Re: Use use type? (Was Re: Safety-critical development in Ada and Eiffel)

[dewar@merv.cs.nyu.edu (Robert Dewar)]

Brian said

<<I have always wondered about this, but I never asked. Perhaps we need a
"Design and Evolution of Ada 95" to answer some of these questions. Do
you have the values of YY, MM, and DD that I need to find the discussion
in the mrtcomments list? I'm curious because I want to know how you all
resolved the tension between fixing (perceived) flaws and breaking
backwards compatibility. Making primitive operators always directly visible
seems like the right approach; what would the extent of the
incompatibility have been? Personally, I don't like "use type", and think
it was a mistake, but my mind has been changed on important Ada issues
(downard funargs :-) after I read the discussions.
>>


I don't understand the motivation at all behind making an exception
for primitive operators. I still don't see why people have such a
different view of

  Add (x, y)

and

  x + y

I appreciate (though usually do not agree with) the position that everything
must be dotted so you can see where it came from in the absence of tools
to identify defining occurrences easily.

But I do not see why the above two examples strike people as so different.
To me the infix notation is simply minor syntactic sugar for the function
call. Minor from a semantic point of view.

I understand that people like the convenience of infix notation, but if
you really believe that dots are needed so that the maintainer knows
exactly where everything comes from, then why the exception?

If your principle is on the other hand: use dots where needed for clarity
(and I have some ideas of what this means). Then why not use "use" clauses
and dot where you want, and don't dot where you don't want.

Sure that is a bit more of a burden on the writer, who has to understand
the principles of when dots are or are not required, but not on the
reader, and it is the reader we are concerned with.

Why not use "use" systematically, and then have the compiler or some other
tool enforce whatever restriction you want on the use of undotted
references.

Note that a pragma like

  pragma Restrictions ("Dot_Where_Required_By_ACME_Tools_Conventions")

is a perfectly valid one.