SRE - adding chars to pattern produces invalid results

[jchd]

That would work but for general use (international, since the word has been used), one has to handle grapheme clusters correctly.

That's way beyond what a simple and efficient function can handle. The same question also points its nose with StringLen about its semantics: does an unavoidable grapheme cluster count for "as many codepoints it uses" or for "one user character"?

[czardas]

The same question also points its nose with StringLen about its semantics: does an unavoidable grapheme cluster count for "as many codepoints it uses" or for "one user character"?

I think this is an unanswerable question and functions to handle specific syntax should be separate specialized functions. Core functions ought to be indifferent. Why? - Because there is no end to the different possible syntax structures you might encounter, now or in the future. So the core function _StringReverse should produce horrors like LFCR and messed up clusters.

Edited April 8, 2013 by czardas