On Oct 5, 2006, at 11:35 AM, Mark Wieder wrote:
Wednesday, October 4, 2006, 5:47:16 PM, you wrote:
I think this underscores the need for REAL type checking. I was told
that a positive integer would return true if "is a date" is used.
BZ #2783
People have created interesting bugs related to types in languages
with strong typing.
Many languages have type integer which is not an integer but some
uniform representation of some subset of integer.
Suppose some number in a program should only be a prime. Should a
prime type be invented?
Some languages infer types.
Types are a hint to compilers and in some languages are a ball and
chain to the compiler. Types allow a developer to organize thoughts
and to catch errors. However, types are part of a more general
mechanism related to constraints on the ranges and domains functions
and commands. This mechanism might involve good comments, runtime
assertions, assertions for analysis, assertions for compiling, and
typing. This can be generalized to I/O and sometimes intermediate
values.
In this particular case, the common Revolution approaches are 1. to
comment constraints or 2. to check for value errors and do something
about those.
In functions I use internally, I usually use comments to describe the
sets of allowed values for parameters. In functions made available
to customers, I use one or both. Sometimes, I am pretty casual about
what a parameter might be.
Revolution lends itself to runtime assertions and I expect that many
folks have rolled their own.
The detailed syntax of Revolution is poorly defined, but I would not
be surprised if someone someday creates a smart lint that can handle
simple proofs. That may be a while. I would not be surprised if the
compiler might catch more in the future based on ranges of values.
As far as 'is a date', I tried '"dinner and movie" is a date' in the
message box and got false, so 'is a date' might be built with an
assumption of a specific meaning for 'date'.
I have enjoyed strong types in programming, but I don't think typing
is what is needed here.
Dar
_______________________________________________
use-revolution mailing list
use-revolution@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription
preferences:
http://lists.runrev.com/mailman/listinfo/use-revolution
