What's the difference between hex code (\x) and unicode (\u) chars?

Question

From ?Quotes:

\xnn   character with given hex code (1 or 2 hex digits)  
\unnnn Unicode character with given code (1--4 hex digits)

In the case where the Unicode character has only one or two digits, I would expect these characters to be the same. In fact, one of the examples on the ?Quotes help page shows:

"\x48\x65\x6c\x6c\x6f\x20\x57\x6f\x72\x6c\x64\x21"
## [1] "Hello World!"
"\u48\u65\u6c\u6c\u6f\u20\u57\u6f\u72\u6c\u64\u21"
## [1] "Hello World!"

However, under Linux, when trying to print a pound sign, I see

cat("\ua3")
## £
cat("\xa3")
## �

That is, the \x hex code fails to display correctly. (This behaviour persisted with any locale that I tried.) Under Windows 7 both versions show a pound sign.

If I convert to integer and back then the pound sign displays correctly under Linux.

cat(intToUtf8(utf8ToInt("\xa3")))
## £

Incidentally, this doesn't work under Windows, since utf8ToInt("\xa3") returns NA.

Some \x characters return NA under Windows but throw an error under Linux. For example:

utf8ToInt("\xf0")
## Error in utf8ToInt("\xf0") : invalid UTF-8 string

("\uf0" is a valid character.)

These examples show that there are some differences between \x and \u forms of characters, which seem to be OS-specific, but I can't see any logic in how they are defined.

What are the difference between these two character forms?

Answer 1

The escape sequence \xNN inserts the raw byte NN into a string, whereas \uNN inserts the UTF-8 bytes for the Unicode code point NN into a UTF-8 string:

> charToRaw('\xA3') [1] a3 > charToRaw('\uA3') [1] c2 a3 

These two types of escape sequence cannot be mixed in the same string:

> '\ua3\xa3' Error: mixing Unicode and octal/hex escapes in a string is not allowed 

This is because the escape sequences also define the encoding of the string. A \uNN sequence explicitly sets the encoding of the entire string to "UTF-8", whereas \xNN leaves it in the default "unknown" (aka. native) encoding:

> Encoding('\xa3') [1] "unknown" > Encoding('\ua3') [1] "UTF-8" 

This becomes important when printing strings, as they need to be converted into the appropriate output encoding (e.g., that of your console). Strings with a defined encoding can be converted appropriately (see enc2native), but those with an "unknown" encoding are simply output as-is:

• On Linux, your console is probably expecting UTF-8 text, and as 0xA3 is not a valid UTF-8 sequence, it gives you "�".
• On Windows, your console is probably expecting Windows-1252 text, and as 0xA3 is the correct encoding for "£", that's what you see. (When the string is \uA3, a conversion from UTF-8 to Windows-1252 takes place.)

If the encoding is set explicitly, the appropriate conversion will take place on Linux:

> s <- '\xa3' > Encoding(s) <- 'latin1' > cat(s) £