I have graph traversal algorithm that needs to keep track of the N "best" node visit. Every time a visit a new node I get its goodness along with a ref to it. I then want to compare it to every currently best node in this list and replace it with the worst one if its better than the worst. How do I most easily do this with regards to minimizing GC-usage.

For N = 3 I mean something like this:

(A,1) => [(A,1)]
(B,2) => [(A,1), (B,2)]
(C,3) => [(A,1), (B,2), (C,3)]
(X,0) => [(X,0), (A,1), (B,2)]
(Y,1) => [(X,0), (Y,1), (A,1)]
...

(A,1) means we just visited node with goodness (distance) 1

Nordlöw:

> I have graph traversal algorithm that needs to keep track of the N "best" node visit.

std.algorithm.topNCopy?

Bye,
bearophile

On Wednesday, 25 March 2015 at 13:55:29 UTC, bearophile wrote:
> Nordlöw:
>
>> I have graph traversal algorithm that needs to keep track of the N "best" node visit.
>
> std.algorithm.topNCopy?
>
> Bye,
> bearophile

Could you please elaborate a bit how you mean this should be used. Notice that the number of visited nodes are in the millions or perhaps even tens of millions. And N is typically 100-1000.

On Wednesday, 25 March 2015 at 13:55:29 UTC, bearophile wrote:
> Nordlöw:
>
>> I have graph traversal algorithm that needs to keep track of the N "best" node visit.
>
> std.algorithm.topNCopy?
>
> Bye,
> bearophile

Notice that, ideally, I would like my list of top-nodes to have a fixed size during the whole algorithm (99.9 % of time) as soon it has reached the length of N.

On Wednesday, 25 March 2015 at 14:40:28 UTC, Per Nordlöw wrote:
> On Wednesday, 25 March 2015 at 13:55:29 UTC, bearophile wrote:
>> Nordlöw:
>>
>>> I have graph traversal algorithm that needs to keep track of the N "best" node visit.
>>
>> std.algorithm.topNCopy?
>>
>> Bye,
>> bearophile
>
> Notice that, ideally, I would like my list of top-nodes to have a fixed size during the whole algorithm (99.9 % of time) as soon it has reached the length of N.

What's wrong with binary heaps?

On Wednesday, 25 March 2015 at 17:49:49 UTC, Tobias Pankrath wrote:
> What's wrong with binary heaps?

Ahh, a Binary Heap perfectly matches my needs.

https://en.wikipedia.org/wiki/Binary_heap
http://dlang.org/phobos/std_container_binaryheap.html

Thx

Nordlöw:

> Ahh, a Binary Heap perfectly matches my needs.
>
> https://en.wikipedia.org/wiki/Binary_heap
> http://dlang.org/phobos/std_container_binaryheap.html

But isn't topNCopy using a heap?

Bye,
bearophile