So there was some kind of misunderstanding.
Perceived point of view:
There should be a total ordering on the (in theory)
infinite set of J values. Fine.
And maybe more than one (Elijah already pointed at <. vs /:)
in different contexts.

This again drives us to the question:
Do we actually want an AA or a dict?
First, we need to talk terminology:
what /is/ an AA versus a dict?
• order?
• homogenous?
• defined primitives?

And then again, quote

But what we need here, more than anything, are use cases.

So what are the odds of J dict (or AA?) users needing
• order (amoung values with/out order in current J)?
• homogenous/not?
• one primitive or another?

And dict/AA aren’t the only alternatives that have been
taken into consideration on this thread:
inverted table based and tree-like structures
should be compared against as well I think

Many ?s to resolve …


Am 02.02.22 um 00:34 schrieb Raul Miller:
On Tue, Feb 1, 2022 at 3:21 PM Elijah Stone <elro...@elronnd.net> wrote:
I don't care what you call it; forcing an order on keys removes
implementation freedom which may have a negative impact on performance.

At no point did I suggest "forcing an order on keys" -- I instead said
that there needed to be an order.

I have worked with implementations in other languages where the
implementation disregarded my key order unless I went to great lengths
to retain that ordering outside of the dictionary. This was documented
as "unordered". From a language semantics point of view, I feel that
that "unordered" mechanism was a mistake. (A premature optimization.)

Thanks,


--
Raul
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

--
----------------------
mail written using NEO
neo-layout.org
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to