getindex for infinite range

[pwl]

It seems to me that this should work

r=1:Inf
r[2]   # should give 2.0

but I get (in 0.5-dev)

julia> r[2]
ERROR: InexactError()
 in indices at ./abstractarray.jl:45 [inlined]
 in checkbounds at ./abstractarray.jl:180 [inlined]
 in checkbounds at ./abstractarray.jl:194 [inlined]
 in getindex(::FloatRange{Float64}, ::Int64) at ./range.jl:461
 in eval(::Module, ::Any) at ./boot.jl:234
 in macro expansion at ./REPL.jl:92 [inlined]
 in (::Base.REPL.##1#2{Base.REPL.REPLBackend})() at ./event.jl:46

Is this intended, or is it just a bug? From the looks of it the error comes from somewhere in checkbounds, as the value of r[i] is simply (r.start + (i-1)*r.step)/r.divisor and it can be computed by hand without any errors.

[pwl]

Also, shouldn't pop! and similar functions be defined for Range? You could just modify the upper/lower bounds when removing elements from both ends.

[yuyichao]

shouldn't pop! and similar functions be defined for Range

No, ranges are immutable.

[pwl]

No, ranges are immutable.

It makes sense now, I didn't know that.

[JeffBezanson]

I don't think Ranges are designed to support infinite length. I would say 1:Inf and 1:NaN should be errors.

[tkelman]

we have countfrom for this, though it doesn't implement indexing

[pwl]

@JeffBezanson is that because 1:Inf has undefined length? Is there any indexible alternative to an infinite list? @tkelman mentioned countfrom, but it isn't indexible, same with repeated. I think having `getindex for these wouldn't hurt.

[mbauman]

I would say 1:Inf and 1:NaN should be errors.

Both now throw InexactErrors at construction time. Adding indexing to countfrom and repeated could be considered separately — see #15988 for an initial start.

[tkelman]

should add tests accordingly