I'm thinking of adding an escape mechanism to be able to format wide strings a bit better. Right now there's the choice between sprintf %s which just let wide chars straight through, and %O which quotes them along with all other control chars. Neither is really good if you want to e.g. print a string to stdout which possibly can contain some wide chars.

For that I'd like to add a %S that only quotes wide chars in some way. The question is what quoting method to use. Java and many other languages has one with "\u" which always is followed by four hexadecimal digits. (Java doesn't say anything about chars wider than 16 bits, but I guess it's "\U" followed by eight hexadecimal digits, as in some generic Unicode technical papers.)

Java also specifies a way to keep it unambiguous: If a string already contains a sequence like "\u4711" it's quoted by inserting another "u", i.e. it becomes "\uu4711".

The \u and \U scheme is better than the \4711 and \x9c9 escapes that Pike uses in strings since there's no ambiguity where the escape sequence ends (%O uses tricks like "\4322""33" to cope with them but that doesn't work very well in strings without quotes around them).

The ad-hoc quoting scheme "<nnnnn>" used in some places in describe_svalue would also be replaced by this for consistency.