Re: playing around with dicts

Hi Michael,

Thanks for the feedback.

On 12/13/2013 06:57 PM, Michael Hendricks wrote:
> I've been playing with dicts in V7.  Here are some observations/questions.
>  This is all done with V7.1.3 on OS X Mavericks.
>
> ?- assert(person{name:john}).
>
> gives a segfault.  Full trace is at https://gist.github.com/mndrix/7947317
>
> I was glad to see that dicts are allowed as facts and clause heads.
>  However, calling listing/0 on such programs gives an error about
> sub_atom/5.  See https://gist.github.com/mndrix/7947389

I think that dicts being allowed as clause head is more a coincident
resulting from the fact that they currently are implemented as ordinary
compound terms with a reserved symbol as functor. I think that should
not be allowed. For one thing,

	person{name:john}

is a different predicate than

	person{name:john, age:42}

I'd also like to have the option to make dicts completely distinct from
compounds.  For this type of usage, the ECLiPSe struct approach would
be more interesting.

> Section 5.4 of the manual says, "If a dict is written, the keys are written
> according to the standard order of terms".  Do we want to commit to that
> behavior so early in the design?  If the implementation changes to
> something based on hashing, it could be expensive to generate sorted
> output.  We can always commit to sorted output later.

I consider allowing them as predicate heads a more involved restriction.
Dicts are already not stored in the standard order of terms, but in the
order of internal handles. So, printing already needs to (re-)sort them.
Having predictable printed output is (IMHO) a nice property.

	Cheers --- Jan

... happy after doing a bit of application programming using dics today.