gecko/js/src/jsnum.cpp

1556 lines
41 KiB
C++
Raw Normal View History

/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 4 -*-
*
2012-05-21 04:12:37 -07:00
* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
/*
* JS number type and wrapper class.
*/
#include "mozilla/FloatingPoint.h"
#include "mozilla/RangedPtr.h"
#include "double-conversion.h"
// Avoid warnings about ASSERT being defined by the assembler as well.
#undef ASSERT
#ifdef XP_OS2
#define _PC_53 PC_53
#define _MCW_EM MCW_EM
#define _MCW_PC MCW_PC
#endif
#include <locale.h>
#include <limits.h>
#include <math.h>
#include <stdlib.h>
#include <string.h>
#include "jstypes.h"
#include "jsutil.h"
#include "jsapi.h"
#include "jsatom.h"
#include "jscntxt.h"
#include "jsversion.h"
#include "jsdtoa.h"
#include "jsgc.h"
#include "jsinterp.h"
#include "jsnum.h"
#include "jsobj.h"
#include "jsopcode.h"
#include "jsprf.h"
#include "jsscope.h"
#include "jsstr.h"
#include "jslibmath.h"
#include "vm/GlobalObject.h"
#include "vm/NumericConversions.h"
#include "vm/StringBuffer.h"
#include "jsatominlines.h"
#include "jsinferinlines.h"
#include "jsnuminlines.h"
#include "jsobjinlines.h"
#include "vm/NumberObject-inl.h"
#include "vm/String-inl.h"
using namespace js;
using namespace js::types;
/*
* If we're accumulating a decimal number and the number is >= 2^53, then the
* fast result from the loop in GetPrefixInteger may be inaccurate. Call
* js_strtod_harder to get the correct answer.
*/
static bool
ComputeAccurateDecimalInteger(JSContext *cx, const jschar *start, const jschar *end, double *dp)
{
size_t length = end - start;
char *cstr = cx->pod_malloc<char>(length + 1);
if (!cstr)
return false;
for (size_t i = 0; i < length; i++) {
char c = char(start[i]);
JS_ASSERT(('0' <= c && c <= '9') || ('a' <= c && c <= 'z') || ('A' <= c && c <= 'Z'));
cstr[i] = c;
}
cstr[length] = 0;
char *estr;
int err = 0;
*dp = js_strtod_harder(cx->runtime->dtoaState, cstr, &estr, &err);
if (err == JS_DTOA_ENOMEM) {
JS_ReportOutOfMemory(cx);
js_free(cstr);
return false;
}
if (err == JS_DTOA_ERANGE && *dp == HUGE_VAL)
*dp = js_PositiveInfinity;
js_free(cstr);
return true;
}
class BinaryDigitReader
{
const int base; /* Base of number; must be a power of 2 */
int digit; /* Current digit value in radix given by base */
int digitMask; /* Mask to extract the next bit from digit */
const jschar *start; /* Pointer to the remaining digits */
const jschar *end; /* Pointer to first non-digit */
public:
BinaryDigitReader(int base, const jschar *start, const jschar *end)
: base(base), digit(0), digitMask(0), start(start), end(end)
{
}
/* Return the next binary digit from the number, or -1 if done. */
int nextDigit() {
if (digitMask == 0) {
if (start == end)
return -1;
int c = *start++;
JS_ASSERT(('0' <= c && c <= '9') || ('a' <= c && c <= 'z') || ('A' <= c && c <= 'Z'));
if ('0' <= c && c <= '9')
digit = c - '0';
else if ('a' <= c && c <= 'z')
digit = c - 'a' + 10;
else
digit = c - 'A' + 10;
digitMask = base >> 1;
}
int bit = (digit & digitMask) != 0;
digitMask >>= 1;
return bit;
}
};
/*
* The fast result might also have been inaccurate for power-of-two bases. This
* happens if the addition in value * 2 + digit causes a round-down to an even
* least significant mantissa bit when the first dropped bit is a one. If any
* of the following digits in the number (which haven't been added in yet) are
* nonzero, then the correct action would have been to round up instead of
* down. An example occurs when reading the number 0x1000000000000081, which
* rounds to 0x1000000000000000 instead of 0x1000000000000100.
*/
static double
ComputeAccurateBinaryBaseInteger(JSContext *cx, const jschar *start, const jschar *end, int base)
{
BinaryDigitReader bdr(base, start, end);
/* Skip leading zeroes. */
int bit;
do {
bit = bdr.nextDigit();
} while (bit == 0);
JS_ASSERT(bit == 1); // guaranteed by GetPrefixInteger
/* Gather the 53 significant bits (including the leading 1). */
double value = 1.0;
for (int j = 52; j > 0; j--) {
bit = bdr.nextDigit();
if (bit < 0)
return value;
value = value * 2 + bit;
}
/* bit2 is the 54th bit (the first dropped from the mantissa). */
int bit2 = bdr.nextDigit();
if (bit2 >= 0) {
double factor = 2.0;
int sticky = 0; /* sticky is 1 if any bit beyond the 54th is 1 */
int bit3;
while ((bit3 = bdr.nextDigit()) >= 0) {
sticky |= bit3;
factor *= 2;
}
value += bit2 & (bit | sticky);
value *= factor;
}
return value;
}
namespace js {
bool
GetPrefixInteger(JSContext *cx, const jschar *start, const jschar *end, int base,
const jschar **endp, double *dp)
{
JS_ASSERT(start <= end);
JS_ASSERT(2 <= base && base <= 36);
const jschar *s = start;
double d = 0.0;
for (; s < end; s++) {
int digit;
jschar c = *s;
if ('0' <= c && c <= '9')
digit = c - '0';
else if ('a' <= c && c <= 'z')
digit = c - 'a' + 10;
else if ('A' <= c && c <= 'Z')
digit = c - 'A' + 10;
else
break;
if (digit >= base)
break;
d = d * base + digit;
}
*endp = s;
*dp = d;
/* If we haven't reached the limit of integer precision, we're done. */
if (d < DOUBLE_INTEGRAL_PRECISION_LIMIT)
return true;
/*
* Otherwise compute the correct integer from the prefix of valid digits
* if we're computing for base ten or a power of two. Don't worry about
* other bases; see 15.1.2.2 step 13.
*/
if (base == 10)
return ComputeAccurateDecimalInteger(cx, start, s, dp);
if ((base & (base - 1)) == 0)
*dp = ComputeAccurateBinaryBaseInteger(cx, start, s, base);
return true;
}
} // namespace js
static JSBool
num_isNaN(JSContext *cx, unsigned argc, Value *vp)
{
if (argc == 0) {
2010-07-14 23:19:36 -07:00
vp->setBoolean(true);
return JS_TRUE;
}
double x;
if (!ToNumber(cx, vp[2], &x))
return false;
vp->setBoolean(MOZ_DOUBLE_IS_NaN(x));
return JS_TRUE;
}
static JSBool
num_isFinite(JSContext *cx, unsigned argc, Value *vp)
{
if (argc == 0) {
2010-07-14 23:19:36 -07:00
vp->setBoolean(false);
return JS_TRUE;
}
double x;
if (!ToNumber(cx, vp[2], &x))
return JS_FALSE;
vp->setBoolean(MOZ_DOUBLE_IS_FINITE(x));
return JS_TRUE;
}
static JSBool
num_parseFloat(JSContext *cx, unsigned argc, Value *vp)
{
JSString *str;
double d;
const jschar *bp, *end, *ep;
if (argc == 0) {
2010-07-14 23:19:36 -07:00
vp->setDouble(js_NaN);
return JS_TRUE;
}
str = ToString(cx, vp[2]);
if (!str)
return JS_FALSE;
bp = str->getChars(cx);
if (!bp)
return JS_FALSE;
end = bp + str->length();
if (!js_strtod(cx, bp, end, &ep, &d))
return JS_FALSE;
if (ep == bp) {
2010-07-14 23:19:36 -07:00
vp->setDouble(js_NaN);
return JS_TRUE;
}
2010-07-14 23:19:36 -07:00
vp->setNumber(d);
return JS_TRUE;
}
static bool
ParseIntStringHelper(JSContext *cx, const jschar *ws, const jschar *end, int maybeRadix,
bool stripPrefix, double *dp)
{
JS_ASSERT(maybeRadix == 0 || (2 <= maybeRadix && maybeRadix <= 36));
JS_ASSERT(ws <= end);
const jschar *s = SkipSpace(ws, end);
JS_ASSERT(ws <= s);
JS_ASSERT(s <= end);
/* 15.1.2.2 steps 3-4. */
bool negative = (s != end && s[0] == '-');
/* 15.1.2.2 step 5. */
if (s != end && (s[0] == '-' || s[0] == '+'))
s++;
/* 15.1.2.2 step 9. */
int radix = maybeRadix;
if (radix == 0) {
if (end - s >= 2 && s[0] == '0' && (s[1] != 'x' && s[1] != 'X')) {
/*
* Non-standard: ES5 requires that parseInt interpret leading-zero
* strings not starting with "0x" or "0X" as decimal (absent an
* explicitly specified non-zero radix), but we continue to
* interpret such strings as octal, as per ES3 and web practice.
*/
radix = 8;
} else {
radix = 10;
}
}
/* 15.1.2.2 step 10. */
if (stripPrefix) {
if (end - s >= 2 && s[0] == '0' && (s[1] == 'x' || s[1] == 'X')) {
s += 2;
radix = 16;
}
}
/* 15.1.2.2 steps 11-14. */
const jschar *actualEnd;
if (!GetPrefixInteger(cx, s, end, radix, &actualEnd, dp))
return false;
if (s == actualEnd)
*dp = js_NaN;
else if (negative)
*dp = -*dp;
return true;
}
/* See ECMA 15.1.2.2. */
JSBool
js::num_parseInt(JSContext *cx, unsigned argc, Value *vp)
{
CallArgs args = CallArgsFromVp(argc, vp);
/* Fast paths and exceptional cases. */
if (args.length() == 0) {
args.rval().setDouble(js_NaN);
return true;
}
if (args.length() == 1 ||
(args[1].isInt32() && (args[1].toInt32() == 0 || args[1].toInt32() == 10))) {
if (args[0].isInt32()) {
args.rval().set(args[0]);
return true;
}
/*
* Step 1 is |inputString = ToString(string)|. When string >=
* 1e21, ToString(string) is in the form "NeM". 'e' marks the end of
* the word, which would mean the result of parseInt(string) should be |N|.
*
* To preserve this behaviour, we can't use the fast-path when string >
* 1e21, or else the result would be |NeM|.
*
* The same goes for values smaller than 1.0e-6, because the string would be in
* the form of "Ne-M".
*/
if (args[0].isDouble()) {
double d = args[0].toDouble();
if (1.0e-6 < d && d < 1.0e21) {
args.rval().setNumber(floor(d));
return true;
}
if (-1.0e21 < d && d < -1.0e-6) {
args.rval().setNumber(-floor(-d));
return true;
}
if (d == 0.0) {
args.rval().setInt32(0);
return true;
}
}
}
/* Step 1. */
RootedString inputString(cx, ToString(cx, args[0]));
if (!inputString)
return false;
args[0].setString(inputString);
/* 15.1.2.2 steps 6-8. */
bool stripPrefix = true;
int32_t radix = 0;
if (args.length() > 1) {
if (!ToInt32(cx, args[1], &radix))
return false;
if (radix != 0) {
if (radix < 2 || radix > 36) {
args.rval().setDouble(js_NaN);
return true;
}
if (radix != 16)
stripPrefix = false;
}
}
/* Steps 2-5, 9-14. */
const jschar *ws = inputString->getChars(cx);
if (!ws)
return false;
const jschar *end = ws + inputString->length();
double number;
if (!ParseIntStringHelper(cx, ws, end, radix, stripPrefix, &number))
return false;
/* Step 15. */
args.rval().setNumber(number);
return true;
}
static JSFunctionSpec number_functions[] = {
JS_FN(js_isNaN_str, num_isNaN, 1,0),
JS_FN(js_isFinite_str, num_isFinite, 1,0),
JS_FN(js_parseFloat_str, num_parseFloat, 1,0),
JS_FN(js_parseInt_str, num_parseInt, 2,0),
JS_FS_END
};
Class js::NumberClass = {
js_Number_str,
JSCLASS_HAS_RESERVED_SLOTS(1) | JSCLASS_HAS_CACHED_PROTO(JSProto_Number),
JS_PropertyStub, /* addProperty */
JS_PropertyStub, /* delProperty */
JS_PropertyStub, /* getProperty */
JS_StrictPropertyStub, /* setProperty */
JS_EnumerateStub,
JS_ResolveStub,
JS_ConvertStub
};
static JSBool
Number(JSContext *cx, unsigned argc, Value *vp)
{
/* Sample JS_CALLEE before clobbering. */
bool isConstructing = IsConstructing(vp);
if (argc > 0) {
if (!ToNumber(cx, &vp[2]))
return false;
vp[0] = vp[2];
} else {
vp[0].setInt32(0);
}
if (!isConstructing)
return true;
JSObject *obj = NumberObject::create(cx, vp[0].toNumber());
if (!obj)
return false;
vp->setObject(*obj);
return true;
}
JS_ALWAYS_INLINE bool
IsNumber(const Value &v)
{
return v.isNumber() || (v.isObject() && v.toObject().hasClass(&NumberClass));
}
inline double
Extract(const Value &v)
{
if (v.isNumber())
return v.toNumber();
return v.toObject().asNumber().unbox();
}
#if JS_HAS_TOSOURCE
JS_ALWAYS_INLINE bool
num_toSource_impl(JSContext *cx, CallArgs args)
{
double d = Extract(args.thisv());
StringBuffer sb(cx);
if (!sb.append("(new Number(") ||
!NumberValueToStringBuffer(cx, NumberValue(d), sb) ||
!sb.append("))"))
{
return false;
}
JSString *str = sb.finishString();
if (!str)
return false;
args.rval().setString(str);
return true;
}
static JSBool
num_toSource(JSContext *cx, unsigned argc, Value *vp)
{
CallArgs args = CallArgsFromVp(argc, vp);
return CallNonGenericMethod<IsNumber, num_toSource_impl>(cx, args);
}
#endif
ToCStringBuf::ToCStringBuf() :dbuf(NULL)
{
JS_STATIC_ASSERT(sbufSize >= DTOSTR_STANDARD_BUFFER_SIZE);
}
ToCStringBuf::~ToCStringBuf()
{
if (dbuf)
js_free(dbuf);
}
JSFlatString *
js::Int32ToString(JSContext *cx, int32_t si)
{
uint32_t ui;
if (si >= 0) {
if (StaticStrings::hasInt(si))
return cx->runtime->staticStrings.getInt(si);
ui = si;
} else {
ui = uint32_t(-si);
JS_ASSERT_IF(si == INT32_MIN, ui == uint32_t(INT32_MAX) + 1);
}
JSCompartment *c = cx->compartment;
if (JSFlatString *str = c->dtoaCache.lookup(10, si))
return str;
JSShortString *str = js_NewGCShortString(cx);
if (!str)
return NULL;
jschar buffer[JSShortString::MAX_SHORT_LENGTH + 1];
RangedPtr<jschar> end(buffer + JSShortString::MAX_SHORT_LENGTH,
buffer, JSShortString::MAX_SHORT_LENGTH + 1);
*end = '\0';
RangedPtr<jschar> start = BackfillIndexInCharBuffer(ui, end);
if (si < 0)
*--start = '-';
jschar *dst = str->init(end - start);
PodCopy(dst, start.get(), end - start + 1);
c->dtoaCache.cache(10, si, str);
return str;
}
/* Returns a non-NULL pointer to inside cbuf. */
static char *
IntToCString(ToCStringBuf *cbuf, int i, int base = 10)
{
unsigned u = (i < 0) ? -i : i;
RangedPtr<char> cp(cbuf->sbuf + cbuf->sbufSize - 1, cbuf->sbuf, cbuf->sbufSize);
*cp = '\0';
/* Build the string from behind. */
switch (base) {
case 10:
cp = BackfillIndexInCharBuffer(u, cp);
break;
case 16:
do {
unsigned newu = u / 16;
*--cp = "0123456789abcdef"[u - newu * 16];
u = newu;
} while (u != 0);
break;
default:
JS_ASSERT(base >= 2 && base <= 36);
do {
unsigned newu = u / base;
*--cp = "0123456789abcdefghijklmnopqrstuvwxyz"[u - newu * base];
u = newu;
} while (u != 0);
break;
}
if (i < 0)
*--cp = '-';
return cp.get();
}
static JSString * JS_FASTCALL
js_NumberToStringWithBase(JSContext *cx, double d, int base);
JS_ALWAYS_INLINE bool
num_toString_impl(JSContext *cx, CallArgs args)
{
JS_ASSERT(IsNumber(args.thisv()));
double d = Extract(args.thisv());
int32_t base = 10;
if (args.hasDefined(0)) {
double d2;
if (!ToInteger(cx, args[0], &d2))
return false;
if (d2 < 2 || d2 > 36) {
JS_ReportErrorNumber(cx, js_GetErrorMessage, NULL, JSMSG_BAD_RADIX);
return false;
}
base = int32_t(d2);
}
2010-07-14 23:19:36 -07:00
JSString *str = js_NumberToStringWithBase(cx, d, base);
if (!str) {
JS_ReportOutOfMemory(cx);
return false;
}
args.rval().setString(str);
return true;
}
static JSBool
num_toString(JSContext *cx, unsigned argc, Value *vp)
{
CallArgs args = CallArgsFromVp(argc, vp);
return CallNonGenericMethod<IsNumber, num_toString_impl>(cx, args);
}
JS_ALWAYS_INLINE bool
num_toLocaleString_impl(JSContext *cx, CallArgs args)
{
JS_ASSERT(IsNumber(args.thisv()));
double d = Extract(args.thisv());
Rooted<JSString*> str(cx, js_NumberToStringWithBase(cx, d, 10));
if (!str) {
JS_ReportOutOfMemory(cx);
return false;
}
/*
* Create the string, move back to bytes to make string twiddling
* a bit easier and so we can insert platform charset seperators.
*/
JSAutoByteString numBytes(cx, str);
if (!numBytes)
return false;
const char *num = numBytes.ptr();
if (!num)
return false;
/*
* Find the first non-integer value, whether it be a letter as in
2008-07-24 12:59:11 -07:00
* 'Infinity', a decimal point, or an 'e' from exponential notation.
*/
const char *nint = num;
if (*nint == '-')
nint++;
while (*nint >= '0' && *nint <= '9')
nint++;
int digits = nint - num;
const char *end = num + digits;
if (!digits) {
args.rval().setString(str);
return true;
}
JSRuntime *rt = cx->runtime;
size_t thousandsLength = strlen(rt->thousandsSeparator);
size_t decimalLength = strlen(rt->decimalSeparator);
/* Figure out how long resulting string will be. */
int buflen = strlen(num);
if (*nint == '.')
buflen += decimalLength - 1; /* -1 to account for existing '.' */
const char *numGrouping;
const char *tmpGroup;
numGrouping = tmpGroup = rt->numGrouping;
int remainder = digits;
if (*num == '-')
remainder--;
while (*tmpGroup != CHAR_MAX && *tmpGroup != '\0') {
if (*tmpGroup >= remainder)
break;
buflen += thousandsLength;
remainder -= *tmpGroup;
tmpGroup++;
}
int nrepeat;
if (*tmpGroup == '\0' && *numGrouping != '\0') {
nrepeat = (remainder - 1) / tmpGroup[-1];
buflen += thousandsLength * nrepeat;
remainder -= nrepeat * tmpGroup[-1];
} else {
nrepeat = 0;
}
tmpGroup--;
char *buf = cx->pod_malloc<char>(buflen + 1);
if (!buf)
return false;
char *tmpDest = buf;
const char *tmpSrc = num;
while (*tmpSrc == '-' || remainder--) {
JS_ASSERT(tmpDest - buf < buflen);
*tmpDest++ = *tmpSrc++;
}
while (tmpSrc < end) {
JS_ASSERT(tmpDest - buf + ptrdiff_t(thousandsLength) <= buflen);
strcpy(tmpDest, rt->thousandsSeparator);
tmpDest += thousandsLength;
JS_ASSERT(tmpDest - buf + *tmpGroup <= buflen);
js_memcpy(tmpDest, tmpSrc, *tmpGroup);
tmpDest += *tmpGroup;
tmpSrc += *tmpGroup;
if (--nrepeat < 0)
tmpGroup--;
}
if (*nint == '.') {
JS_ASSERT(tmpDest - buf + ptrdiff_t(decimalLength) <= buflen);
strcpy(tmpDest, rt->decimalSeparator);
tmpDest += decimalLength;
JS_ASSERT(tmpDest - buf + ptrdiff_t(strlen(nint + 1)) <= buflen);
strcpy(tmpDest, nint + 1);
} else {
JS_ASSERT(tmpDest - buf + ptrdiff_t(strlen(nint)) <= buflen);
strcpy(tmpDest, nint);
}
if (cx->localeCallbacks && cx->localeCallbacks->localeToUnicode) {
Rooted<Value> v(cx, StringValue(str));
bool ok = !!cx->localeCallbacks->localeToUnicode(cx, buf, v.address());
if (ok)
args.rval().set(v);
js_free(buf);
return ok;
}
str = js_NewStringCopyN(cx, buf, buflen);
js_free(buf);
if (!str)
return false;
args.rval().setString(str);
return true;
}
JSBool
num_toLocaleString(JSContext *cx, unsigned argc, Value *vp)
{
CallArgs args = CallArgsFromVp(argc, vp);
return CallNonGenericMethod<IsNumber, num_toLocaleString_impl>(cx, args);
}
JS_ALWAYS_INLINE bool
num_valueOf_impl(JSContext *cx, CallArgs args)
{
JS_ASSERT(IsNumber(args.thisv()));
args.rval().setNumber(Extract(args.thisv()));
return true;
}
JSBool
js_num_valueOf(JSContext *cx, unsigned argc, Value *vp)
{
CallArgs args = CallArgsFromVp(argc, vp);
return CallNonGenericMethod<IsNumber, num_valueOf_impl>(cx, args);
}
const unsigned MAX_PRECISION = 100;
static bool
ComputePrecisionInRange(JSContext *cx, int minPrecision, int maxPrecision, const Value &v,
int *precision)
{
double prec;
if (!ToInteger(cx, v, &prec))
return false;
if (minPrecision <= prec && prec <= maxPrecision) {
*precision = int(prec);
return true;
}
ToCStringBuf cbuf;
char *numStr = IntToCString(&cbuf, *precision);
JS_ASSERT(numStr);
JS_ReportErrorNumber(cx, js_GetErrorMessage, NULL, JSMSG_PRECISION_RANGE, numStr);
return false;
}
static bool
DToStrResult(JSContext *cx, double d, JSDToStrMode mode, int precision, CallArgs args)
{
char buf[DTOSTR_VARIABLE_BUFFER_SIZE(MAX_PRECISION + 1)];
char *numStr = js_dtostr(cx->runtime->dtoaState, buf, sizeof buf, mode, precision, d);
if (!numStr) {
JS_ReportOutOfMemory(cx);
return false;
}
JSString *str = js_NewStringCopyZ(cx, numStr);
if (!str)
return false;
args.rval().setString(str);
return true;
}
/*
* In the following three implementations, we allow a larger range of precision
* than ECMA requires; this is permitted by ECMA-262.
*/
JS_ALWAYS_INLINE bool
num_toFixed_impl(JSContext *cx, CallArgs args)
{
JS_ASSERT(IsNumber(args.thisv()));
int precision;
if (args.length() == 0) {
precision = 0;
} else {
if (!ComputePrecisionInRange(cx, -20, MAX_PRECISION, args[0], &precision))
return false;
}
return DToStrResult(cx, Extract(args.thisv()), DTOSTR_FIXED, precision, args);
}
JSBool
num_toFixed(JSContext *cx, unsigned argc, Value *vp)
{
CallArgs args = CallArgsFromVp(argc, vp);
return CallNonGenericMethod<IsNumber, num_toFixed_impl>(cx, args);
}
JS_ALWAYS_INLINE bool
num_toExponential_impl(JSContext *cx, CallArgs args)
{
JS_ASSERT(IsNumber(args.thisv()));
JSDToStrMode mode;
int precision;
if (args.length() == 0) {
mode = DTOSTR_STANDARD_EXPONENTIAL;
precision = 0;
} else {
mode = DTOSTR_EXPONENTIAL;
if (!ComputePrecisionInRange(cx, 0, MAX_PRECISION, args[0], &precision))
return false;
}
return DToStrResult(cx, Extract(args.thisv()), mode, precision + 1, args);
}
JSBool
num_toExponential(JSContext *cx, unsigned argc, Value *vp)
{
CallArgs args = CallArgsFromVp(argc, vp);
return CallNonGenericMethod<IsNumber, num_toExponential_impl>(cx, args);
}
JS_ALWAYS_INLINE bool
num_toPrecision_impl(JSContext *cx, CallArgs args)
{
JS_ASSERT(IsNumber(args.thisv()));
double d = Extract(args.thisv());
if (!args.hasDefined(0)) {
JSString *str = js_NumberToStringWithBase(cx, d, 10);
if (!str) {
JS_ReportOutOfMemory(cx);
return false;
}
args.rval().setString(str);
return true;
}
JSDToStrMode mode;
int precision;
if (args.length() == 0) {
mode = DTOSTR_STANDARD;
precision = 0;
} else {
mode = DTOSTR_PRECISION;
if (!ComputePrecisionInRange(cx, 1, MAX_PRECISION, args[0], &precision))
return false;
}
return DToStrResult(cx, d, mode, precision, args);
}
JSBool
num_toPrecision(JSContext *cx, unsigned argc, Value *vp)
{
CallArgs args = CallArgsFromVp(argc, vp);
return CallNonGenericMethod<IsNumber, num_toPrecision_impl>(cx, args);
}
static JSFunctionSpec number_methods[] = {
#if JS_HAS_TOSOURCE
JS_FN(js_toSource_str, num_toSource, 0, 0),
#endif
JS_FN(js_toString_str, num_toString, 1, 0),
JS_FN(js_toLocaleString_str, num_toLocaleString, 0, 0),
JS_FN(js_valueOf_str, js_num_valueOf, 0, 0),
JS_FN("toFixed", num_toFixed, 1, 0),
JS_FN("toExponential", num_toExponential, 1, 0),
JS_FN("toPrecision", num_toPrecision, 1, 0),
JS_FS_END
};
// ES6 draft ES6 15.7.3.10
static JSBool
Number_isNaN(JSContext *cx, unsigned argc, Value *vp)
{
CallArgs args = CallArgsFromVp(argc, vp);
if (args.length() < 1 || !args[0].isDouble()) {
args.rval().setBoolean(false);
return true;
}
args.rval().setBoolean(MOZ_DOUBLE_IS_NaN(args[0].toDouble()));
return true;
}
// ES6 draft ES6 15.7.3.11
static JSBool
Number_isFinite(JSContext *cx, unsigned argc, Value *vp)
{
CallArgs args = CallArgsFromVp(argc, vp);
if (args.length() < 1 || !args[0].isNumber()) {
args.rval().setBoolean(false);
return true;
}
args.rval().setBoolean(args[0].isInt32() ||
MOZ_DOUBLE_IS_FINITE(args[0].toDouble()));
return true;
}
// ES6 draft ES6 15.7.3.12
static JSBool
Number_isInteger(JSContext *cx, unsigned argc, Value *vp)
{
CallArgs args = CallArgsFromVp(argc, vp);
if (args.length() < 1 || !args[0].isNumber()) {
args.rval().setBoolean(false);
return true;
}
Value val = args[0];
args.rval().setBoolean(val.isInt32() ||
(MOZ_DOUBLE_IS_FINITE(val.toDouble()) &&
ToInteger(val.toDouble()) == val.toDouble()));
return true;
}
// ES6 drafult ES6 15.7.3.13
static JSBool
Number_toInteger(JSContext *cx, unsigned argc, Value *vp)
{
CallArgs args = CallArgsFromVp(argc, vp);
if (args.length() < 1) {
args.rval().setInt32(0);
return true;
}
double asint;
if (!ToInteger(cx, args[0], &asint))
return false;
args.rval().setNumber(asint);
return true;
}
static JSFunctionSpec number_static_methods[] = {
JS_FN("isFinite", Number_isFinite, 1, 0),
JS_FN("isInteger", Number_isInteger, 1, 0),
JS_FN("isNaN", Number_isNaN, 1, 0),
JS_FN("toInteger", Number_toInteger, 1, 0),
JS_FS_END
};
/* NB: Keep this in synch with number_constants[]. */
enum nc_slot {
NC_NaN,
NC_POSITIVE_INFINITY,
NC_NEGATIVE_INFINITY,
NC_MAX_VALUE,
NC_MIN_VALUE,
NC_LIMIT
};
/*
* Some to most C compilers forbid spelling these at compile time, or barf
* if you try, so all but MAX_VALUE are set up by InitRuntimeNumberState
* using union jsdpun.
*/
static JSConstDoubleSpec number_constants[] = {
{0, "NaN", 0,{0,0,0}},
{0, "POSITIVE_INFINITY", 0,{0,0,0}},
{0, "NEGATIVE_INFINITY", 0,{0,0,0}},
{1.7976931348623157E+308, "MAX_VALUE", 0,{0,0,0}},
{0, "MIN_VALUE", 0,{0,0,0}},
{0,0,0,{0,0,0}}
};
double js_NaN;
double js_PositiveInfinity;
double js_NegativeInfinity;
#if (defined __GNUC__ && defined __i386__) || \
(defined __SUNPRO_CC && defined __i386)
/*
* Set the exception mask to mask all exceptions and set the FPU precision
* to 53 bit mantissa (64 bit doubles).
*/
inline void FIX_FPU() {
short control;
asm("fstcw %0" : "=m" (control) : );
control &= ~0x300; // Lower bits 8 and 9 (precision control).
control |= 0x2f3; // Raise bits 0-5 (exception masks) and 9 (64-bit precision).
asm("fldcw %0" : : "m" (control) );
}
#else
#define FIX_FPU() ((void)0)
#endif
namespace js {
bool
InitRuntimeNumberState(JSRuntime *rt)
{
FIX_FPU();
double d;
/*
* Our NaN must be one particular canonical value, because we rely on NaN
* encoding for our value representation. See jsval.h.
*/
d = MOZ_DOUBLE_SPECIFIC_NaN(0, 0x8000000000000ULL);
number_constants[NC_NaN].dval = js_NaN = d;
rt->NaNValue.setDouble(d);
d = MOZ_DOUBLE_POSITIVE_INFINITY();
number_constants[NC_POSITIVE_INFINITY].dval = js_PositiveInfinity = d;
rt->positiveInfinityValue.setDouble(d);
d = MOZ_DOUBLE_NEGATIVE_INFINITY();
number_constants[NC_NEGATIVE_INFINITY].dval = js_NegativeInfinity = d;
rt->negativeInfinityValue.setDouble(d);
number_constants[NC_MIN_VALUE].dval = MOZ_DOUBLE_MIN_VALUE();
/* Copy locale-specific separators into the runtime strings. */
const char *thousandsSeparator, *decimalPoint, *grouping;
#ifdef HAVE_LOCALECONV
struct lconv *locale = localeconv();
thousandsSeparator = locale->thousands_sep;
decimalPoint = locale->decimal_point;
grouping = locale->grouping;
#else
thousandsSeparator = getenv("LOCALE_THOUSANDS_SEP");
decimalPoint = getenv("LOCALE_DECIMAL_POINT");
grouping = getenv("LOCALE_GROUPING");
#endif
if (!thousandsSeparator)
thousandsSeparator = "'";
if (!decimalPoint)
decimalPoint = ".";
if (!grouping)
grouping = "\3\0";
/*
* We use single malloc to get the memory for all separator and grouping
* strings.
*/
size_t thousandsSeparatorSize = strlen(thousandsSeparator) + 1;
size_t decimalPointSize = strlen(decimalPoint) + 1;
size_t groupingSize = strlen(grouping) + 1;
char *storage = js_pod_malloc<char>(thousandsSeparatorSize +
decimalPointSize +
groupingSize);
if (!storage)
return false;
js_memcpy(storage, thousandsSeparator, thousandsSeparatorSize);
rt->thousandsSeparator = storage;
storage += thousandsSeparatorSize;
js_memcpy(storage, decimalPoint, decimalPointSize);
rt->decimalSeparator = storage;
storage += decimalPointSize;
js_memcpy(storage, grouping, groupingSize);
rt->numGrouping = grouping;
return true;
}
void
FinishRuntimeNumberState(JSRuntime *rt)
{
/*
* The free also releases the memory for decimalSeparator and numGrouping
* strings.
*/
char *storage = const_cast<char *>(rt->thousandsSeparator);
js_free(storage);
}
} /* namespace js */
JSObject *
js_InitNumberClass(JSContext *cx, JSObject *obj)
{
JS_ASSERT(obj->isNative());
/* XXX must do at least once per new thread, so do it per JSContext... */
FIX_FPU();
Rooted<GlobalObject*> global(cx, &obj->asGlobal());
RootedObject numberProto(cx, global->createBlankPrototype(cx, &NumberClass));
if (!numberProto)
return NULL;
numberProto->asNumber().setPrimitiveValue(0);
RootedFunction ctor(cx);
ctor = global->createConstructor(cx, Number, cx->names().Number, 1);
if (!ctor)
return NULL;
if (!LinkConstructorAndPrototype(cx, ctor, numberProto))
return NULL;
/* Add numeric constants (MAX_VALUE, NaN, &c.) to the Number constructor. */
if (!JS_DefineConstDoubles(cx, ctor, number_constants))
return NULL;
if (!DefinePropertiesAndBrand(cx, ctor, NULL, number_static_methods))
return NULL;
if (!DefinePropertiesAndBrand(cx, numberProto, NULL, number_methods))
return NULL;
if (!JS_DefineFunctions(cx, global, number_functions))
return NULL;
RootedValue valueNaN(cx, cx->runtime->NaNValue);
RootedValue valueInfinity(cx, cx->runtime->positiveInfinityValue);
/* ES5 15.1.1.1, 15.1.1.2 */
if (!DefineNativeProperty(cx, global, cx->names().NaN, valueNaN,
JS_PropertyStub, JS_StrictPropertyStub,
JSPROP_PERMANENT | JSPROP_READONLY, 0, 0) ||
!DefineNativeProperty(cx, global, cx->names().Infinity, valueInfinity,
JS_PropertyStub, JS_StrictPropertyStub,
JSPROP_PERMANENT | JSPROP_READONLY, 0, 0))
{
return NULL;
}
if (!DefineConstructorAndPrototype(cx, global, JSProto_Number, ctor, numberProto))
return NULL;
return numberProto;
}
namespace js {
static char *
FracNumberToCString(JSContext *cx, ToCStringBuf *cbuf, double d, int base = 10)
{
#ifdef DEBUG
{
int32_t _;
JS_ASSERT(!MOZ_DOUBLE_IS_INT32(d, &_));
}
#endif
char* numStr;
if (base == 10) {
/*
* This is V8's implementation of the algorithm described in the
* following paper:
*
* Printing floating-point numbers quickly and accurately with integers.
* Florian Loitsch, PLDI 2010.
*/
const double_conversion::DoubleToStringConverter &converter
= double_conversion::DoubleToStringConverter::EcmaScriptConverter();
double_conversion::StringBuilder builder(cbuf->sbuf, cbuf->sbufSize);
converter.ToShortest(d, &builder);
numStr = builder.Finalize();
} else {
numStr = cbuf->dbuf = js_dtobasestr(cx->runtime->dtoaState, base, d);
}
return numStr;
}
char *
NumberToCString(JSContext *cx, ToCStringBuf *cbuf, double d, int base/* = 10*/)
{
2010-07-14 23:19:36 -07:00
int32_t i;
return MOZ_DOUBLE_IS_INT32(d, &i)
? IntToCString(cbuf, i, base)
: FracNumberToCString(cx, cbuf, d, base);
}
}
static JSString * JS_FASTCALL
js_NumberToStringWithBase(JSContext *cx, double d, int base)
{
ToCStringBuf cbuf;
char *numStr;
/*
* Caller is responsible for error reporting. When called from trace,
* returning NULL here will cause us to fall of trace and then retry
* from the interpreter (which will report the error).
*/
if (base < 2 || base > 36)
return NULL;
JSCompartment *c = cx->compartment;
2010-07-14 23:19:36 -07:00
int32_t i;
if (MOZ_DOUBLE_IS_INT32(d, &i)) {
if (base == 10 && StaticStrings::hasInt(i))
return cx->runtime->staticStrings.getInt(i);
if (unsigned(i) < unsigned(base)) {
if (i < 10)
return cx->runtime->staticStrings.getInt(i);
jschar c = 'a' + i - 10;
JS_ASSERT(StaticStrings::hasUnit(c));
return cx->runtime->staticStrings.getUnit(c);
}
if (JSFlatString *str = c->dtoaCache.lookup(base, d))
return str;
numStr = IntToCString(&cbuf, i, base);
JS_ASSERT(!cbuf.dbuf && numStr >= cbuf.sbuf && numStr < cbuf.sbuf + cbuf.sbufSize);
} else {
if (JSFlatString *str = c->dtoaCache.lookup(base, d))
return str;
numStr = FracNumberToCString(cx, &cbuf, d, base);
if (!numStr) {
JS_ReportOutOfMemory(cx);
return NULL;
}
JS_ASSERT_IF(base == 10,
!cbuf.dbuf && numStr >= cbuf.sbuf && numStr < cbuf.sbuf + cbuf.sbufSize);
JS_ASSERT_IF(base != 10,
cbuf.dbuf && cbuf.dbuf == numStr);
}
JSFlatString *s = js_NewStringCopyZ(cx, numStr);
c->dtoaCache.cache(base, d, s);
return s;
}
JSString *
js_NumberToString(JSContext *cx, double d)
{
return js_NumberToStringWithBase(cx, d, 10);
}
namespace js {
JSFlatString *
NumberToString(JSContext *cx, double d)
{
if (JSString *str = js_NumberToStringWithBase(cx, d, 10))
return &str->asFlat();
return NULL;
}
JSFlatString *
IndexToString(JSContext *cx, uint32_t index)
{
if (StaticStrings::hasUint(index))
return cx->runtime->staticStrings.getUint(index);
JSCompartment *c = cx->compartment;
if (JSFlatString *str = c->dtoaCache.lookup(10, index))
return str;
JSShortString *str = js_NewGCShortString(cx);
if (!str)
return NULL;
jschar buffer[JSShortString::MAX_SHORT_LENGTH + 1];
RangedPtr<jschar> end(buffer + JSShortString::MAX_SHORT_LENGTH,
buffer, JSShortString::MAX_SHORT_LENGTH + 1);
*end = '\0';
RangedPtr<jschar> start = BackfillIndexInCharBuffer(index, end);
jschar *dst = str->init(end - start);
PodCopy(dst, start.get(), end - start + 1);
c->dtoaCache.cache(10, index, str);
return str;
}
bool JS_FASTCALL
NumberValueToStringBuffer(JSContext *cx, const Value &v, StringBuffer &sb)
{
/* Convert to C-string. */
ToCStringBuf cbuf;
const char *cstr;
2010-07-14 23:19:36 -07:00
if (v.isInt32()) {
cstr = IntToCString(&cbuf, v.toInt32());
} else {
cstr = NumberToCString(cx, &cbuf, v.toDouble());
if (!cstr) {
JS_ReportOutOfMemory(cx);
return JS_FALSE;
}
}
/*
* Inflate to jschar string. The input C-string characters are < 127, so
* even if jschars are UTF-8, all chars should map to one jschar.
*/
size_t cstrlen = strlen(cstr);
JS_ASSERT(!cbuf.dbuf && cstrlen < cbuf.sbufSize);
return sb.appendInflated(cstr, cstrlen);
}
JS_PUBLIC_API(bool)
ToNumberSlow(JSContext *cx, Value v, double *out)
{
#ifdef DEBUG
/*
* MSVC bizarrely miscompiles this, complaining about the first brace below
* being unmatched (!). The error message points at both this opening brace
* and at the corresponding SkipRoot constructor. The error seems to derive
* from the presence guard-object macros on the SkipRoot class/constructor,
* which seems well in the weeds for an unmatched-brace syntax error.
* Otherwise the problem is inscrutable, and I haven't found a workaround.
* So for now just disable it when compiling with MSVC -- not ideal, but at
* least Windows debug shell builds complete again.
*/
#ifndef _MSC_VER
{
SkipRoot skip(cx, &v);
MaybeCheckStackRoots(cx);
}
#endif
#endif
2010-07-14 23:19:36 -07:00
JS_ASSERT(!v.isNumber());
goto skip_int_double;
for (;;) {
2010-07-14 23:19:36 -07:00
if (v.isNumber()) {
*out = v.toNumber();
return true;
}
skip_int_double:
if (v.isString())
return StringToNumberType<double>(cx, v.toString(), out);
2010-07-14 23:19:36 -07:00
if (v.isBoolean()) {
if (v.toBoolean()) {
*out = 1.0;
2010-07-14 23:19:36 -07:00
return true;
}
*out = 0.0;
2010-07-14 23:19:36 -07:00
return true;
}
2010-07-14 23:19:36 -07:00
if (v.isNull()) {
*out = 0.0;
2010-07-14 23:19:36 -07:00
return true;
}
2010-07-14 23:19:36 -07:00
if (v.isUndefined())
break;
2010-07-14 23:19:36 -07:00
JS_ASSERT(v.isObject());
if (!ToPrimitive(cx, JSTYPE_NUMBER, &v))
2010-07-14 23:19:36 -07:00
return false;
if (v.isObject())
break;
}
*out = js_NaN;
2010-07-14 23:19:36 -07:00
return true;
}
/*
* Convert a value to an int64_t, according to the WebIDL rules for long long
* conversion. Return converted value in *out on success, false on failure.
*/
JS_PUBLIC_API(bool)
ToInt64Slow(JSContext *cx, const Value &v, int64_t *out)
{
JS_ASSERT(!v.isInt32());
double d;
if (v.isDouble()) {
d = v.toDouble();
} else {
if (!ToNumberSlow(cx, v, &d))
return false;
}
*out = ToInt64(d);
return true;
}
/*
* Convert a value to an uint64_t, according to the WebIDL rules for unsigned long long
* conversion. Return converted value in *out on success, false on failure.
*/
JS_PUBLIC_API(bool)
ToUint64Slow(JSContext *cx, const Value &v, uint64_t *out)
{
JS_ASSERT(!v.isInt32());
double d;
if (v.isDouble()) {
d = v.toDouble();
} else {
if (!ToNumberSlow(cx, v, &d))
return false;
}
*out = ToUint64(d);
return true;
}
JS_PUBLIC_API(bool)
ToInt32Slow(JSContext *cx, const Value &v, int32_t *out)
{
2010-07-14 23:19:36 -07:00
JS_ASSERT(!v.isInt32());
double d;
2010-07-14 23:19:36 -07:00
if (v.isDouble()) {
d = v.toDouble();
2008-03-06 13:40:43 -08:00
} else {
if (!ToNumberSlow(cx, v, &d))
return false;
2008-03-06 13:40:43 -08:00
}
*out = ToInt32(d);
return true;
}
JS_PUBLIC_API(bool)
ToUint32Slow(JSContext *cx, const Value &v, uint32_t *out)
{
2010-07-14 23:19:36 -07:00
JS_ASSERT(!v.isInt32());
double d;
2010-07-14 23:19:36 -07:00
if (v.isDouble()) {
d = v.toDouble();
2008-03-06 13:40:43 -08:00
} else {
if (!ToNumberSlow(cx, v, &d))
return false;
2008-03-06 13:40:43 -08:00
}
*out = ToUint32(d);
return true;
}
JS_PUBLIC_API(bool)
ToUint16Slow(JSContext *cx, const Value &v, uint16_t *out)
{
2010-07-14 23:19:36 -07:00
JS_ASSERT(!v.isInt32());
double d;
2010-07-14 23:19:36 -07:00
if (v.isDouble()) {
d = v.toDouble();
} else if (!ToNumberSlow(cx, v, &d)) {
2010-07-14 23:19:36 -07:00
return false;
}
if (d == 0 || !MOZ_DOUBLE_IS_FINITE(d)) {
*out = 0;
return true;
}
uint16_t u = (uint16_t) d;
if ((double)u == d) {
*out = u;
return true;
}
bool neg = (d < 0);
d = floor(neg ? -d : d);
d = neg ? -d : d;
unsigned m = JS_BIT(16);
d = fmod(d, (double) m);
if (d < 0)
d += m;
*out = (uint16_t) d;
return true;
}
} /* namespace js */
JSBool
js_strtod(JSContext *cx, const jschar *s, const jschar *send,
const jschar **ep, double *dp)
{
size_t i;
char cbuf[32];
char *cstr, *istr, *estr;
JSBool negative;
double d;
const jschar *s1 = SkipSpace(s, send);
size_t length = send - s1;
/* Use cbuf to avoid malloc */
if (length >= sizeof cbuf) {
cstr = (char *) cx->malloc_(length + 1);
if (!cstr)
return JS_FALSE;
} else {
cstr = cbuf;
}
for (i = 0; i != length; i++) {
if (s1[i] >> 8)
break;
cstr[i] = (char)s1[i];
}
cstr[i] = 0;
istr = cstr;
if ((negative = (*istr == '-')) != 0 || *istr == '+')
istr++;
if (*istr == 'I' && !strncmp(istr, "Infinity", 8)) {
d = negative ? js_NegativeInfinity : js_PositiveInfinity;
estr = istr + 8;
} else {
int err;
d = js_strtod_harder(cx->runtime->dtoaState, cstr, &estr, &err);
if (d == HUGE_VAL)
d = js_PositiveInfinity;
else if (d == -HUGE_VAL)
d = js_NegativeInfinity;
}
i = estr - cstr;
if (cstr != cbuf)
js_free(cstr);
*ep = i ? s1 + i : s;
*dp = d;
return JS_TRUE;
}