Implementation of high/low for all uint types

[bluenote10]

Since uint and uint64 are not Ordinal, it is currently not possible to use high/low on them in contrast to uint8/uint16/uint32. This is unfortunate when writing generic algorithms which make use of high/low and should work on all number types. As a workaround it is currently necessary to bring some manual implementations into scope like:

proc low*[T: uint|uint64](x: typedesc[T]): T =
  cast[T](0)

proc high*[T: uint|uint64](x: typedesc[T]): T =
  cast[T](-1)

For consistency it would be nice to have them by default. Note: I'm not trying to make them Ordinals, but just because they don't have all properties of an Ordinal (I assume overflow checks), the types still have a smallest/largest value.

Also there is currently an inconsistency in high and low, because high is defined for high[T: Ordinal](x: T) while low uses no type selector, i.e., low[T](x: T). As a result, the following code

var x: uint
echo low(x)

errors with Error: invalid argument for 'low' (a local error from semLowHigh in semexprs.nim when handling the magic), while trying to use high gives the clearer type mismatch.

If desired, I can work on that.

[markus-oberhumer]

Actually I'd rather like to understand why uint and uint64 are not Ordinal - system.nim vaguely talks about implementation reasons.

[markus-oberhumer]

Here is the version I'm currently using in my local system.nim.

Note that I do not use typeclasses (issue #6840) and there still is a codegen bug (see #6839).

template low* (_: uint): uint = 0'u
template high*(_: uint): uint = not 0'u
template low* (_: uint64): uint64 = 0'u64
template high*(_: uint64): uint64 = not 0'u64
template low* (_: typedesc[uint]): uint = 0'u
template high*(_: typedesc[uint]): uint = not 0'u
template low* (_: typedesc[uint64]): uint64 = 0'u64
template high*(_: typedesc[uint64]): uint64 = not 0'u64

[kaushalmodi]

Getting high(uint) and low(uint) will probably help fix #11551 and #11552.

[kaushalmodi]

/cc @Araq Referencing your comment back in 2013 ( https://forum.nim-lang.org/t/313#1630 ), does that still hold true?

[Araq]

This has finally been implemented.