Re: Float numbers equality.

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

2013/8/12 Andrew Starks <andrew.starks@trms.com>:

> Any library function offered would be used exclusively by people too
> inexperienced to know that they can't solve their problem through an
> abstraction provided by an opaque library.

That is very well put.

The only library function I might be prepared to support, would go
something like this:

math.near(x,y[,ballpark])

    Returns `true` if `x` can be considered to be so near to `y` that the
    difference between them could well be explained solely by the effect
    of roundoff error, given that `x` and `y` are expected to be of the
    same sort of size as the positive quantity `ballpark`, which has the
    default value 1.

I.e. the name would stress that `x` and `y` are not equal; the
documentation would be riddled with vague terms, except for
"roundoff error" appearing explicitly; and the tolerance, though
not constant, would not automatically depend on `x` or `y`.