iota's floating point implementation does not conform to documentation

[dlangBugzillaToGithub]

acehreli (@acehreli) reported this on 2023-01-07T02:32:50Z

Transfered from https://issues.dlang.org/show_bug.cgi?id=23604

Description

The documentation at

  https://dlang.org/library/std/range/iota.html

currently states

   "The two-argument overloads have step = 1. If begin < end
    && step < 0 or begin > end && step > 0 or begin == end,
    then an empty range is returned."

However, the implementation for floating point types does not match documentation:

import std;

void test(T)() {
    assert(iota(T(1), T(2), T(-1)).empty);
    assert(iota(T(2), T(1)).empty);        // By-default step==1
    assert(iota(T(3), T(3)).empty);        // By-default step==1
}

void main() {
    test!int;     // Produces empty ranges: GOOD
    test!double;  // asserts:               BAD
}

See the following thread for further discussion:

  https://forum.dlang.org/thread/rdbvirgshewaujtsscnp@forum.dlang.org