| bellerophon.rs |
An implementation of Clinger's Bellerophon algorithm.
This is a moderate path algorithm that uses an extended-precision
float, represented in 80 bits, by calculating the bits of slop
and determining if those bits could prevent unambiguous rounding.
This algorithm requires less static storage than the Lemire algorithm,
and has decent performance, and is therefore used when non-decimal,
non-power-of-two strings need to be parsed. Clinger's algorithm
is described in depth in "How to Read Floating Point Numbers Accurately.",
available online [here](http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.45.4152&rep=rep1&type=pdf).
This implementation is loosely based off the Golang implementation,
found [here](https://github.com/golang/go/blob/b10849fbb97a2244c086991b4623ae9f32c212d0/src/strconv/extfloat.go).
This code is therefore subject to a 3-clause BSD license. |
13650 |
- |
| bigint.rs |
A simple big-integer type for slow path algorithms.
This includes minimal stackvector for use in big-integer arithmetic. |
24688 |
- |
| extended_float.rs |
|
633 |
- |
| fpu.rs |
Platform-specific, assembly instructions to avoid
intermediate rounding on architectures with FPUs.
This is adapted from the implementation in the Rust core library,
the original implementation can be [here](https://github.com/rust-lang/rust/blob/master/library/core/src/num/dec2flt/fpu.rs).
It is therefore also subject to a Apache2.0/MIT license. |
3890 |
- |
| heapvec.rs |
Simple heap-allocated vector. |
4568 |
- |
| lemire.rs |
Implementation of the Eisel-Lemire algorithm.
This is adapted from [fast-float-rust](https://github.com/aldanor/fast-float-rust),
a port of [fast_float](https://github.com/fastfloat/fast_float) to Rust. |
9131 |
- |
| lib.rs |
Fast, minimal float-parsing algorithm.
minimal-lexical has a simple, high-level API with a single
exported function: [`parse_float`].
[`parse_float`] expects a forward iterator for the integer
and fraction digits, as well as a parsed exponent as an [`i32`].
For more examples, please see [simple-example](https://github.com/Alexhuszagh/minimal-lexical/blob/master/examples/simple.rs).
EXAMPLES
--------
```
extern crate minimal_lexical;
// Let's say we want to parse "1.2345".
// First, we need an external parser to extract the integer digits ("1"),
// the fraction digits ("2345"), and then parse the exponent to a 32-bit
// integer (0).
// Warning:
// --------
// Please note that leading zeros must be trimmed from the integer,
// and trailing zeros must be trimmed from the fraction. This cannot
// be handled by minimal-lexical, since we accept iterators.
let integer = b"1";
let fraction = b"2345";
let float: f64 = minimal_lexical::parse_float(integer.iter(), fraction.iter(), 0);
println!("float={:?}", float); // 1.235
```
[`parse_float`]: fn.parse_float.html
[`i32`]: https://doc.rust-lang.org/stable/std/primitive.i32.html |
1990 |
- |
| libm.rs |
A small number of math routines for floats and doubles.
These are adapted from libm, a port of musl libc's libm to Rust.
libm can be found online [here](https://github.com/rust-lang/libm),
and is similarly licensed under an Apache2.0/MIT license |
38671 |
- |
| mask.rs |
Utilities to generate bitmasks. |
1261 |
- |
| num.rs |
Utilities for Rust numbers. |
10249 |
- |
| number.rs |
Representation of a float as the significant digits and exponent.
This is adapted from [fast-float-rust](https://github.com/aldanor/fast-float-rust),
a port of [fast_float](https://github.com/fastfloat/fast_float) to Rust. |
3723 |
- |
| parse.rs |
Parse byte iterators to float. |
6618 |
- |
| rounding.rs |
Defines rounding schemes for floating-point numbers. |
4286 |
- |
| slow.rs |
Slow, fallback cases where we cannot unambiguously round a float.
This occurs when we cannot determine the exact representation using
both the fast path (native) cases nor the Lemire/Bellerophon algorithms,
and therefore must fallback to a slow, arbitrary-precision representation. |
14611 |
- |
| stackvec.rs |
Simple stack-allocated vector. |
8878 |
- |
| table.rs |
Pre-computed tables for parsing float strings. |
308 |
- |
| table_bellerophon.rs |
Cached exponents for basen values with 80-bit extended floats.
Exact versions of base**n as an extended-precision float, with both
large and small powers. Use the large powers to minimize the amount
of compounded error. This is used in the Bellerophon algorithm.
These values were calculated using Python, using the arbitrary-precision
integer to calculate exact extended-representation of each value.
These values are all normalized.
DO NOT MODIFY: Generated by `etc/bellerophon_table.py` |
4049 |
- |
| table_lemire.rs |
Pre-computed tables powers-of-5 for extended-precision representations.
These tables enable fast scaling of the significant digits
of a float to the decimal exponent, with minimal rounding
errors, in a 128 or 192-bit representation.
DO NOT MODIFY: Generated by `etc/lemire_table.py`
This adapted from the Rust implementation, based on the fast-float-rust
implementation, and is similarly subject to an Apache2.0/MIT license. |
36927 |
- |
| table_small.rs |
Pre-computed small tables for parsing decimal strings. |
2127 |
- |