[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: gnubol: subsets
In a message dated 11/25/99 11:28:14 AM EST, TIMJOSLING@prodigy.net writes:
<<
The code can be in the reductions for a and b so there is no real problem
with
this (when there are two options; when there are many as in the data
definitions,
you need to manually count them to avoid exponential explosion)..
>>
These ideas are exactly correct. Except, IMHO, entirely optimisitic. In the
code base are mistakes so even though you have expectation of two conditional
clauses, that is not what you encounter, instead you can get many. We want to
stay on our feet. Especially when you see that there could be the kind of
errors that make later conditionals fly back to the outer arithmetic where
numerous can accumulate.
And this is where the seeming theory becomes a practical issue for the
compiler writer. Any where you have an expected count, say in the limit of
three phrases in a VARYING clause, or just one VARYING clause in a perform;
you could of course see something quite different. Any time you expect an
order of clauses, you could see something quite different.
The order of the day is manual count. It is real unhappy. The optimistic
grammar will not do it. You will loose too much code and user will complain
that subsequent compiles discover problems you should have flagged before.
The real issue here is the kind of thinking the tools encourage. We can not
have the nieve hierarchy of rules that we at first dream of. Many source
programs do not fit together! Many people want to use the compiler for early
code review. Even in mostly good code minor error can cause serious
disruptions.
In the procedure division, IMHO, we will need much manual counting, and we
will also need to review the order of occurence after the fact. Most tedius.
Best Wishes,
Bob Rayhawk
RKRayhawk@aol.com
--
This message was sent through the gnu-cobol mailing list. To remove yourself
from this mailing list, send a message to majordomo@lusars.net with the
words "unsubscribe gnu-cobol" in the message body. For more information on
the GNU COBOL project, send mail to gnu-cobol-owner@lusars.net.