How much faster is fast_float compared to strtod?

fast_float is typically 4-10 times faster, parsing at speeds up to 1 GB/s in benchmarks. It's optimized for high-throughput scenarios like database processing and web browser engines, where performance gains are critical.

Can fast_float parse integers in binary or hexadecimal?

Yes, it supports parsing integer types with different bases, including binary and hexadecimal, as shown in the README examples. You use the same from_chars function with an optional base parameter for flexible integer conversion.

How to parse numbers with a comma as decimal separator in fast_float?

Use the from_chars_advanced function with parse_options set to a custom decimal separator like comma. The README provides code snippets for this, allowing locale-dependent parsing while maintaining high performance.

Does fast_float support compile-time number parsing in C++20?

Yes, it supports constexpr evaluation in C++20, enabling compile-time parsing for scenarios like embedded systems or template metaprogramming where runtime overhead must be minimized.

fast_float or fast_double_parser: which should I use?

fast_float is generally preferred for its updated algorithm and standard C++ API, with wider adoption in projects like GCC and Chromium. fast_double_parser is older and used in specific frameworks, but fast_float offers better integration and ongoing support.

How to handle out-of-range values like 1e1000 in fast_float?

By default, it sets the value to infinity or zero and returns std::errc::result_out_of_range. You can override this by checking the error and preserving the original value, as detailed in the README's behavior section.

Open-Awesome

fast_float

Apache-2.0C++v8.2.10

A high-performance C++ library for parsing floating-point and integer numbers from strings, 4x to 10x faster than strtod.

Visit Website GitHub

2.1k stars193 forks0 contributors

What is fast_float?

fast_float is a high-performance C++ library for converting ASCII strings representing decimal numbers (like "1.3e10") into binary floating-point and integer types. It solves the problem of slow number parsing in data-intensive applications by providing an exact, locale-independent implementation that is many times faster than standard library functions.

Target Audience

C++ developers working on performance-critical systems such as databases, web browsers, scientific computing, and data processing pipelines where fast and accurate number parsing is essential.

Value Proposition

Developers choose fast_float for its exceptional speed (4x-10x faster than strtod), exact IEEE-compliant rounding, and ease of integration as a header-only library. It is widely adopted in major projects like GCC, Chromium, and MySQL, ensuring reliability and performance.

Overview

Fast and exact implementation of the C++ from_chars functions for number types: 4x to 10x faster than strtod, part of GCC 12, MySQL, DuckDB, Chromium, Redis and WebKit/Safari

Use Cases

Best For

Parsing large volumes of numeric data in databases or data pipelines
Improving number parsing performance in web browsers and JavaScript engines
High-speed logging and monitoring systems that process numeric metrics
Scientific computing applications requiring exact floating-point conversion
Embedded systems where memory allocation and performance are constrained
Implementing JSON parsers or other data serialization formats efficiently

Not Ideal For

Systems requiring hexadecimal floating-point number parsing
Applications dependent on long double precision for extended accuracy
Projects needing full locale-aware parsing with complex cultural formatting rules beyond custom decimal separators

Pros & Cons

Pros

Exceptional Parsing Speed

Benchmarks in the README show parsing at up to 1 GB/s, making it 4-10 times faster than standard strtod, essential for high-throughput data processing in projects like Chromium and Apache Arrow.

Exact IEEE Rounding

Provides round-to-even behavior for precise binary representation, adhering to C++17 specifications and ensuring correctness in scientific and financial applications.

Header-Only Integration

Easy to add to C++ projects without complex build systems; just include the header, and it supports CMake, FetchContent, or single-header amalgamation for flexibility.

Broad Type and Format Support

Handles float, double, integers (with bases like 2, 10, 16), UTF-16/32 inputs, and advanced options like JSON, Fortran notation, and custom decimal separators via parse_options.

Cons

No Long Double Support

The README explicitly states it does not support long double, limiting use in applications like scientific computing that rely on extended precision floating-point types.

Decimal-Only Parsing

Only supports decimal format; hexadecimal strings are not supported, which can be a drawback for systems parsing hexadecimal numeric data common in low-level programming.

C++-Exclusive Implementation

Primarily for C++ users; C programmers must use a separate port like ffc.h, adding complexity for mixed-language projects or legacy C codebases.

Frequently Asked Questions

Related Projects

stb

stb single-file public domain libraries for C/C++

Stars34,612

Forks8,088

Last commit1 month ago

{fmt}

A modern formatting library

Stars25,655

Forks3,045

Last commit10 hours ago

xxHash

Extremely fast non-cryptographic hash algorithm

Stars11,242

Forks914