move DirectIndexString from Base

[bkamins]

PR following JuliaLang/julia#24259

[nalimilan]

I think you'll have to define this earlier, since DirectIndexString is used above. You'll be able to check that once you'll have a version of Julia without it.

[bkamins]

moved

[nalimilan]

Looks good to me, the code is tested via ASCIIString. Though we should probably wait until JuliaLang/julia#24259 has been merged to check that tests pass without DirectIndexString in Base.

EDIT: the problem is, tests currently fail on 0.7...

[ararslan]

Looks good to me as well. I assume the UTF16 BoundsError is preexisting.

[bkamins]

Yes - the error seems to be line 763 of abstractarray.jl which is called with a number.

[ararslan]

Are there tests for DirectIndexString in Base that could be copied over?

[bkamins]

I have moved everything from Base that referenced DirectIndexString.
No tests used DirectIndexString as no type in Base was subtype of DirectIndexString.

This part of code:

# isvalid(), chr2ind() and ind2chr() for SubString{DirectIndexString}
let ss, s="lorem ipsum",
    sdict=Dict(SubString(s,1,11)=>s,
               SubString(s,1,6)=>"lorem ",
               SubString(s,1,0)=>"",
               SubString(s,2,4)=>"ore",
               SubString(s,2,11)=>"orem ipsum",
               SubString(s,15,14)=>""
               )
    for (ss,s) in sdict
        local ss
        for i in -1:12
            @test isvalid(ss,i)==isvalid(s,i)
        end
    end
    for (ss,s) in sdict
        local ss
        for i in 1:length(ss)
            @test ind2chr(ss,i)==ind2chr(s,i)
        end
    end
    for (ss,s) in sdict
        local ss
        for i in 1:length(ss)
            @test chr2ind(ss,i)==chr2ind(s,i)
        end
    end
end #let

seems to have been testing DirectIndexString in the past but now it tests String containing only ASCII (I have corrected the description in JuliaLang/julia#24262).

[ararslan]

We should try to get this merged and tagged ASAP to avoid a gap in availability on 0.7. Any ideas about the 0.7 failures, @nalimilan?

[nalimilan]

Honestly I don't think anybody cares much about DirectIndexString... :-)

I've traced the failure to ReinterpretArray, see JuliaLang/julia#23750 (comment). The problematic call is convert(UTF16String, Int16[[0x65, 0x66] [0x67, 0x68]]), which I really wonder how somebody thought it was worth supporting... The next level of absurdity will be to allow constructing strings from BitArray{3}.

[nalimilan]

I have a fix at #25.

[stevengj]

ASCIIString and UTF32String are both subtypes of DirectIndexString and get methods like nextind and length from DirectIndexString. So, the tests for ASCIIString and UTF32String are implicitly testing DirectIndexString. And people do care about those types.