Subject | Re: [Firebird-Architect] RFC: Please unify stored procedure execution |
---|---|
Author | Ann W. Harrison |
Post date | 2004-12-24T15:33:38Z |
At 11:11 PM 12/23/2004, Fabricio Araujo wrote:
got legacy behavior and your proposal, though simple
and consistent internally, does have the unfortunate
side-effect of breaking existing behavior that follows
the old rules and documentation.
My suggestion is that we add the RDB$PROCEDURE_TYPE
field, filled in when the procedure is compiled and
do nothing else. No additional checks, errors, or
anything.
1) Existing stuff, including really warped procedures
that do one thing when invoked with EXECUTE PROCEDURE
and another when SELECTED still work - as well as they
ever did.
2) Interface layers no longer have to parse procedures
to guess the best way to invoke them.
3) Naive users get a hint of how the system thinks their
procedure should be invoked.
Regards,
Ann
>I'm adept of the KIS principle: Keep It Simple.Simplicity is a great virtue, but in this case we've
got legacy behavior and your proposal, though simple
and consistent internally, does have the unfortunate
side-effect of breaking existing behavior that follows
the old rules and documentation.
My suggestion is that we add the RDB$PROCEDURE_TYPE
field, filled in when the procedure is compiled and
do nothing else. No additional checks, errors, or
anything.
1) Existing stuff, including really warped procedures
that do one thing when invoked with EXECUTE PROCEDURE
and another when SELECTED still work - as well as they
ever did.
2) Interface layers no longer have to parse procedures
to guess the best way to invoke them.
3) Naive users get a hint of how the system thinks their
procedure should be invoked.
Regards,
Ann