From: Hal Fulton <hal9000@...>
Date: 2004-03-04T08:07:15+09:00
Subject: Re: why are Hashes so unsorted? what's your solution?

John W. Kennedy wrote:
> David Garamond wrote:
> 
>> Most language don't guarantee insert order for hash, in fact IIRC some 
>> deliberately randomize it. 
> 
> 
> They /all/ do, or else they're calling something a "hash" that ain't.
> 
> They're /called/ "hashes" (and have been since the days of vacuum-tube 
> mainframes) /because/ the process "makes a hash of" the ordering.

You're correct, of course, as is Lothar.

The reason for the new (mis)use of the term, I believe, is this.

The language feature is named for its implementation (an unfortunate
historical quirk). I am more interested in its interface.

Ideally the language feature should have a name that reflects how it
can be used rather than how it is implemented.

I'm as guilty as anyone else of saying "ordered hash." When I say it,
I mean a data structure whose elements can be addressed like an array,
but with a non-numeric key which is an arbitrary object. Those are
interface issues.

I care about how I put things in and how I get them out. I don't want
to know how it's implemented (except that the implementation "pokes
through" when I retrieve the list and it's not in the original order).

This reminds me a little of the people who say you can't talk about
the "screen" of a laptop. It's a "display"; there's no network of wires
crisscrossing in it, and hence no "screen."

The term "hash" is so much more fun and elegant than "associative
array." Let's come up with a term that is both elegant and accurate
for an ordered entity with a hash-like interface.

Hal