From: matz@... (Yukihiro Matsumoto)
Date: 2000-12-03T23:33:44+09:00
Subject: [ruby-talk:6765] Re: Array Intersect (&) question

Hi,

In message "[ruby-talk:6755] Re: Array Intersect (&) question"
    on 00/12/03, "Brian F. Feldman" <green@FreeBSD.org> writes:

|Personally, I feel that a non-ordered &() method would violate the POLS for 
|a lot of people, including myself.  How about adding a faster version as 
|intersection(), then?  I'm sure a similar situation exists with |() and can 
|be optimized by making a new union().

I think I'm going to re-implement & and | using hash, preserving
current result order.

|I like highly deterministic API calls - preserving order and object 
|duplication would be the only reasonable default, in my humble opinion.

But what do you mean by `preserving object duplication'?  Do you agree
with Ross's opinion?

[ruby-talk:6752]
>   a2 = %w# 1 1 y #      -> ["1", "1", "y"]
>   a3 = %w# 1 y 1 z #    -> ["1", "y", "1", "z"]
>   a2 & a3               -> ["1", "y"]
>
> Whereas I believe the true intersection of a2 & a3 should be
> ["1", "1", "y"] . Opinions?

FYI, Current behavior does not preserve duplication.

							matz.