fastnum

Fixed

• Incorrect division result for numbers with large amount of decimal digits #49.
• Fix no_std build #50.

Changed

• Improve debug formatting of Decimal and UnsignedDecimal.

Examples

• Add no_std wasm example #50.

Fixed

• .ln() panic regression (0.2.10 -> 0.7.0) #46.

Changed

• Improve log implementation #38.

Documentation

• Documentation for Decimal::with_rounding_mode and Decimal::with_ctx:
  • Clarified that when changing the rounding mode, any existing extra precision (if present) is immediately rounded
    using the new rule.
  • Clarified that applying a new context may trigger traps if the current value already has signaling flags set (
    e.g., INEXACT) and the new context enables traps for those signals; this may result in a panic depending on
    build/configuration.
• Minor fixes.

Added

• Big-integer conversion APIs #44:

  • Parsing from raw bytes:

    • from_radix_be(buf, radix) -> Option<Self>
    • from_radix_le(buf, radix) -> Option<Self>

  • Encoding to raw bytes:

    • to_radix_be(radix) -> Vec<u8>
    • to_radix_le(radix) -> Vec<u8>

  • String conversion helpers:

    • parse_str_radix(s, radix) -> Self (panicking variant)
    • parse_bytes(buf, radix) -> Option<Self>
    • to_str_radix(radix) -> String

Documentation

• Conversion documentation expanded with clearer semantics, base handling, and examples.
• Improved doc comments for:
  • Base-aware parsing behavior and panic conditions.
  • Byte-level import/export helpers.
  • parse_str(s) and from_str(s) for big integers recognize 0x and 0b prefixes (decimal remains default without a prefix).

Internal

• Conversion code reorganized into focused modules:
  • bint/convert/{from_bytes,to_bytes,from_str,to_str}.rs
• Macro routing for conversion implementation is simplified and deduplicated.

Changed

• Minor internal performance improvements.

Changed

• Minor performance improvements.

Breaking changes

• Removed previously deprecated Decimal methods:
  • from_scale (use quantum instead).
  • normalized (use reduce instead).
  • with_scale (use rescale instead).

Added

• Decimal truncation API (#39):
  • Decimal::trunc() — truncates to integral with no fractional portion without rounding.
  • Decimal::trunc_with_scale(scale) — truncates to the specified scale without rounding.

Changed

• Internal documentation macro routing for decimal type aliases refined (no public API impact).

Documentation

• Minor fixes.
• Added a dedicated “Truncate” section with behavior details and examples.

Internal

• Introduced an internal truncation implementation integrated with scaling and extra-precision handling to ensure true
  truncation semantics (no rounding).

Added

• Type conversion traits for integers and decimals:
  • Cast — type-safe, infallible conversion (e.g., widening or lossless transforms).
  • TryCast — checked conversion returning an error on overflow, sign mismatch, or incompatible scale.
  • Supported families: unsigned/signed big integers (U64, U128, U256, U512, U1024, and I64, I128,
    I256, I512, I1024) #42 and decimals (UD64, UD128, UD256, UD512, UD1024, and D64, D128, D256,
    D512, D1024).
  • Common scenarios:
    • Widening cast for integers — via Cast.
    • Checked narrowing cast — via TryCast, with overflow errors.
    • Conversions between unsigned and signed integers — TryCast rejects negative/out-of-range values.
    • Conversions between UD* and D* — proper sign/scale propagation with checks in TryCast.
• Const-generic “dimensions” for conversion direction checks:
  • Internal Widen/Narrow markers ensure Cast/TryCast correctness without unstable generic const expressions.

Deprecated

• Decimal .transmute() is now deprecated (since "0.5.0"); removal is planned for a future major release.
  Use .resize() or cast via Cast and TryCast traits instead.

Documentation

• Minor fixes and clarifications; added usage examples for type conversions.

Changed

• Internal performance optimizations.

Changed

This release is mostly focused on fix: #34.

Converting decimal numbers to floating point IEEE 754 was completely redesigned. Implementation is based on LLVM’s libc experience: https://llvm.org/devmtg/2022-11/slides/QuickTalk3-ApproximatingatScale-StringToFloat.pdf Algorithms:

• First pass: Clinger’s Fast Path: Clinger WD. How to Read Floating Point Numbers Accurately.
• Second Pass: Eisel-Lemire fast_float: Number Parsing at a Gigabyte per Second and Noble Mushtak, Daniel Lemire. Fast number parsing without fallback.
• Third Pass: Simple Decimal Conversion Nigel Tao’s description of the Simple Decimal Conversion algorithm.

Documentation

• Add decimal -> f64 conversion benchmarks.

Changed

• Bump Rust version to 1.87 (stabilize #![feature(integer_sign_cast)]).
• Replace big integer's implementation from bnum type aliasing to wrapped types.
• Some minor performance optimizations.

Fixed

• Fix incorrect D64 to_f64 implementation: #29.

Fixed

• Fix inaccurate pow: #28.

Added

• Add impl From<UnsignedDecimal> for Decimal convertion trait implementation.

Fixed

• Fix display rounding error: #24.

Fixed

• Incorrect cmp implementation: #22.

Fixed

• The comparison function does not correctly implement a total order: #21.

Added

• Examples for compile time environment variables.

Fixed

• Failing to converge the Brent iterative method for calculating reciprocals (infinity loop): #20.

Added

• Add public constant functions for convert Decimal and UnsignedDecimal from/to Rust's primitive numeric types: #15.
• Introduce .transmute() operation for N-to-M bit unsigned decimal conversion: #18.

Fixed

• Internal conversion to NBase fails for some numbers: #16.

Changed

• Some float2decimal performance optimizations.

Fixed

• Minor fixes in float2decimal conversion.
• Minor fixes with extra precision digits.

Fixed

• Incorrect .ceil() and floor() behavior for negative values.

This release primarily focuses on:

• Increasing the accuracy of approximated mathematical calculations due to the use of 7 extra precision digits instead of 4.
• Performance optimizations.

Expect better performance, precision handling, and expanded documentation covering new methods and examples.

• Bump Rust version to 1.83
  • Stabilize &mut, *mut, &Cell, and *const Cell in const.

Added

• floor() and ceil() const methods.
• Add basic tests for trivial trigonometric functions for 0–360 degree angles.

Changed

• Control block is completely refactored for better performance and more compact memory layout.
• Micro-optimizations in rounding.
• Inline optimizations and better precision handling for mathematical rounding.
• Remove custom const float helpers as they're stabilized as const in rust 1.83.

Fixed

• Incorrect floor() and ceil() methods in num_traits Float implementation.

Documentation

• Minor fixes.

Added

• Add tokio-postgres feature which enables serialization and deserialization of fastnum decimals for tokio-postgres crate #8.
• Add tokio-postgres and sqlx_postgres examples.

Documentation

• Minor fixes.

Fixed

• Incorrect to_f64 impl on Decimal: #7.

Documentation

• Minor fixes.

Fixed

• Fixed compilation warnings

This release primarily focuses on:

• Implementing a full range of advanced mathematical functions (exponential, roots, power, logarithmic, and trigonometric functions) for working with exact precision decimal numbers.
• Increasing the accuracy of approximated mathematical calculations due to the use of extra precision digits and the absence of intermediate rounding.
• Fixing and improving conversions between Float and Decimal types.

Expect better performance, precision handling, and expanded documentation covering new methods, constants, and examples. Highlights include removing libm dependency and major refinements for no-std environments.

Added

• Implement exponential exp and binary exponential exp2 functions for decimals.
• Implement logarithmic ln, log10, log2, log functions for decimals.
• Implement sqrt, cbrt, and nth_root roots functions for decimals.
• Refactor and extend pow to support non-integer exponents.
• Implement base trigonometric functions: sin, cos, tan, sin_cos for decimals.
• Implement inverse trigonometric functions: asin, acos, atan, atan2 for decimals.
• Implement hyperbolic functions: sinh, cosh, tanh for decimals.
• Implement inverse hyperbolic functions: asinh, acosh, atanh for decimals.
• Implement hypotenuse calculation hypot function for decimals.
• Introduce fused multiply-add mul_add operation without intermediate rounding.
• Introduce transmute operation for N-to-M bit decimal conversion.
• Add integer and unsigned integer conversion utilities (from_int, from_uint).
• Implement num_traits::float::Float trait.

Changed

• Improved arithmetic precision of approximated mathematical calculations due to the use of extra precision digits and the absence of intermediate rounding.
• Overhauled mathematical constants. Add INEXACT flag and extra precision digits.
• Mark some methods as inexact by design.
• Micro-optimizations in rounding.
• Inline optimizations and better precision handling for mathematical rounding.
• Re-implement recip method without division.
• Improve test coverage for mathematical operations (e.g., sqrt, ln, exp).
• Remove libm dependency for no-std environment.

Fixed

• Corrected edge cases for float (f32/f64) from/to Decimal conversions (fix #5).
• Fix long rounding issues with improved context precision.
• General code quality improvements and bug fixes, including better inexact flag handling.

Documentation

• Expanded documentation for all major operations with examples.
• Added comprehensive examples and descriptions to LIB.md.
• Minor fixes.

Breaking changes

• Replace decimal::Category with core::num::FpCategory.

Changed

• Micro-optimizations in rounding.
• Make .from_parts() constructor public (Way to directly create decimals with scale without dividing? #3).

Added

• Implement TryFrom<Decimal> for UnsignedDecimal and From<UnsignedDecimal> for Decimal traits.

Fixed

• Fixed minor issues with ceil/floor rounding.

Fixed

• Fixed a performance issue with parsing and rescaling during arithmetic operations.

Changed

• Replace ilog10 method for integers with faster algorithm.
• Replace x 10 multiplication with a faster algorithm (const static lookup table).
• Some performance improvements.

Fixed

• Division by divisor with non-scalable coefficient.

Changed

• More strict #[repr] for Decimal type.
• Documentation improvements.

Added

• Implement num_traits::float::FloatCore trait.
• Add MIN_POSITIVE and EPSILON constants.
• Add .powi(), .recip(), .to_degrees(), .to_radians() methods.

Changed

• Documentation improvements.

Added

• Add basic mathematical constants (PI, E, etc.).
• Implement num_traits::FloatConst trait.

Fixed

• cargo test failing two tests on macOS #2

The main goals of this release are:

• Stabilize API for more stringent compliance with the recommendations of IEEE 754 and IEEE 854 standards.
• Refactor decimal module and simplify context usage.

Breaking changes

• Remove [Context] from a most methods argument list and put it into decimal control block for better context handling. Now [Context] is a property of any Decimal number instance.

Added

• Extend test coverage.

Changed

• Documentation improvements.
• Deprecate and replace some outdated methods to align with the updated structure and functionality.
• Benchmarks were modularized into separate operations, enhancing clarity and maintainability.

Added

• Extend test coverage.

Fixed

• sqlx support for PostgreSQL.

Changed

• Minor documentation fixes.
• Remove const_str dependency.

Added

• Added signals! macro.
• Extend test coverage.

Changed

• Deprecated with_scale! in favor of quantum!.
• Documentation improvements.

• Minor fixes.

Breaking changes

• Internal decimal representation for signed and unsigned decimal types.
• Reduce exponent from 64 bit to 16 bit.
• Remove DecimalResult and unwrap methods.
• Replace ArithmeticPolicy, OverflowPolicy, RoundingPolicy with Context.

Added

• Add NaN and ±Infinity special values.
• Add Flags and Signal.
• Add Context and SignalingTraps.
• Add with_context! macro.

Fixed

• Fix panic shift with overflow.

Changed

• Stabilize API.
• Re-implement most methods.
• Documentation improvements.

• Fix rounding subtraction.
• Extend test coverage for numtraits feature.
• Documentation improvements.

• Stabilize API

• Stabilize API

• Bump utoipa to 5.0.x version

• Stabilize API
• Improve docs

• Stabilize API

• Technical releases

• Initial commit