display something useful for text/plain output of invalid String

[stevengj]

Rather than throwing an exception in the REPL when it tries to show a String containing invalid UTF-8 data, this displays something useful.

[stevengj]

Sample output:

julia> String(rand(UInt8, 10))
10-byte String of invalid UTF-8 data:
 0x17
 0xc6
 0x08
 0xde
 0x53
 0x2e
 0xae
 0xc3
 0x1b
 0x49

[nalimilan]

Good idea. Though it would be even more useful if you could also print the beginning of the string up to the first invalid character. That would make it easier to spot where the problem is (especially in long strings).

[stevengj]

@nalimilan, my thinking was that strings containing invalid UTF-8 are most likely because the user was stuffing arbitrary binary data into a String, in which case there is no point in printing any of it as a string.

[stevengj]

(We can always do this for now, and decide later if we need more verbose output to explain where the data stops being valid UTF-8 if that turns out to be useful in practice.)

[nalimilan]

@nalimilan, my thinking was that strings containing invalid UTF-8 are most likely because the user was stuffing arbitrary binary data into a String, in which case there is no point in printing any of it as a string.

One can also frequently get partially invalid strings, e.g. when reading text in the wrong encoding (ISO-8859-* as UTF-8), or with corrupt filenames. It's very useful in that case to know where's the invalid character.

(We can always do this for now, and decide later if we need more verbose output to explain where the data stops being valid UTF-8 if that turns out to be useful in practice.)

Sure.

[StefanKarpinski]

This will get re-revised in the 0.6 release as part of #16107 but for now this is fine.