Big-Float constants `ee` and `pi` don't change to match precision. #1013

Izaakwltn · 2023-10-11T17:38:31Z

These variables are locked in with the initial big-float precision, and don't adjust as precision changes.

Initial precision is 53 bits so they're basically locked in as Double-Floats.

stylewarning · 2023-10-11T20:49:03Z

note: this is only an issue with the MPFR version of the library. The portable version handles this fine.

Izaakwltn · 2023-10-13T19:06:51Z

I've been thinking about this bug, and thought of a couple quick but potentially janky approaches for fixing it.

Every time precision is set, redefine Trigonometric Big-Float and Exponentiable Big-Float, which will automatically update the definition of pi and ee respectively. This could be kept in separate functions like reset-bf-constants or set-precision-and-constants. Bubble-gum and paperclip approach.
- Note: Is it possible to redefine one member of a type class or does every redefinition have to be applied as a whole?
Convert Big-Float to MPFR-Float and make a new Big-Float type in the style of impl-default.lisp that uses MPFR-Floats instead of Dyadics. This is probably the nuclear option.
Simplifying the possibility above, give Big-Float two constructors that contain either an MPFR-Float object or a (Unit -> MPFR-Float) function:

(define-type Big-Float
  (BFValue MPFR-Float)
  (BFConst (Unit -> MPFR-Float))

stylewarning · 2023-10-14T03:38:21Z

Ignoring potential language changes, the last suggestion would be the right approach.

Izaakwltn added the bug label Oct 11, 2023

stylewarning added the standard library label Oct 11, 2023

Izaakwltn mentioned this issue Oct 18, 2023

Adding Big-Float constructors for variable prec constants #1024

Closed

stylewarning removed the bug label Apr 24, 2025

Provide feedback