Arbitrary-precision binary, decimal, and ball arithmetic for MoonBit, with checked operations and explicit numeric semantics.

floating-point

arbitrary-precision

decimal

interval

ieee-754

Download zip

Author

Luna-Flow

Version

0.9.0

License

Apache-2.0

Repository

github.com/Luna-Flow/floating

Last updated

5 hours ago

Downloads

101

Dependencies

#floating

Luna-Flow/floating provides arbitrary-precision binary, IEEE decimal, General Decimal Arithmetic, and certified interval arithmetic for MoonBit. Binary values follow IEEE 754 at any precision, decimals keep their quantum, and intervals are rounded outward so that they always contain the exact result. Precision, rounding, exponent range, special values, status flags, traps and enclosures are explicit values in every API rather than hidden global state.

#Install

moon add Luna-Flow/floating@0.9.0

The module needs the MoonBit toolchain 0.10 or later (moonc ≥ 0.10). Import only the packages you use in your moon.pkg:

import {
  "Luna-Flow/floating/bin_float",
  "Luna-Flow/floating/decimal",
  "Luna-Flow/floating/ball_float",
}

#Quick start

///|
test "floating quick start" {
  // One third, correctly rounded to binary64, with its IEEE status.
  let ctx = @bin_float.BinaryContext::binary64()
  let (third, flags) = @bin_float.BinFloat::from_int(1).div_ctx(
    @bin_float.BinFloat::from_int(3),
    ctx,
  )
  inspect(third.to_shortest_string(), content="0.3333333333333333")
  inspect(flags.inexact(), content="true")

  // A decimal keeps the quantum of its literal.
  let (price, _) = @decimal.Decimal::from_string_ctx(
    "12.3400",
    @decimal.DecimalContext::decimal64(),
  )
  inspect(price.quantum(), content="-4")

  // An interval contains every exact result.
  let one = @ball_float.BallFloat::from_int(1, precision=53)
  let enclosure = one.div(@ball_float.BallFloat::from_int(3, precision=53))
  inspect(enclosure.contains(third), content="true")
}

The three results have different contracts: third is one rounded point with the flags its rounding raised, price retains the quantum -4 of 12.3400, and enclosure is a set of reals guaranteed to contain .

#Packages

Package	Purpose	Result model
bin_float	arbitrary-precision binary floating point, IEEE 754 binary operations, binary16/32/64/128 interchange	value, or (value, BinaryFlags) under a BinaryContext
decimal	IEEE 754 decimal arithmetic, decimal32/64/128 DPD and BID interchange	value, or (value, DecimalFlags) under a DecimalContext
decimal_gda	General Decimal Arithmetic with sticky status and traps	GdaOutcome with the defined result and the next context
ball_float	outward-rounded bare and decorated intervals (IEEE 1788)	interval, or (interval, BallFlags) under a BallContext
bin_float_checked	binary pipeline that stops at the first error	Result[BinFloat, ArithmeticError] in a wrapper
decimal_checked	IEEE decimal pipeline that accumulates flags	value with latest and accumulated flags
decimal_gda_checked	GDA pipeline that stops at a trap	one threaded GdaOutcome
ball_float_checked	interval pipeline that stops at the first error	Result[BallFloat, ArithmeticError] in a wrapper
def	shared vocabulary: Sign, PartialOrder, the Floating trait	—
semantic	exact projection for comparing values across packages	exact rational, signed infinity or NaN

Expression, corpus-frontend, CLI, benchmark, consistency and internal/* packages are repository infrastructure. The package map lists all of them with their tutorial, API and design pages.

#Documentation

The manual is published at https://lunaflow.cn/en/floating/ in English, Chinese and Japanese; its English source is in doc/manual. Good starting points:

Getting started: choosing a package and first programs.
Numeric semantics: rounding, ulp, flags, quantum, signed zero, NaN and enclosures.
Architecture: layers, the numeric core pipeline and certified elementary functions.
Verification: gates and the scope of every conformance claim.

Conformance evidence is finite and pinned: the GDA official corpus passes 64,986/64,986 legal executable rows, the binary TestFloat level-1 matrix 254,227,872 vectors, and the strict ITF1788 aggregate 4,656/4,656 cases. Read the conformance pages before turning a result into a compatibility claim.

#Development

The repository uses just as its task runner.

just pr 8                    # pull-request gate
just fmt                     # format MoonBit sources
just docs                    # manual checks and documentation examples
just gate binary 8           # TestFloat and MPFR
just gate decimal 8          # IEEE decimal vectors
just gate decimal_gda 8      # GDA decTest corpora
just gate interval 8         # strict ITF1788
just ci 8                    # everything, before a release

Smoke fixtures, plans, pinned corpora, targets and phases go through one entry point:

just conformance smoke binary
just conformance run decimal --run-target native --run-target wasm
just conformance run interval --phase trigonometric --strict-supported

Corpus provenance and options are described in testdata/bin_float, testdata/decimal and testdata/interval.

#Contributing

Contributions to correctness, documentation and test coverage are welcome. Read CONTRIBUTING for the workflow and the repository conventions for documentation and API snapshot rules. Release history is in CHANGELOG.

#License

Apache-2.0. See LICENSE.

Site source Report issue Packages Build queue Skills Statistics