Hi,
I was ready to send a message about the usefulness of Integer
literal parsing on this list, but I ended with a really long text :)
So I'm only posting the preamble here. Please, contact me directly if
you are interested in this problem, or tell me if this discussion is
in the scope of the list, in case I will post the whole thing later :)
------------ Preamble of my initial post ------------
the subject of this message is about the usefulness of integer
parsing (as integer constants) in the postgres backend. From my
understanding of backend source code, integer expressions are parsed
as integers constants (files 'scan.l' and 'gram.y') *then* transformed
back in strings in order to be processed by the right reader procedure
(for example, the function 'int4in' in the case of four bytes
integers).
Apart from removing the overhead involved in these
transformations, I think dropping the parsing of Integer constants
would simplify the declaration of new types and correct a possible bug
with parsing of oid litterals. As I'm surely wrong :), the remaining
of this message explains how I came to this conclusion and *try to
explain* some funny results I got with the regress tests after the
modifications.
------------ End of preamble of my initial post ------------
Remaining of the message will be send to people interested in the
problem :)
Thanks,
Jean-Marie
------------------------------
