BSD-3-Clause licensed by Andrew Martin
Maintained by [email protected]
This version can be pinned in stack with:natural-arithmetic-0.2.1.0@sha256:1296ff4149f878df02d15a9b0197c6f27f311d56d478223d7e7935f030e293f0,3719
Depends on 2 packages(full list with versions):
Used by 5 packages in nightly-2024-11-30(full list with versions):

A search for terms like arithmetic and natural on hackage reveals no shortage of libraries for handling the arithmetic of natural numbers. How is this library any different some of the others? It has a particular purpose: providing a foundation on top on which other libraries may define types indexed by sizes. This uses GHC's non-inductively-defined GHC.TypeNats.Nat. As a rule, this does not use unsafeCoerce internally anywhere.

Perhaps the most direct competitor to `natural-arithmetic` is a typechecker plugin like type-nat-solver. The big difference is that `type-nat-solver` can really only be used in application code, not in library code. This is because libraries should not require the presence of typechecker plugins. Technically, they can (you could document it), but many developers will not use libraries that have unusual install procedures like this.

This library, in places, requires users to use the TypeApplications language extension. This is done when a number is only need at the type level (without a runtime witness).

This library uses a non-minimal core, providing redundant primitives in Arithmetic.Lt and Arithmetic.Lte. This is done in the interest of making it easy for user to assemble proofs. Recall that proof assembly is done by hand rather than by an SMT solver, so removing some tediousness from this is helpful to users.

This library provides left and variants variants of several functions. For example, Arithmetic.Lte provides both substituteL and substituteR. This is only done when there are two variants of a function. For substitution, this is the case because we have `b = c, a ≤ b ==> a ≤ c` and `a = c, a ≤ b ==> c ≤ b`. So, we provide both substituteL and substituteR. However, for addition of inequalities, we have four possible variants: `a ≤ b, c ≤ d ==> a + c ≤ b + d`, `a ≤ b, c ≤ d ==> c + a ≤ b + d`, `a ≤ b, c ≤ d ==> a + c ≤ d + b`, `a ≤ b, c ≤ d ==> c + a ≤ d + b`. Consequently, we only provide a single plus function, and users must use Arithmetic.Plus.commutative to further manipulate the inequality.

Here are the proof-manipulation vocabulary used by this library. Many of these terms are not standard, but we try to be consistent in this library:

  • Weaken: Increase an upper bound without changing the bounded value

  • Increment: Increase an upper bound along with the bounded value

  • Decrement: Decrease an upper bound along with the bounded value

  • Substitute: Replace a number with an equal number

Changes

Revision history for natural-arithmetic

0.2.1.0 – 2024-02-02

  • Add fromInt and fromInt# to Arithmetic.Fin.
  • Update package metadata.

0.2.0.0 – 2024-01-09

  • Change the types of with# and construct# (both in Arithmetic.Fin) to use an unboxed “less than” instead of a boxed one. This is a breaking change.
  • Add patterns for the natural numbers 5, 6, 7.
  • Add a lot of primitives for working with unboxed natural and inequalities. GHC is unable to eliminate the boxed Fin type in a suprisingly large number of cases, and Fin# helps a lot with this.

0.1.4.0 – 2023-05-31

  • Add unboxed Nat type
  • Add nominal role for Fin# type constructor. Technically, this is a breaking change, but if anyone was using coerce on a Fin#, they were already in a bunch of trouble. So, there is not going to be a major version bump for this.

0.1.3.0 – 2022-05-23

  • Add strict variant of descend.
  • Add unboxed Fin type

0.1.2.0 – 2019-01-20

  • Add strict variant of descend.

0.1.1.0 – 2019-11-22

  • Undocumented

0.1.0.0 – 2019-09-04

  • Initial release.