Support the flag ' for thousand separator in Printf, issue #29077

[ki-chi]

Closes #29077.
The flag ' is valid only for the decimal notation, i.e., the following formats: %d, %u, %i, %f, %F, %g, and %G.

julia> using Printf

julia> @sprintf "%'d" 1000000
"1,000,000"

julia> @sprintf "%'.2f" 1000000
"1,000,000.00"

julia> @sprintf "%-'10d" 1000
"1,000     "

julia> @sprintf "%'10d" 1000
"     1,000"

[KristofferC]

@quinnj, maybe you could take a look at this?

[quinnj]

This overall looks like useful functionality. I've left a few comments about improving the performance, which we want to be sensitive too. i.e. for formats not using the apostrophe, we'd ideally have zero impact from this net-new code. I think the BigInt trunc is the only really tricky part; we need to find another, more efficient way to get that information.

[ki-chi]

@quinnj Thanks for the review. I will try to fix the code, and you are right about the BigInt issue; I need to think about it more carefully.

[nathanrboyer]

Now that #40105 is released in Julia 1.10, I want this functionality even more (to eliminate my need for the Formatting.jl package). Maybe @johanmon, having worked on this function recently, has some ideas to avoid the performance issues above?

[GregPlowman]

It seems this PR has stalled because of some performance concerns.

I think the BigInt trunc is the only really tricky part; we need to find another, more efficient way to get that information.

Would something like the following work as an alternative to the BigInt version?

function countthousandsep(f::Spec{T}, x) where {T <: Union{Ints, Floats}}
    (isnan(x) || isinf(x) || base(T) != 10) && return 0
    x < 0 && return countthousandsep(f, -x)
    x2 = round(x, digits=f.precision)
    x2 < 1000 && return 0
    numdig = log10(x2)
    numsep = Int(div(numdig, 3))
    return numsep
end

I think the other issues raised were using view to avoid copying, which seems simple enough.

@ki-chi, are you still willing to work on this?

[ki-chi]

@GregPlowman,

Sorry for leaving that pull request hanging. I need more time to work on this PR now. But I always welcome any suggestion to resolve the performance issue above.

[GregPlowman]

@ki-chi, is this PR finished on your end? If you're happy with everything, should you the mark the requested changes as addressed?

@quinnj, would you be willing to continue your review?