Re: Speed query

[David Given <dg@...>]

Geoff Leyland wrote:
> Good question.  I just tested (b) vs a pure lua binary heap and the 
> binary heap is so much better (the sort version still hasn't finished) 
> that I think I must have got (b) wrong (does table.qsort exist?  I can 
> only find table.sort documented.  Which is better out of table_insert(t, 
> x) and t[#t+1] = x?)
>
> If you want to try my heap code I can send it, but I think there's one 
> in loop that's bound to be better.

That deeply surprises me, but I can hardly argue with results --- so, 
yes please, I'd love a copy.

(For context: it's to handle a message queue. The priorities are 
timestamps and will be floats. Therefore bundling up equal priorities 
isn't going to be any use, although it might be worth special-casing the 
'now' timestamp.)

I've looked at loop. Unfortunately, loop uses lots of infrastructure I 
don't want to use, and is also very focused on the Javascripty 
data-and-methods-in-the-same-namespace model which makes me rather 
uneasy (although priority queues are one exception), so I'd rather have 
my own version where I know what it does.

Thanks for all the comments, people!

--
David Given
dg@cowlark.com