On Mon, Dec 10, 2012 at 10:49 AM, Marco Antoniotti
<marcoxa@cs.nyu.edu> wrote:
On Dec 9, 2012, at 21:31 , Nick Levine <
ndl@ravenbrook.com> wrote:
> 1. Typo "devbelopers" for "developers".
>
> 2. Maybe I'm just dim, but I really don't understand what this is
> about. The functional interface looks like that of a queue (or a
> stack, depending on whether insert and remove work on the same end as
> each other or not). Can you explain? give a couple of examples?
I have nothing against this (*), although the spec leaves out a few corner cases. E.g., what happens when the keys are equal, but the values/content associated are not?
That is unclear in the specification. The example implementations would do different things depending on if you use something that is (essentially) "less than" or "less than or equal" (in the former case, they'd not re-order the equal values whenever there's an operation on the queue, in the latter case, they would).
Apart from that, I do not understand the use of DECLARE-HEAP and I believe that there should be no provisions to define a package for these sort of data structures. Or better: I believe a discussion about how to structure packages and libraries should be started. But this is an old pet peeve of mine.
That is definitely something that is worth discussing. The main reason fro submitting GENHEAP is that I didn't see anything similar and, well, I did submit GENHASH back in the early days of CDRs. If people are, generally, hesitant, I am more than happy to withdraw the submission.
The main reason for DECLARE-HEAP is to allow for the example implementation that defines all underlying operations in terms of methods on a specialized class to be easily created on "this is the type of heap I want" rather than on the first creation.