RFC: Updates to IO, IOString

[kmsquire]

In the original versions of read(s::GZipStream, ::Type{Char}) and write(s::GZipStream, Char) in #1355, Jeff suggested generalizing the functions so they work with all IO streams.

Edit: removed error message, fixed by Jameson's comment below.

Edit: The read and write methods above work, but don't quite go far enough. So, additionally, this patch

• rearraged io.jl, so that all functions operating on abstract IO types are collected together. IOStream is located below. This makes it easier to see what needs to be implemented for IO subtypes
• promotes each_line/EachLine to work with generic IO types
• adds seek_end, position, truncate, eof, and close methods to IOString, to make it a (nearly) fully functional IO class
• adds a bytestring method to IOString, so that its contents can easily be converted to a string

The bytestring(io::IOString) method isn't (yet) a replacement for memio(), since it copies the data. This is actually necessary, currently, because utf8 string written to IOString are converted to disk format (cf. the original read/write functions above). It may be worthwhile NOT to do so for IOString.

[vtjnash]

on line 271 error("write(IO, Char): illegal character $(uint(c))"), the $ interpolation macro hasn't been defined yet in the bootstrap sequence. Instead, writing it something like error("write(IO, Char): illegal character"*string(uint(c))) should resolve the error

[kmsquire]

Perfect--thanks Jameson, it works!

[JeffBezanson]

It seems to me the worst that can happen in writing a utf8 string is canonicalization. I think it would be ok to write a UTF8String to an IOString and get back slightly different bytes on reading, as long as valid characters remain the same.