I've committed changes to gram.y to get PATH accepted as both a data
type and a column label (the latter is what the regression test has been
complaining about).
There is some difference in "edge case" names allowed for ColId and, for
example, function names, though I think the current state is more
consistant than earlier. There are likely to be more changes coming, and
*possibly* these changes will include ugly restrictions on where SQL92
type names are allowed to appear.
This last point comes from SQL99 allowing
CREATE FUNCTION name (IN argname argtype,...) CREATE FUNCTION name (argname argtype,...) CREATE FUNCTION name
(argtype,...)
and afaict most other permutations. yacc has trouble with the "argname
argtype" pair since it can't tell by looking ahead just one token
whether what is reading is an "argname" or an "argtype". I'll keep
poking at it...
- Thomas