Re: Speed query

[Geoff Leyland <geoff_leyland@...>]

On 28/03/2008, at 12:36 AM, Tony Finch wrote:
> On Wed, 26 Mar 2008, Javier Guerra wrote:
> > but also noted that your speed test of the sort()ed queue just  
> > inserts
> > numbers in the queue, while the heap uses a {key,value} table for  
> > each
> > item... no fair!
>
> You could avoid this by storing all the keys and values in one  
> table and
> the ordering in another, i.e.
> 	pairs[key] = value
> 	index[n] = key

Good idea, but I tried it and it didn't seem to make a significant  
difference with lua or luajit.

I also tried changing the entries to { k, v } rather than { key=k,  
value=v } figuring I'd swap some hash lookups for index lookups, but  
it was slightly slower on lua and only a little faster with luajit,  
so I figured it wasn't worth it.

Javier's #self improvement is even better with luajit.  I tried going  
further by having a manually updated self.size to avoid *all* the  
#selfs, but it was no better.

Cheers,
Geoff