Lists/Maps/Sets (Was Re: The story of E, part 2 (fwd))

[Mark S. Miller]

At 10:05 PM 10/11/98 -0700, Ka-Ping Yee wrote:
>First, personal opinions about the individual names...
>
>I have to say "TableEditor" sounds particularly awful to me;

It sucks.


>             |    mapping       sequence
>  -----------+------------------------------
>   immutable |     table          tuple
>             |
>    mutable  |   dictionary       list
>

>             |    ********       sequence
>  -----------+------------------------------
>   immutable |    mapping         tuple
>             |
>    mutable  |   dictionary       list

>             |    mapping         tuple 
>  -----------+------------------------------
>   immutable |    mapping         tuple
>             |
>    mutable  |   dictionary       list


Here are the Java-imposed naming constraints:

* Java already has a "Dictionary", and it has a different contract
(guaranteed not to admit nulls as keys or values), so I can't reuse that term.

* Java already has a well known mutable sequence type, unfortunately called
Vector, whose contract is compatible with my needs.

* Java has used up "Array" for a contract that is element-mutable but not
length-mutable.


Here are the MarkM-esthetics-imposed constraints:

* The items on the immutable row correspond to mathematical objects, as
might be taught in a course on set theory.  They should have names
suggestive of their math-value nature, which is why I like "Mapping" better
than "Table" for the immutable first-column thingo.  

* Even though none of your suggested matrices have "Table" as the
first-column label, I didn't catch an argument against it.  

* There remains the issue of what call a mutable first-column thingo.
While cute, I don't want to re-use Xanadu's choice of "MuTable".


Putting these together, we've got

 	              Table			Sequence
 	       ----+----------------------------
 	immutable  |  Mapping		Tuple
 	           |
 	mutable    |  ???????		Vector


For the missing one, I'm at a loss.  Suggestions solicited.  Consider
coming up with terms for "snapshot"/"newMutable" at the same time.  

If a generalizable naming pattern can be found that generalizes to other
such column-triples, I'd be willing to waive many of the above constraints.
 The memorability of a reusable naming pattern is worth a lot, and the same
three-way distinction keeps coming up elsewhere (eg, Slot, FinalSlot,
SettableSlot).