Re: The border crossing - integers in a ghetto

[Dirk Laurie <dirk.laurie@...>]

2018-07-28 13:14 GMT+02:00 Gavin Wraith <gavin@wra1th.plus.com>:
> In message <CABcj=tniVP-0vhACzCv7XtinecPkPPkj2w6rpMx=qYwRq2hkTg@mail.gmail.com>
>           Dirk Laurie <dirk.laurie@gmail.com> wrote:
>
>>If you only intend to use those builds for programs you wrote
>>completely by yourself, it may well be a good thing.
>
> Myself, and just possibly a handful of other users of RISC OS.
>
>>No more 2*math.pi, it's got to be 2.0*math.pi.  Not even C demands that.
>
> By way of mitigation, I have to say that I am a mathematician, a
> category theorist. So the idea of testing the equality of things of
> different type goes against my faith. Categorification is much
> in vogue nowadays. Please take cum grano salis :)-.

Well, now! Before posting, I edited out as too abstruse a section
which referred to Dedekind cuts, Peano's axioms and Bourbaki
[1] as examples of the sort of thing we do _not_ need in Lua. But
since you seem to be fluent in Haskell, isn't that a better space
for such notions?

>  I grant you that floating point formats are convenient for engineers.
>  The dominance of IEEE floating point formats makes the general public
>  forget that there are other possible formats, which might be useful
>  for specific purposes. Interval arithmetic, lazy streams of integral
>  unimodular 2x2 matrices, finite fields, ... . But if I was into
>  number crunching I would not be using Lua and a Raspberry Pi.

Actually just LUA_NOCVTS2N à la Lua 5.4 is enough to get you
any of that, by replacing the string metamethods.

[1] Was it Peter Henrici who referred to Wirth's post-Pascal
languages as "the Bourbakification of computer science"?