I find these really useful --- finally I can isearch-regexp for
identifiers in code and match exactly what I mean.  Synthesizing these
operators from the existing ones is a pain, partly because "symbol
constituents" are those characters whose syntax is *either* 'word' or
'symbol'.

This also makes etags / ctags regexps behave a bit more like Emacs
regexps: '_' is no longer a word constituent.  I don't know if that's
actually a feature or not.