From: Alex Gutteridge <alexg@...>
Date: 2007-04-03T09:50:46+09:00
Subject: Re: implementing an array based class

On 3 Apr 2007, at 08:35, bwv549 wrote:

> Some of the data I deal with involves hundreds of thousands of records
> where each is very predictable in the number of attributes.  By going
> from a hash based object to an array based object I am able to
> substantially decrease memory usage.  However, the array based object
> I have made (see below), is really lame for two reasons:
>
> 1. It is a pain to make a different array based class
> 2. When I call 'flatten' on an array of these objects, my objects get
> flattened, too (how do I prevent this?)
>
> I realize that I could just use arrays of arrays for all the objects,
> but for ease of programming I'd like to be able to access an object's
> attributes by name.  Does anyone know of some memory efficient and
> fast way of doing this (that's insensitive to 'flatten')?

snip

> Ideas and suggestions are most welcome.  Thanks.

I don't know about speed/memory use, but perhaps Struct (in stdlib)  
or Arrayfields:

http://raa.ruby-lang.org/project/arrayfields/

does what you want?

Alex Gutteridge

Bioinformatics Center
Kyoto University