Re: [open-cobol-list] Sort utility within OC

[Roger While]

>  > Some thoughts -
>  > 1)
>  > I don't think we should clutter the command line.
>  >
>  > MF's mfsort has precisely 2 calling sequences :
>  > mfsort <instructions>
>  > and
>  > mfsort take <filename>
>  >
>  > ie. All input/output file(s) are in the instructions.
>
> The SU I had envisioned was modeled on main-frame methods.
> Thus the actual input/output file(s) names are defined externally.
>
> This could be achieved by using predefined environment variables, and
> command line overrides.
> The command line interface would only consist of about 3 options.
>
> In my view this approach would result in a cleaner SORT grammar.

OK. I don't have strong views either way on this.



>  > 2)
>  > We need to define a (preferably compatible) grammar
>  > for the instructions.
>  >
>  > 3)
>  > Given the above, we could take MF's syntax as a start point.
>
> I have written a grammar which is based on main-frame SORT utilities.
>
> Due to IP considerations, I do not think we should base the SORT grammar
> on any specific vendor.


Well, there is bound to be some overlap in the pseudo-language.


>  > 4)
>  > Note that mfsort does not offer conditional evaluation
>  > as Cris's examples.
>  > (I am not sure we would want to support this)
>
> These SORT utilities became popular because of these EXTRA features,
> such as omit-include conditionals.
> In many cases, it is eisier and faster, to write a SORT script than to
> write a COBOL (or other) program.
>
> I don't think it is in the interest of any COBOL vendor, to create a
> product which would duplicate features, and compete with the main product.
>
> In my view, I think that omit-include conditionals are an important
> feature to have in a SORT utility program.

Sorry, I should have added "in the initial implementation".
I agree it would be useful.



>  > 5)
>  > Given the above, it should be possible to "flex" a grammar.
>  > We could then define a standard API for what the flexxer
>  > produces. This would be the specific hooks into the
>  > Cobol runtime of OC/TC.
>  > Dependant on how the flexxer works out, then David and I can
>  > agree on a common API.
>
> Perhaps first we should agree on what this SORT utility should be able
> to do.
>
> I think it should perform the following tasks.
>
> 1)
> Sort ASCII char-sets on descending/ascending fields,
> with the ability to remove duplicates.
> 2)
> Reformat input records (before sort).
> Reformat output records (after sort).
> 3)
> Sum numeric fields (packed, binary, zone decimals { EBCDIC binary ?})
> 4)
> Some form of omit-include conditionals.

Well, I sincerely hope we would be able to sort on ANY field type.

Roger