Re: with and use again (was: Manifesto against Vector)

[Martin Krischik <krischik@users.sourceforge.net>]

Dmitry A. Kazakov wrote:

> Martin Krischik wrote:
> 
>> Dmitry A. Kazakov wrote:
>> 
>>> On Thu, 6 May 2004 10:55:01 +0100, Marius Amado Alves
>>> <maa@liacc.up.pt> wrote:
>>> 
>>>>On Thursday 06 May 2004 10:29, Dmitry A.Kazakov wrote:
>>>>> On Thu, 6 May 2004 09:03:54 +0200, "Jean-Pierre Rosen"
>>>>> > ... With is a library level
>>>>> > clause; it tells: "this unit, somewhere needs that unit".
>>>>> Why that should be told? I'd say that "with" is superfluous.
>>>>
>>>>I think it's to make it easier for the compiler. Without the with a lot
>>>>more semantic processing would be required, and probably an additional
>>>>pass.
>>> 
>>> Huh, you will never extort that from a with-lover. (:-)) Guess why?
>>> Because if that were the only reason, then there would be no more
>>> arguments against use-ing without with-ing!
>> 
>> We love with not because of compiler technicalities but because we like
>> our programs to be readable. To take your Example:
>> 
>> procedure P (I : in Something);
>> 
>> procedure P (I : in Foo.Something);
>> 
>> Now imagine P inside at least 500 lines of specification with at least 5
>> with statements.
>> 
>> And now try to argue that the first version is more readable.
> 
> I will not, because I agree. Of course, if you really want to have
> specifications 500 lines long.

The GNAT specification if Ada.Text_IO has 445 lines. That's with comments
but without the generic child packages. In GNAT they are special child
packages drawn in by compiler magic.

So 500 lines is not that much.

> But this is not an argument *for* placing 
> "with Foo" before that mess! It is an argument for full names. Which is a
> quite different thing. And note, usually a necessity to use full names
> indicates bad design of the packages involved, a good opportunity to
> review them!

True, I am from the Type.Object school - as opposed to the Types.Type
school. But so are most UML -> Ada generators. You would not want to
declare UML users bad designers, would you?

With Regards

Martin
-- 
mailto://krischik@users.sourceforge.net
http://www.ada.krischik.com