Re: The real problem with the SQL precompilers (was SQL on AS400...)

On Fri, Jul 18, 2008 at 12:06 PM, Jon Paris <Jon.Paris@xxxxxxxxxxxxxx> wrote:

I agree with your assessment. I know IBM has made some improvements
to the pre-compiler (thanks Gina!), but this disconnect still bites
us all the time. WDSC (RDi) falls over when reviewing compiler
errors, debugging is ugly, x-ref and outline features are
incomplete, etc..... IBM needs to eliminate this asap. Seems to be
in IBM's best interest to make their tools usable, without the work-
arounds that we currently are stuck with. IBM should find this
situation embarrassing.

I think the opportunity has already been missed.

When they were contemplating making the changes that we have recently
seen (and I agree that Gina did good work) they had the opportunity to
throw it all away and start over and do it properly. They chose
instead to put more chewing gum and baler twine on an already flawed
approach. We can't blame Gina for that - she wasn't responsible for
the architectural decisions. I suspect that budget (both in Toronto
and Rochester) was the ultimate stumbling block.

The ridiculous aspect of this is that the compiler on which RPG IV was
based already used the "compiler in charge" approach and it had to be
disabled for the ILE implementation. The result of course is that the
pre-compiler is almost doomed to be forever behind the compiler as it
always has to be updated whenever the compiler changes. That's
constant duplication of effort which could have been avoided with a re-
write.

I think the pre compiler concept, with its integration thru views with
the interactive debugger is the best feature found in the IBM i
compilers. Hypothetically you can use the pre compiler to add new
feature to ILE C or ILE RPG to the extent that a new language can be
created. The only failing of the pre compiler I see is it is not
complete. Variable names used by the programmer cant be mapped to the
actual code that was passed onto the compiler. ( you see the impact of
this when debugging an sql procedure ) The associated space API which
allows a precompiler to store additional info in the module and
program objects dont appear to work at all. Maybe those who did not
like the pre compiler idea in the first place spread enough doubt to
convince IBM mgmt that there was no use funding the effort.

here is a video of a computer language team in it for the long haul,
with decision making management which lives and breaths their
products.
http://channel9.msdn.com/posts/Charles/C-40-Meet-the-Design-Team/

-Steve