From: John Miller <jfmiller28@...>
Date: 2007-08-24T04:59:23+09:00
Subject: Re: Rope Data Structures

>Just a musing on this, but if we want a fast and scalable rope structure
>in Ruby (and speed and scalability are the reasons for using a rope, I
>believe?), wouldn't it be better (though not quite so much fun) to
>simply wrap a c / c++ rope library, rather than implement it from
>scratch in Ruby?
>
>Perhaps it'd even be possible to get Ruby's string to switch internal
>implementation using some huristics about its size.
>
>Cheers,
>  Benjohn

I agree that for something like this a binary distribution is best, but 
they still cause problems in cross platform environments. (eg. JRuby)  I 
think the best solutions are 1.) Have Multiple Gem distros (java, 
mswin32, source) and include a pure ruby version.  Or 2.) integrate this 
into the interpreter core as another built in class.


Pe単a, Botp wrote:
> On Behalf Of John Miller:
> # I tried writing a rope library in ruby and got an order of magnitude
> # better performance when compared to a String implementation (6.5 to 
> 48.5
> 
> wow, that's quite a helpful boost. specially since a large percentage of 
> the web and database time is spent on string manipulation. post your 
> work as a gem, pls.
> 
> kind regards -botp

The ICFP task was written in such a way as to make maximum use of Ropes. 
I think that processing eRB template would also see pretty good gains. 
However, one must be careful to use Ropes only when string concatenation 
and slicing (Ropes do these in O(1) ) are much more common then 
accessing the string content (O(log(n)).  An application where a string 
is build once and the accessed repeatedly will have poorer performance 
if it is converted to a rope.  On the other hand a string that is 
repeatedly manipulated before finally being printed a single time will 
see performance gains proportional to the average length of the 
String/Rope.

Databases do spend lots of time with strings, but there are many more 
accesses to strings then changes. Beyond that most of the strings that 
databases deal with are "short."

I will see what I can do about getting out some sample code.

John Miller

-- 
Posted via http://www.ruby-forum.com/.