util.c lib/bigdecimal/jacobian.rb lib/bigdecimal/ludcmp.rb lib/bigdecimal/math.rb @@ -31,9 +33,11 @@ Gem::Specification.new do |s| sample/pi.rb ] + s.required_ruby_version = Gem::Requirement.new(">= 2.3.0".freeze) + s.add_development_dependency "rake", "~> 10.0" s.add_development_dependency "rake-compiler", ">= 0.9" s.add_development_dependency "rake-compiler-dock", ">= 0.6.1" - s.add_development_dependency "minitest", "~> 4.7.5" + s.add_development_dependency "minitest", "< 5.0.0" s.add_development_dependency "pry" end diff --git a/ext/bigdecimal/bigdecimal.h b/ext/bigdecimal/bigdecimal.h index 3297f18867..b7b0abe28c 100644 --- a/ext/bigdecimal/bigdecimal.h +++ b/ext/bigdecimal/bigdecimal.h @@ -308,7 +308,7 @@ VP_EXPORT size_t VpInit(BDIGIT BaseVal); VP_EXPORT void *VpMemAlloc(size_t mb); VP_EXPORT void *VpMemRealloc(void *ptr, size_t mb); VP_EXPORT void VpFree(Real *pv); -VP_EXPORT Real *VpAlloc(size_t mx, const char *szVal); +VP_EXPORT Real *VpAlloc(size_t mx, const char *szVal, int strict_p); VP_EXPORT size_t VpAsgn(Real *c, Real *a, int isw); VP_EXPORT size_t VpAddSub(Real *c,Real *a,Real *b,int operation); VP_EXPORT size_t VpMult(Real *c,Real *a,Real *b); diff --git a/ext/bigdecimal/extconf.rb b/ext/bigdecimal/extconf.rb index e565da891a..602c7d6162 100644 --- a/ext/bigdecimal/extconf.rb +++ b/ext/bigdecimal/extconf.rb @@ -1,6 +1,10 @@ # frozen_string_literal: false require 'mkmf' +def windows_platform? + /cygwin|mingw|mswin/ === RUBY_PLATFORM +end + gemspec_name = gemspec_path = nil unless ['', '../../'].any? {|dir| gemspec_name = "#{dir}bigdecimal.gemspec" @@ -17,8 +21,6 @@ bigdecimal_version = $defs << %Q[-DRUBY_BIGDECIMAL_VERSION=\\"#{bigdecimal_version}\\"] -alias __have_macro__ have_macro - have_func("labs", "stdlib.h") have_func("llabs", "stdlib.h") have_func("finite", "math.h") @@ -30,6 +32,38 @@ have_func("rb_rational_den", "ruby.h") have_func("rb_array_const_ptr", "ruby.h") have_func("rb_sym2str", "ruby.h") +if windows_platform? + library_base_name = "ruby-bigdecimal" + case RUBY_PLATFORM + when /cygwin|mingw/ + import_library_name = "libruby-bigdecimal.a" + when /mswin/ + import_library_name = "bigdecimal-$(arch).lib" + end +end + +checking_for(checking_message("Windows")) do + if windows_platform? + case RUBY_PLATFORM + when /cygwin|mingw/ + $DLDFLAGS << " $(srcdir)/bigdecimal.def" + $DLDFLAGS << " -Wl,--out-implib=$(TARGET_SO_DIR)#{import_library_name}" + when /mswin/ + $DLDFLAGS << " /DEF:$(srcdir)/bigdecimal.def" + end + $cleanfiles << import_library_name + true + else + false + end +end + create_makefile('bigdecimal') {|mf| mf << "\nall:\n\nextconf.h: $(srcdir)/#{gemspec_name}\n" + case RUBY_PLATFORM + when /mswin/ + mf << "\nall:\n\tdir $(TARGET_SO_DIR)" + else + mf << "\nall:\n\tls $(TARGET_SO_DIR)" + end } diff --git a/ext/bigdecimal/lib/bigdecimal/jacobian.rb b/ext/bigdecimal/lib/bigdecimal/jacobian.rb index 9cad06c09b..84c50248b7 100644 --- a/ext/bigdecimal/lib/bigdecimal/jacobian.rb +++ b/ext/bigdecimal/lib/bigdecimal/jacobian.rb @@ -21,6 +21,9 @@ # # fx is f.values(x). # + +require 'bigdecimal' + module Jacobian module_function diff --git a/ext/bigdecimal/lib/bigdecimal/util.rb b/ext/bigdecimal/lib/bigdecimal/util.rb index 911fa6fe3a..88f490cb45 100644 --- a/ext/bigdecimal/lib/bigdecimal/util.rb +++ b/ext/bigdecimal/lib/bigdecimal/util.rb @@ -5,6 +5,8 @@ # and provides BigDecimal#to_d and BigDecimal#to_digits. #++ +require 'bigdecimal' +require 'bigdecimal/util.so' class Integer < Numeric # call-seq: @@ -42,8 +44,8 @@ class Float < Numeric # # See also BigDecimal::new. # - def to_d(precision=nil) - BigDecimal(self, precision || Float::DIG) + def to_d(precision=Float::DIG) + BigDecimal(self, precision) end end @@ -64,13 +66,6 @@ class String # # See also BigDecimal::new. # - def to_d - begin - BigDecimal(self) - rescue ArgumentError - BigDecimal(0) - end - end end @@ -132,3 +127,20 @@ class Rational < Numeric BigDecimal(self, precision) end end + + +class NilClass + # call-seq: + # nil.to_d -> bigdecimal + # + # Returns nil represented as a BigDecimal. + # + # require 'bigdecimal' + # require 'bigdecimal/util' + # + # nil.to_d # => 0.0 + # + def to_d + BigDecimal(0) + end +end diff --git a/ext/bigdecimal/sample/linear.rb b/ext/bigdecimal/sample/linear.rb index 3b23269f8a..516c2473be 100644 --- a/ext/bigdecimal/sample/linear.rb +++ b/ext/bigdecimal/sample/linear.rb @@ -28,8 +28,8 @@ def rd_order(na) end na = ARGV.size -zero = BigDecimal.new("0.0") -one = BigDecimal.new("1.0") +zero = BigDecimal("0.0") +one = BigDecimal("1.0") while (n=rd_order(na))>0 a = [] @@ -37,27 +37,28 @@ while (n=rd_order(na))>0 b = [] if na <= 0 # Read data from console. - printf("\nEnter coefficient matrix element A[i,j]\n"); + printf("\nEnter coefficient matrix element A[i,j]\n") for i in 0...n do for j in 0...n do printf("A[%d,%d]? ",i,j); s = ARGF.gets - a << BigDecimal.new(s); - as << BigDecimal.new(s); + a << BigDecimal(s) + as << BigDecimal(s) end - printf("Contatant vector element b[%d] ? ",i); b << BigDecimal.new(ARGF.gets); + printf("Contatant vector element b[%d] ? ",i) + b << BigDecimal(ARGF.gets) end else # Read data from specified file. - printf("Coefficient matrix and constant vector.\n"); + printf("Coefficient matrix and constant vector.\n") for i in 0...n do s = ARGF.gets printf("%d) %s",i,s) s = s.split for j in 0...n do - a << BigDecimal.new(s[j]); - as << BigDecimal.new(s[j]); + a << BigDecimal(s[j]) + as << BigDecimal(s[j]) end - b << BigDecimal.new(s[n]); + b << BigDecimal(s[n]) end end x = lusolve(a,b,ludecomp(a,n,zero,one),zero) diff --git a/ext/bigdecimal/sample/nlsolve.rb b/ext/bigdecimal/sample/nlsolve.rb index b1dd08e0a3..c2227dac73 100644 --- a/ext/bigdecimal/sample/nlsolve.rb +++ b/ext/bigdecimal/sample/nlsolve.rb @@ -12,11 +12,11 @@ include Newton class Function # :nodoc: all def initialize() - @zero = BigDecimal.new("0.0") - @one = BigDecimal.new("1.0") - @two = BigDecimal.new("2.0") - @ten = BigDecimal.new("10.0") - @eps = BigDecimal.new("1.0e-16") + @zero = BigDecimal("0.0") + @one = BigDecimal("1.0") + @two = BigDecimal("2.0") + @ten = BigDecimal("10.0") + @eps = BigDecimal("1.0e-16") end def zero;@zero;end def one ;@one ;end diff --git a/ext/bigdecimal/util/extconf.rb b/ext/bigdecimal/util/extconf.rb new file mode 100644 index 0000000000..023b19ba7e --- /dev/null +++ b/ext/bigdecimal/util/extconf.rb @@ -0,0 +1,40 @@ +# frozen_string_literal: false +require 'mkmf' + +def windows_platform? + /cygwin|mingw|mswin/ === RUBY_PLATFORM +end + +if windows_platform? + library_base_name = "ruby-bigdecimal" + case RUBY_PLATFORM + when /cygwin|mingw/ + import_library_name = "libruby-bigdecimal.a" + when /mswin/ + import_library_name = "bigdecimal-$(arch).lib" + end +end + +checking_for(checking_message("Windows")) do + if windows_platform? + if defined?($extlist) + build_dir = "$(TARGET_SO_DIR)../" + else + base_dir = File.expand_path('../../../..', __FILE__) + build_dir = File.join(base_dir, "tmp", RUBY_PLATFORM, "bigdecimal", RUBY_VERSION) + end + case RUBY_PLATFORM + when /cygwin|mingw/ + $LDFLAGS << " -L#{build_dir} -L.. -L .." + $libs << " -l#{library_base_name}" + when /mswin/ + $DLDFLAGS << " /libpath:#{build_dir} /libpath:.." + $libs << " #{import_library_name}" + end + true + else + false + end +end + +create_makefile('bigdecimal/util') diff --git a/ext/bigdecimal/util/util.c b/ext/bigdecimal/util/util.c new file mode 100644 index 0000000000..8d38d87852 --- /dev/null +++ b/ext/bigdecimal/util/util.c @@ -0,0 +1,9 @@ +#include "ruby.h" + +RUBY_EXTERN VALUE rmpd_util_str_to_d(VALUE str); + +void +Init_util(void) +{ + rb_define_method(rb_cString, "to_d", rmpd_util_str_to_d, 0); +} -- cgit v1.2.3