Fix 0.6 breakages and depwarns

[timholy]

I think this gets most of the recent depwarns & breakages. Will require a new Compat tag (JuliaLang/Compat.jl#329). Also doesn't fix

ArgumentError: colons must be converted by to_indices(...)
  Stacktrace:
   [1] macro expansion at /home/tim/.julia/v0.6/AxisArrays/src/indexing.jl:47 [inlined]
   [2] reaxis at /home/tim/.julia/v0.6/AxisArrays/src/indexing.jl:18 [inlined]
   [3] getindex at /home/tim/.julia/v0.6/AxisArrays/src/indexing.jl:52 [inlined]
   [4] getindex(::AxisArrays.AxisArray{Float64,2,Array{Float64,2},Tuple{AxisArrays.Axis{:Timestamp,StepRange{DateTime,Base.Dates.Hour}},AxisArrays.Axis{:Cols,Array{Symbol,1}}}}, ::Colon, ::Symbol) at /home/tim/.julia/v0.6/AxisArrays/src/indexing.jl:76
   [5] include_from_node1(::String) at ./loading.jl:539
   [6] include(::String) at ./sysimg.jl:14
   [7] include_from_node1(::String) at ./loading.jl:539
   [8] include(::String) at ./sysimg.jl:14
   [9] eval(::Module, ::Any) at ./boot.jl:235
   [10] eval_user_input(::Any, ::Base.REPL.REPLBackend) at ./REPL.jl:66
   [11] macro expansion at ./REPL.jl:97 [inlined]
   [12] (::Base.REPL.##1#2{Base.REPL.REPLBackend})() at ./event.jl:73
ERROR: LoadError: LoadError: There was an error during testing
while loading /home/tim/.julia/v0.6/AxisArrays/test/core.jl, in expression starting on line 177
while loading /home/tim/.julia/v0.6/AxisArrays/test/runtests.jl, in expression starting on line 8

Wouldn't mind a little collaboration in fixing that particular problem 😄. If anyone is interested, feel free to grab this and run with it.

[mbauman]

I have something locally that hacks around to_indices. It does a lot of other, much crazier things, though. I'll push it as a separate PR… and you can take a look.

[timholy]

OK, I think the breakages are mostly fixed, though I noticed a couple of small issues in Base julia that needed fixing. (JuliaLang/julia#20686)

Also, I switched to Unitful for the README and tests. Most urgent because of JuliaLang/julia#20671, but I think it's also the future so we might as well switch. It passes tests locally if I have the master branch of Unitful checked out, but it doesn't on the current release. (CC @ajkeller34.)

[timholy]

OK to merge and tag? Locally, this passes on both 0.5 and 0.6 (I have master checked out on Unitful).

[mbauman]

Sure, since it's a test-only dependency that makes sense. I'd prefer to not have CI broken for too long, so if a tag isn't expected for a while I'd like to Pkg.checkout Unitful on travis.

[ajkeller34]

I hope to tag Unitful today or tomorrow, since it looks like Attobot has made that a whole lot easier.

Which reminds me, I still have that open PR on AxisArrays... sorry for the delay, I just haven't found the time. I hope to return to it soon, and am also excited to see what happens with #56 or #58.

[tkelman]

why are these here instead of base?

[mbauman]

Because this was easier than going back and reverting that portion of JuliaLang/julia#14957.

[tkelman]

what needs reverting?

[timholy]

The point of JuliaLang/julia#14957 was to centralize/simplify, but for ranges there do turn out to be cases where you want to perform the computation even when bounds-checking is forced on. Since a certain number of unsafe_getindex calls have crept back into range.jl, maybe it's not unreasonable to add them back systematically.

[mbauman]

I'd rather have unsafe_getindex be removed completely and use a new name for the range computation since it's not really unsafe (in terms of memory or validity) at all. For this use-case we really don't want this function to apply at all to non-ranges. But this really isn't a high priority for me — unsafe_getindex is an undocumented function that's rarely used these days.