That way we get proper unicode support in regular expressions as long as the current locale treats wchar_t as unicode codepoints. Fixes #638 Fixes #595 Fixes #162