Documentation of utf32 incorrectly states that an Array of UInt32 is not copied by UTF32String

[ScottPJones]

While it is true that an Array of Char is not copied, the same as an Array of UInt8 for UTF8String, or of UInt16 for UTF16String, UTF32String copies the array, if it is an Array of UInt32.
Either the documentation should point out this inconsistency, or else the code needs to change to have a constructor that does a reinterpret(Char, array) so as not to copy (which is my vote).
Either way, the documentation should be changed (to say that Array of UInt32 is copied, and Char not, or with the code changed, that both are not copied...)
See https://groups.google.com/forum/#!topic/julia-dev/ZBSelpTlw58

[stevengj]

We should just add the missing constructor that does not do the copy:

UTF32String(data::Vector{UInt32}) = UTF32String(reinterpret(Char, data))

(The convert functions should continue to make a copy.)

[ScottPJones]

That's what I had thought... but I am thinking that this is rather dangerous... you basically have mutable strings, whether you like it or not... It might be nicer if there were some way of having an "unsafe" constructor, that could be used by the internals (or where people felt sure they knew what they are doing). For now I'll just add the missing constructor.