From: Gavin Sinclair <gsinclair@...>
Date: 2003-02-04T22:27:41+09:00
Subject: Re: Native rational and complex number implementations?

On Tuesday, February 4, 2003, 11:33:10 PM, Tobias wrote:

> On Tue, 4 Feb 2003, Gavin Sinclair wrote:
>> I believe it would be beneficial to have Rational and Complex
>> as built-in classes.

> Could you tell some concrete expectations from making them built-in, i.e. 
> the reasons why you think it would be beneficial?

Small benefit, really.  It would be easier to document because
everything is designed to interoperate up-front.  The current state
is: you require some files which butcher the existing classes.  Thus
the API documentation is split.  Not that I expect anything to change
for that reason; I'm just curious, really.

I do think it would be elegant to include them, and that's a good, but
not sufficient, reason.

> All that I would like to have built in is a ruby literal for specifying 
> complex numbers (a literal for purely imaginary numbers would be
> sufficient, of course) because it's tedious to write
> Complex(2,3) if what you want is 2+3i

True, but how many literals do you really define?  Numbers tend to be
kept internal to algorithms.

Also, alias C Complex :)


> I don't think significant performance gains are possible by making Complex
> built-in because of the templated nature of the Complex class (real and 
> imaginary part can be any Numeric except Complex itself).

Probably not, but if a method is creating lots of complex #s then what
would you prefer: Ruby methods to create them, or machine code?


Again, just curious to see if there are good reasons to keep them out,
apart from the "too busy; leave me alone" reason, which is very good.

Gavin