Re: [Spark] Converting Arrays

[Lutz Donnerhacke <lutz@iks-jena.de>]

* James S. Rogers wrote:
> "Lutz Donnerhacke" <lutz@iks-jena.de> wrote in message
>> I run into a difficult problem (for me):
> 
> change the initialization of "path" as follows:
> 
> path := (others => ASCII.NUL);

path := (1 .. s'Length => s, others => ASCII.NUL);
would be the approbriate Ada construct.

But I do not talk about Ada. I talk about the Spark subset of Ada which
refuses such construct because:
  - it's a untyped aggragate (illegal in Spark).
  - constains non static components (slicing in illegal in Spark).
  - assignes to non statically known bounds. (illegal in Spark).

> Note that another problem looms in your code.
> The "for" loop will attempt to access one element
> beyond the end of string "s".

Yep. Spark would found this later. But currently it refuses the construct as
a whole. What does not compile, is already erronous. ;-)

> It appears that you are trying to copy s into path
> while skipping the first element of s. This can be done
> more directly and correctly with array slices:

Slicing is illegal in Spark.

A possible solution is to assign the values in two loops. But Spark refuses
this construct, too, because the data flow analysis handle arrays as a
single variable and therefore compains about uninitialized values, reuse of
uninitialized values and contradiction to the predicted information flow.

> You will need to decide how to handle the case when
> path is shorter than s.

Very simple: I have a precondition on the specification claiming that s is
really shorter than path, so on every call to this function, Spark proofs
that this procedure will not fail due to unexpected parameters.

I fear, that I have to use a constraint array type for out variables. :-(