Updated repository to Julia 0.7+ and 1.1+

[chakravala]

This pull-request updates the CBOR repository to be compatible with Julia 0.7+ and 1.1+

This also resolves the issue discussed in #7 with @stevengj

Unfortunately, the tests are not all passing yet. Some more tweaks need to be made, but I am not familiar with what is causing the problem.

Test Failed at /home/flow/.julia/dev/CBOR/test/runtests.jl:83
  Expression: isequal(data, decode(bytes))
   Evaluated: isequal(6.103515625e-5, 4.683311023e-315)
ERROR: LoadError: There was an error during testing
in expression starting at /home/flow/.julia/dev/CBOR/test/runtests.jl:81
ERROR: Package CBOR errored during testing
Stacktrace:

Could you please review this pull-request and see if you can fix the rest of the testing?

[stevengj]

You shouldn't need to use bytes2hex or parse at all here. For example, you could just do:

reinterpret(Float64, ntoh(reinterpret(UInt64, bytes[(start_idx + 1):(start_idx + float_byte_len)])))

This could be made even more efficient (if you care) by eliminating the creation and reinterpretation of a array for the bytes slice:

reinterpret(Float64, UInt64(bytes[start_idx + 1]) << 56 +
                     UInt64(bytes[start_idx + 2]) << 48 +
                     UInt64(bytes[start_idx + 3]) << 40 +
                     UInt64(bytes[start_idx + 4]) << 32 +
                     UInt64(bytes[start_idx + 5]) << 24 +
                     UInt64(bytes[start_idx + 6]) << 16 +
                     UInt64(bytes[start_idx + 7]) << 8 +
                     UInt64(bytes[start_idx + 8]))

[stevengj]

But if instead of working with bytes arrays, you read directly from an IO stream (or an IOBuffer), you could just do:

ntoh(read(io, Float64))

(where the ntoh is to fixe the byte order).

[stevengj]

If instead of using an array of bytes, you used an IO stream (or an IOBuffer), you could just write(io, hton(float)).

By converting the value to a string, then parsing into an array of bytes, and then pushing the array of bytes to another array, you are probably at least an order of magnitude slower, as well as requiring much more code.

[chakravala]

Thanks for your comments @stevengj , I think it would be best if @saurvs could take your advice and implement it, since all I'm trying to do is to get the tests working reliably first.

Is the numerical error in the tests related to the hex2bytes conversions?

[SimonDanisch]

Oh woops, why do I never look into PRs when upgrading a package...
Well, my PR passes tests: #9

[chakravala]

Thanks for making progress on it, was hoping somebody would keep this package in order because I wanted to use it for one of my upcoming projects.

How is it coming along at the moment? I see you have not merged anything yet

[SimonDanisch]

#10 took this over