Re: filling area by color atack safety

On 19/03/2024 06:10, Tim Rentsch wrote:

Malcolm McLean <malcolm.arthur.mclean@gmail.com> writes:
 

On 18/03/2024 18:36, Tim Rentsch wrote:

>
>
It doesn't scale well.  In particular worst case performance
scaling is worse than O(N) (as determined experimentally, not
theoretically).

>
Is that because the queue is being memmoved instead of using a
circular buffer when it gets towards the end?

 I'm sure I don't know, and I'm astonished that you would ask.
It's your code after all.  IMO it should simply be thrown out and
re-written;  it pains me just to look at it, let alone to try to
understand or fix it.

Yes, but I wrote it years ago. I can't remember why the value of 256 pixels was put in. But I do remember why the queue isn't very efficient - for the small images I expected to be processed, I judged that the added complexity of maintaining a circular queue wouldn't be worth it, given that I wanted the routine to be leaf. However if you can somehow trigger catastrophic big O behaviour, it won't be a surprise.
--
Check out Basic Algorithms and my other books:
https://www.lulu.com/spotlight/bgy1mm