Re: ada and robots

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

Bob Duff wrote

(regarding variable length parameter lists in Ada, lack thereof)

<<Granted.  IMHO, a "nice to have" feature, but certainly nothing to do
with why C is better or worse at accessing low-level hardware
functionality.  It was left out of Ada to simplify the language.
 

I disagree, there are two issues here. Variable length lists of homogenous
types (which can include of course collectoins of pointers to different
subtype instances of a class) can perfectly well be passed. For example,
a routine called put_lines that outputs a series of lines can be called

put_lines (("line1", "line2", "line3"));

This kind of "implicit rowing" first appeared in Algol68, and is a very
convenient way of dealing with variable length parameter lists.

However, what the C folks might want is the completely general capability
for variable length lists of arbitrary heterogenous types. This is possible
in C, but is seriously type unsafe. Furthermore, there is no obvious way it
can be made type safe. Hence this feature definitely should be rejected, as
should any programmer who feels they cannot program without it.

To be fair, it is possible to get something like this by means of an implicit
uniting coercion, but it is really very little trouble to introduce this
explicitly.

So for example, suppose you want a routine that takes an arbitrary list
of integers and floats (could obviously be extended), then you could make
a routine in Ada that looked like

  printf ("the value of x is % and the value of y is %",
            (I(integer expr), F(float expr)));

which seems close enough to the C notation if this is really what you want.