Name: clC74495 Date: 09/22/98


Dear Sir/Madam,

When using containers, such as Vectors, Sets,
Lists, and Maps, type casting for objects is necessary
when elements are returned from the containers in
current specification and implementation.
Usually, there are three kinds of uses after
an element is returned:
(1) Assigned directly to a variable;
(2) Used as an argument to a function call;
(3) One of its methods called.

The first two constitute substantial part of their
usage although I don't have the statistics. And
I argue that the type casting from Object to a proper
type is somewhat redundant for these two cases
(it's necessary for the third case though). It requires
programmers to cast an object to a type REQUIRED by the
assignment (case 1) or the function (case 2). It
may help programmers to think twice when they type
in the casting. Other than that, there is no use
and I believe it's kind a burden on programmers
and obscures the code due to obvious, lengthy casting
that can be determined by context.

More importantly, the compiler can only check if the
casting meets the requirement, but cannot guarantee
its CORRECTNESS, until at run time. For the first
two cases, the compiler can automatically determine
the cast from context (currently it's an error
message). Why not turning this error message into
action to save people's time and energy by eliminating
the casting requirement before we have a strong typed,
parametric container solution for Java?
I don't think it will compromise Java's strongly-
typedness nature at all. (The only difference
might be the programmers will only think once
not twice :-)

Thank you for serious consideration.
(Review ID: 36944)
======================================================================