Hi,
As part of previous discussions about typmod for user type, Tom
mentioned that you would need to make type and function names
equivalent. As it turns out, if you refactor a few rules, you can
actually make it work and manage them seperately. For this to work the
current "col_name_keyword" non-terminal has to be divided into few more
categories. The criterion is mostly whether it is followed by a left
parenthsis.
1. Some can be moved to unreserved words, mostly the types BIT, VARCHAR,
BIGINT etc since they don't need special rules anymore.
2. Words that have special productions (eg CONVERT, COALESCE etc),
these can still only be used as column names, not types or
(user-defined) functions.
3. Words which can be column names functions but not types. These never
appear normally with a parenthesis, so they will be interpreted as a
function if there is one. (eg SETOF, NATIONAL, etc)
4. Words that can be column names and types but not functions. These
are artifacts of the grammer due to the fact that VARYING is
unrestricted. These are BIT, NCHAR, CHAR_P and CHARACTER.
After this, you can have user-functions that take an arbitrary set of
parameters. There'll need to be a bit of code to verify the arguments.
It results in a net gain of 15 keywords for functions and about 30 for
types.
My question is, should users be able to create types schema.int4 and
schema.integer simultaneously. Currently it allows you but it's not
handled very well (\dT doesn't list both). Should this be allowed?
Should aliasing for DEC and DECIMAL -> NUMERIC be done for
user-defined types?
Have a nice day,
--
Martijn van Oosterhout <kleptog@svana.org> http://svana.org/kleptog/
> Patent. n. Genius is 5% inspiration and 95% perspiration. A patent is a
> tool for doing 5% of the work and then sitting around waiting for someone
> else to do the other 95% so you can sue them.