Re: Community Input for the Maintenance and Revision of the Ada Programming Language

[Vincent DIEMUNSCH <vincent.diemunsch@gmail.com>]

Le mardi 29 août 2017 01:49:15 UTC+2, far...@gmail.com a écrit :
> For me, Ada is a wonderful language for my engineering calculations.
> I have built my own tools including operator overloads for real, imaginary, complex, and integer scalars, > mating with overloads for vector and matrix types.

Would you be ready to share this ?

> Oddly missing from Ada, and commented about on this list before, is why brackets [ ] aren't used for 
> array indexing.

True. This is really not understandable in 2017 ! I think it comes from the DoD's Steelman requirements to have the shortest number of ASCII characters. But that was not their best idea and Ada has now taken the opposite path in allowing Unicode identifiers..

Regarding arrays I propose the following evolutions :

1. Allow discriminants in array types (if they are not unconstrained array types).
type Matrix (M,N : Positive) is array (1 .. M, 1 .. N) of Long_Float.

2. Allow a special indexing syntax, that I called relative indexing. Adding a Unary + operator (that has no other purpose ! :-) ) in front of an index expression means starting from A'First. Unary plus would then become the ordinal operator.

 A(+i) simply means A(A'First +i). 

This is an elegant and above all retro-compatible way of permitting to start every array at zero which is easier in algorithms, cf. Dijkstra (EWD 831 "Why numbering should start at zero") or Wirth's Oberon. Using A'Range is not possible outside the basic cases. Anytime you want to compute indexes you fall on relative indexes from 0 to N-1 and convoluted expressions that add or suppress A'First.

> Also missing is multidimensional array slicing. 
But above dimension 2 one needs Column Major Order as Fortran or Matlab...  And not Row Major Order as C or Pascal. Otherwise a slice from a 3 dimension arrays is not binary compatible with a 2 dimensions array.  This is very well known but has never been corrected. Hence the problems between LAPACK and the Ada «Generic_Arrays».


> Some will question whether these are appropriate to include in the language proper and that is a valid
> conversation. But it is things like this that will help get Ada away from the "embedded" stereotype.

I totally agree.

Regards,

Vincent