rust/src/libstd/num/strconv.rs
Daniel Micay c9d4ad07c4 remove the float type
It is simply defined as `f64` across every platform right now.

A use case hasn't been presented for a `float` type defined as the
highest precision floating point type implemented in hardware on the
platform. Performance-wise, using the smallest precision correct for the
use case greatly saves on cache space and allows for fitting more
numbers into SSE/AVX registers.

If there was a use case, this could be implemented as simply a type
alias or a struct thanks to `#[cfg(...)]`.

Closes #6592

The mailing list thread, for reference:

https://mail.mozilla.org/pipermail/rust-dev/2013-July/004632.html
2013-10-01 14:54:10 -04:00

756 lines
27 KiB
Rust

// Copyright 2013 The Rust Project Developers. See the COPYRIGHT
// file at the top-level directory of this distribution and at
// http://rust-lang.org/COPYRIGHT.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.
#[allow(missing_doc)];
use clone::Clone;
use container::Container;
use std::cmp::{Ord, Eq};
use ops::{Add, Sub, Mul, Div, Rem, Neg};
use option::{None, Option, Some};
use char;
use str::{StrSlice};
use str;
use vec::{CopyableVector, ImmutableVector, MutableVector};
use vec::OwnedVector;
use num::{NumCast, Zero, One, cast, pow_with_uint, Integer};
use num::{Round, Float, FPNaN, FPInfinite};
pub enum ExponentFormat {
ExpNone,
ExpDec,
ExpBin
}
pub enum SignificantDigits {
DigAll,
DigMax(uint),
DigExact(uint)
}
pub enum SignFormat {
SignNone,
SignNeg,
SignAll
}
pub trait NumStrConv {
fn nan() -> Option<Self>;
fn inf() -> Option<Self>;
fn neg_inf() -> Option<Self>;
fn neg_zero() -> Option<Self>;
fn round_to_zero(&self) -> Self;
fn fractional_part(&self) -> Self;
}
macro_rules! impl_NumStrConv_Floating (($t:ty) => (
impl NumStrConv for $t {
#[inline]
fn nan() -> Option<$t> { Some( 0.0 / 0.0) }
#[inline]
fn inf() -> Option<$t> { Some( 1.0 / 0.0) }
#[inline]
fn neg_inf() -> Option<$t> { Some(-1.0 / 0.0) }
#[inline]
fn neg_zero() -> Option<$t> { Some(-0.0 ) }
#[inline]
fn round_to_zero(&self) -> $t { self.trunc() }
#[inline]
fn fractional_part(&self) -> $t { self.fract() }
}
))
macro_rules! impl_NumStrConv_Integer (($t:ty) => (
impl NumStrConv for $t {
#[inline] fn nan() -> Option<$t> { None }
#[inline] fn inf() -> Option<$t> { None }
#[inline] fn neg_inf() -> Option<$t> { None }
#[inline] fn neg_zero() -> Option<$t> { None }
#[inline] fn round_to_zero(&self) -> $t { *self }
#[inline] fn fractional_part(&self) -> $t { 0 }
}
))
// FIXME: #4955
// Replace by two generic impls for traits 'Integral' and 'Floating'
impl_NumStrConv_Floating!(f32)
impl_NumStrConv_Floating!(f64)
impl_NumStrConv_Integer!(int)
impl_NumStrConv_Integer!(i8)
impl_NumStrConv_Integer!(i16)
impl_NumStrConv_Integer!(i32)
impl_NumStrConv_Integer!(i64)
impl_NumStrConv_Integer!(uint)
impl_NumStrConv_Integer!(u8)
impl_NumStrConv_Integer!(u16)
impl_NumStrConv_Integer!(u32)
impl_NumStrConv_Integer!(u64)
// Special value strings as [u8] consts.
static INF_BUF: [u8, ..3] = ['i' as u8, 'n' as u8, 'f' as u8];
static POS_INF_BUF: [u8, ..4] = ['+' as u8, 'i' as u8, 'n' as u8,
'f' as u8];
static NEG_INF_BUF: [u8, ..4] = ['-' as u8, 'i' as u8, 'n' as u8,
'f' as u8];
static NAN_BUF: [u8, ..3] = ['N' as u8, 'a' as u8, 'N' as u8];
/**
* Converts an integral number to its string representation as a byte vector.
* This is meant to be a common base implementation for all integral string
* conversion functions like `to_str()` or `to_str_radix()`.
*
* # Arguments
* - `num` - The number to convert. Accepts any number that
* implements the numeric traits.
* - `radix` - Base to use. Accepts only the values 2-36.
* - `sign` - How to emit the sign. Options are:
* - `SignNone`: No sign at all. Basically emits `abs(num)`.
* - `SignNeg`: Only `-` on negative values.
* - `SignAll`: Both `+` on positive, and `-` on negative numbers.
* - `f` - a callback which will be invoked for each ascii character
* which composes the string representation of this integer
*
* # Return value
* A tuple containing the byte vector, and a boolean flag indicating
* whether it represents a special value like `inf`, `-inf`, `NaN` or not.
* It returns a tuple because there can be ambiguity between a special value
* and a number representation at higher bases.
*
* # Failure
* - Fails if `radix` < 2 or `radix` > 36.
*/
pub fn int_to_str_bytes_common<T:NumCast+Zero+Eq+Ord+Integer+
Div<T,T>+Neg<T>+Rem<T,T>+Mul<T,T>>(
num: T, radix: uint, sign: SignFormat, f: &fn(u8)) {
assert!(2 <= radix && radix <= 36);
let _0: T = Zero::zero();
let neg = num < _0;
let radix_gen: T = cast(radix);
let mut deccum = num;
// This is just for integral types, the largest of which is a u64. The
// smallest base that we can have is 2, so the most number of digits we're
// ever going to have is 64
let mut buf = [0u8, ..64];
let mut cur = 0;
// Loop at least once to make sure at least a `0` gets emitted.
loop {
// Calculate the absolute value of each digit instead of only
// doing it once for the whole number because a
// representable negative number doesn't necessary have an
// representable additive inverse of the same type
// (See twos complement). But we assume that for the
// numbers [-35 .. 0] we always have [0 .. 35].
let current_digit_signed = deccum % radix_gen;
let current_digit = if current_digit_signed < _0 {
-current_digit_signed
} else {
current_digit_signed
};
buf[cur] = match current_digit.to_u8() {
i @ 0..9 => '0' as u8 + i,
i => 'a' as u8 + (i - 10),
};
cur += 1;
deccum = deccum / radix_gen;
// No more digits to calculate for the non-fractional part -> break
if deccum == _0 { break; }
}
// Decide what sign to put in front
match sign {
SignNeg | SignAll if neg => { f('-' as u8); }
SignAll => { f('+' as u8); }
_ => ()
}
// We built the number in reverse order, so un-reverse it here
while cur > 0 {
cur -= 1;
f(buf[cur]);
}
}
/**
* Converts a number to its string representation as a byte vector.
* This is meant to be a common base implementation for all numeric string
* conversion functions like `to_str()` or `to_str_radix()`.
*
* # Arguments
* - `num` - The number to convert. Accepts any number that
* implements the numeric traits.
* - `radix` - Base to use. Accepts only the values 2-36.
* - `negative_zero` - Whether to treat the special value `-0` as
* `-0` or as `+0`.
* - `sign` - How to emit the sign. Options are:
* - `SignNone`: No sign at all. Basically emits `abs(num)`.
* - `SignNeg`: Only `-` on negative values.
* - `SignAll`: Both `+` on positive, and `-` on negative numbers.
* - `digits` - The amount of digits to use for emitting the
* fractional part, if any. Options are:
* - `DigAll`: All calculatable digits. Beware of bignums or
* fractions!
* - `DigMax(uint)`: Maximum N digits, truncating any trailing zeros.
* - `DigExact(uint)`: Exactly N digits.
*
* # Return value
* A tuple containing the byte vector, and a boolean flag indicating
* whether it represents a special value like `inf`, `-inf`, `NaN` or not.
* It returns a tuple because there can be ambiguity between a special value
* and a number representation at higher bases.
*
* # Failure
* - Fails if `radix` < 2 or `radix` > 36.
*/
pub fn float_to_str_bytes_common<T:NumCast+Zero+One+Eq+Ord+Float+Round+
Div<T,T>+Neg<T>+Rem<T,T>+Mul<T,T>>(
num: T, radix: uint, negative_zero: bool,
sign: SignFormat, digits: SignificantDigits) -> (~[u8], bool) {
assert!(2 <= radix && radix <= 36);
let _0: T = Zero::zero();
let _1: T = One::one();
match num.classify() {
FPNaN => { return ("NaN".as_bytes().to_owned(), true); }
FPInfinite if num > _0 => {
return match sign {
SignAll => ("+inf".as_bytes().to_owned(), true),
_ => ("inf".as_bytes().to_owned(), true)
};
}
FPInfinite if num < _0 => {
return match sign {
SignNone => ("inf".as_bytes().to_owned(), true),
_ => ("-inf".as_bytes().to_owned(), true),
};
}
_ => {}
}
let neg = num < _0 || (negative_zero && _1 / num == Float::neg_infinity());
let mut buf: ~[u8] = ~[];
let radix_gen: T = cast(radix as int);
// First emit the non-fractional part, looping at least once to make
// sure at least a `0` gets emitted.
let mut deccum = num.trunc();
loop {
// Calculate the absolute value of each digit instead of only
// doing it once for the whole number because a
// representable negative number doesn't necessary have an
// representable additive inverse of the same type
// (See twos complement). But we assume that for the
// numbers [-35 .. 0] we always have [0 .. 35].
let current_digit = (deccum % radix_gen).abs();
// Decrease the deccumulator one digit at a time
deccum = deccum / radix_gen;
deccum = deccum.trunc();
buf.push(char::from_digit(current_digit.to_int() as uint, radix)
.unwrap() as u8);
// No more digits to calculate for the non-fractional part -> break
if deccum == _0 { break; }
}
// If limited digits, calculate one digit more for rounding.
let (limit_digits, digit_count, exact) = match digits {
DigAll => (false, 0u, false),
DigMax(count) => (true, count+1, false),
DigExact(count) => (true, count+1, true)
};
// Decide what sign to put in front
match sign {
SignNeg | SignAll if neg => {
buf.push('-' as u8);
}
SignAll => {
buf.push('+' as u8);
}
_ => ()
}
buf.reverse();
// Remember start of the fractional digits.
// Points one beyond end of buf if none get generated,
// or at the '.' otherwise.
let start_fractional_digits = buf.len();
// Now emit the fractional part, if any
deccum = num.fract();
if deccum != _0 || (limit_digits && exact && digit_count > 0) {
buf.push('.' as u8);
let mut dig = 0u;
// calculate new digits while
// - there is no limit and there are digits left
// - or there is a limit, it's not reached yet and
// - it's exact
// - or it's a maximum, and there are still digits left
while (!limit_digits && deccum != _0)
|| (limit_digits && dig < digit_count && (
exact
|| (!exact && deccum != _0)
)
) {
// Shift first fractional digit into the integer part
deccum = deccum * radix_gen;
// Calculate the absolute value of each digit.
// See note in first loop.
let current_digit = deccum.trunc().abs();
buf.push(char::from_digit(
current_digit.to_int() as uint, radix).unwrap() as u8);
// Decrease the deccumulator one fractional digit at a time
deccum = deccum.fract();
dig += 1u;
}
// If digits are limited, and that limit has been reached,
// cut off the one extra digit, and depending on its value
// round the remaining ones.
if limit_digits && dig == digit_count {
let ascii2value = |chr: u8| {
char::to_digit(chr as char, radix).unwrap() as uint
};
let value2ascii = |val: uint| {
char::from_digit(val, radix).unwrap() as u8
};
let extra_digit = ascii2value(buf.pop());
if extra_digit >= radix / 2 { // -> need to round
let mut i: int = buf.len() as int - 1;
loop {
// If reached left end of number, have to
// insert additional digit:
if i < 0
|| buf[i] == '-' as u8
|| buf[i] == '+' as u8 {
buf.insert((i + 1) as uint, value2ascii(1));
break;
}
// Skip the '.'
if buf[i] == '.' as u8 { i -= 1; loop; }
// Either increment the digit,
// or set to 0 if max and carry the 1.
let current_digit = ascii2value(buf[i]);
if current_digit < (radix - 1) {
buf[i] = value2ascii(current_digit+1);
break;
} else {
buf[i] = value2ascii(0);
i -= 1;
}
}
}
}
}
// if number of digits is not exact, remove all trailing '0's up to
// and including the '.'
if !exact {
let buf_max_i = buf.len() - 1;
// index to truncate from
let mut i = buf_max_i;
// discover trailing zeros of fractional part
while i > start_fractional_digits && buf[i] == '0' as u8 {
i -= 1;
}
// Only attempt to truncate digits if buf has fractional digits
if i >= start_fractional_digits {
// If buf ends with '.', cut that too.
if buf[i] == '.' as u8 { i -= 1 }
// only resize buf if we actually remove digits
if i < buf_max_i {
buf = buf.slice(0, i + 1).to_owned();
}
}
} // If exact and trailing '.', just cut that
else {
let max_i = buf.len() - 1;
if buf[max_i] == '.' as u8 {
buf = buf.slice(0, max_i).to_owned();
}
}
(buf, false)
}
/**
* Converts a number to its string representation. This is a wrapper for
* `to_str_bytes_common()`, for details see there.
*/
#[inline]
pub fn float_to_str_common<T:NumCast+Zero+One+Eq+Ord+NumStrConv+Float+Round+
Div<T,T>+Neg<T>+Rem<T,T>+Mul<T,T>>(
num: T, radix: uint, negative_zero: bool,
sign: SignFormat, digits: SignificantDigits) -> (~str, bool) {
let (bytes, special) = float_to_str_bytes_common(num, radix,
negative_zero, sign, digits);
(str::from_utf8(bytes), special)
}
// Some constants for from_str_bytes_common's input validation,
// they define minimum radix values for which the character is a valid digit.
static DIGIT_P_RADIX: uint = ('p' as uint) - ('a' as uint) + 11u;
static DIGIT_I_RADIX: uint = ('i' as uint) - ('a' as uint) + 11u;
static DIGIT_E_RADIX: uint = ('e' as uint) - ('a' as uint) + 11u;
/**
* Parses a byte slice as a number. This is meant to
* be a common base implementation for all numeric string conversion
* functions like `from_str()` or `from_str_radix()`.
*
* # Arguments
* - `buf` - The byte slice to parse.
* - `radix` - Which base to parse the number as. Accepts 2-36.
* - `negative` - Whether to accept negative numbers.
* - `fractional` - Whether to accept numbers with fractional parts.
* - `special` - Whether to accept special values like `inf`
* and `NaN`. Can conflict with `radix`, see Failure.
* - `exponent` - Which exponent format to accept. Options are:
* - `ExpNone`: No Exponent, accepts just plain numbers like `42` or
* `-8.2`.
* - `ExpDec`: Accepts numbers with a decimal exponent like `42e5` or
* `8.2E-2`. The exponent string itself is always base 10.
* Can conflict with `radix`, see Failure.
* - `ExpBin`: Accepts numbers with a binary exponent like `42P-8` or
* `FFp128`. The exponent string itself is always base 10.
* Can conflict with `radix`, see Failure.
* - `empty_zero` - Whether to accept a empty `buf` as a 0 or not.
* - `ignore_underscores` - Whether all underscores within the string should
* be ignored.
*
* # Return value
* Returns `Some(n)` if `buf` parses to a number n without overflowing, and
* `None` otherwise, depending on the constraints set by the remaining
* arguments.
*
* # Failure
* - Fails if `radix` < 2 or `radix` > 36.
* - Fails if `radix` > 14 and `exponent` is `ExpDec` due to conflict
* between digit and exponent sign `'e'`.
* - Fails if `radix` > 25 and `exponent` is `ExpBin` due to conflict
* between digit and exponent sign `'p'`.
* - Fails if `radix` > 18 and `special == true` due to conflict
* between digit and lowest first character in `inf` and `NaN`, the `'i'`.
*/
pub fn from_str_bytes_common<T:NumCast+Zero+One+Eq+Ord+Div<T,T>+
Mul<T,T>+Sub<T,T>+Neg<T>+Add<T,T>+
NumStrConv+Clone>(
buf: &[u8], radix: uint, negative: bool, fractional: bool,
special: bool, exponent: ExponentFormat, empty_zero: bool,
ignore_underscores: bool
) -> Option<T> {
match exponent {
ExpDec if radix >= DIGIT_E_RADIX // decimal exponent 'e'
=> fail2!("from_str_bytes_common: radix {:?} incompatible with \
use of 'e' as decimal exponent", radix),
ExpBin if radix >= DIGIT_P_RADIX // binary exponent 'p'
=> fail2!("from_str_bytes_common: radix {:?} incompatible with \
use of 'p' as binary exponent", radix),
_ if special && radix >= DIGIT_I_RADIX // first digit of 'inf'
=> fail2!("from_str_bytes_common: radix {:?} incompatible with \
special values 'inf' and 'NaN'", radix),
_ if (radix as int) < 2
=> fail2!("from_str_bytes_common: radix {:?} to low, \
must lie in the range [2, 36]", radix),
_ if (radix as int) > 36
=> fail2!("from_str_bytes_common: radix {:?} to high, \
must lie in the range [2, 36]", radix),
_ => ()
}
let _0: T = Zero::zero();
let _1: T = One::one();
let radix_gen: T = cast(radix as int);
let len = buf.len();
if len == 0 {
if empty_zero {
return Some(_0);
} else {
return None;
}
}
if special {
if buf == INF_BUF || buf == POS_INF_BUF {
return NumStrConv::inf();
} else if buf == NEG_INF_BUF {
if negative {
return NumStrConv::neg_inf();
} else {
return None;
}
} else if buf == NAN_BUF {
return NumStrConv::nan();
}
}
let (start, accum_positive) = match buf[0] as char {
'-' if !negative => return None,
'-' => (1u, false),
'+' => (1u, true),
_ => (0u, true)
};
// Initialize accumulator with signed zero for floating point parsing to
// work
let mut accum = if accum_positive { _0.clone() } else { -_1 * _0};
let mut last_accum = accum.clone(); // Necessary to detect overflow
let mut i = start;
let mut exp_found = false;
// Parse integer part of number
while i < len {
let c = buf[i] as char;
match char::to_digit(c, radix) {
Some(digit) => {
// shift accum one digit left
accum = accum * radix_gen.clone();
// add/subtract current digit depending on sign
if accum_positive {
accum = accum + cast(digit as int);
} else {
accum = accum - cast(digit as int);
}
// Detect overflow by comparing to last value, except
// if we've not seen any non-zero digits.
if last_accum != _0 {
if accum_positive && accum <= last_accum { return NumStrConv::inf(); }
if !accum_positive && accum >= last_accum { return NumStrConv::neg_inf(); }
// Detect overflow by reversing the shift-and-add proccess
if accum_positive &&
(last_accum != ((accum - cast(digit as int))/radix_gen.clone())) {
return NumStrConv::inf();
}
if !accum_positive &&
(last_accum != ((accum + cast(digit as int))/radix_gen.clone())) {
return NumStrConv::neg_inf();
}
}
last_accum = accum.clone();
}
None => match c {
'_' if ignore_underscores => {}
'e' | 'E' | 'p' | 'P' => {
exp_found = true;
break; // start of exponent
}
'.' if fractional => {
i += 1u; // skip the '.'
break; // start of fractional part
}
_ => return None // invalid number
}
}
i += 1u;
}
// Parse fractional part of number
// Skip if already reached start of exponent
if !exp_found {
let mut power = _1.clone();
while i < len {
let c = buf[i] as char;
match char::to_digit(c, radix) {
Some(digit) => {
// Decrease power one order of magnitude
power = power / radix_gen;
let digit_t: T = cast(digit);
// add/subtract current digit depending on sign
if accum_positive {
accum = accum + digit_t * power;
} else {
accum = accum - digit_t * power;
}
// Detect overflow by comparing to last value
if accum_positive && accum < last_accum { return NumStrConv::inf(); }
if !accum_positive && accum > last_accum { return NumStrConv::neg_inf(); }
last_accum = accum.clone();
}
None => match c {
'_' if ignore_underscores => {}
'e' | 'E' | 'p' | 'P' => {
exp_found = true;
break; // start of exponent
}
_ => return None // invalid number
}
}
i += 1u;
}
}
// Special case: buf not empty, but does not contain any digit in front
// of the exponent sign -> number is empty string
if i == start {
if empty_zero {
return Some(_0);
} else {
return None;
}
}
let mut multiplier = _1.clone();
if exp_found {
let c = buf[i] as char;
let base = match (c, exponent) {
// c is never _ so don't need to handle specially
('e', ExpDec) | ('E', ExpDec) => 10u,
('p', ExpBin) | ('P', ExpBin) => 2u,
_ => return None // char doesn't fit given exponent format
};
// parse remaining bytes as decimal integer,
// skipping the exponent char
let exp: Option<int> = from_str_bytes_common(
buf.slice(i+1, len), 10, true, false, false, ExpNone, false,
ignore_underscores);
match exp {
Some(exp_pow) => {
multiplier = if exp_pow < 0 {
_1 / pow_with_uint::<T>(base, (-exp_pow.to_int()) as uint)
} else {
pow_with_uint::<T>(base, exp_pow.to_int() as uint)
}
}
None => return None // invalid exponent -> invalid number
}
}
Some(accum * multiplier)
}
/**
* Parses a string as a number. This is a wrapper for
* `from_str_bytes_common()`, for details see there.
*/
#[inline]
pub fn from_str_common<T:NumCast+Zero+One+Eq+Ord+Div<T,T>+Mul<T,T>+
Sub<T,T>+Neg<T>+Add<T,T>+NumStrConv+Clone>(
buf: &str, radix: uint, negative: bool, fractional: bool,
special: bool, exponent: ExponentFormat, empty_zero: bool,
ignore_underscores: bool
) -> Option<T> {
from_str_bytes_common(buf.as_bytes(), radix, negative,
fractional, special, exponent, empty_zero,
ignore_underscores)
}
#[cfg(test)]
mod test {
use super::*;
use option::*;
#[test]
fn from_str_ignore_underscores() {
let s : Option<u8> = from_str_common("__1__", 2, false, false, false,
ExpNone, false, true);
assert_eq!(s, Some(1u8));
let n : Option<u8> = from_str_common("__1__", 2, false, false, false,
ExpNone, false, false);
assert_eq!(n, None);
let f : Option<f32> = from_str_common("_1_._5_e_1_", 10, false, true, false,
ExpDec, false, true);
assert_eq!(f, Some(1.5e1f32));
}
#[test]
fn from_str_issue5770() {
// try to parse 0b1_1111_1111 = 511 as a u8. Caused problems
// since 255*2+1 == 255 (mod 256) so the overflow wasn't
// detected.
let n : Option<u8> = from_str_common("111111111", 2, false, false, false,
ExpNone, false, false);
assert_eq!(n, None);
}
#[test]
fn from_str_issue7588() {
let u : Option<u8> = from_str_common("1000", 10, false, false, false,
ExpNone, false, false);
assert_eq!(u, None);
let s : Option<i16> = from_str_common("80000", 10, false, false, false,
ExpNone, false, false);
assert_eq!(s, None);
let f : Option<f32> = from_str_common(
"10000000000000000000000000000000000000000", 10, false, false, false,
ExpNone, false, false);
assert_eq!(f, NumStrConv::inf())
let fe : Option<f32> = from_str_common("1e40", 10, false, false, false,
ExpDec, false, false);
assert_eq!(fe, NumStrConv::inf())
}
}
#[cfg(test)]
mod bench {
use extra::test::BenchHarness;
use rand::{XorShiftRng, Rng};
use to_str::ToStr;
use f64;
#[bench]
fn uint_to_str_rand(bh: &mut BenchHarness) {
let mut rng = XorShiftRng::new();
do bh.iter {
rng.gen::<uint>().to_str();
}
}
#[bench]
fn float_to_str_rand(bh: &mut BenchHarness) {
let mut rng = XorShiftRng::new();
do bh.iter {
f64::to_str(rng.gen());
}
}
}