Methods for iterators

[diamog]

Our old method of iterators comes from PUMI. For example here:
https://github.com/SCOREC/EnGPar/blob/master/test/NgraphTestSuite.cpp#L276-L284

For all iterators besides vertex iterators, this requires allocating memory and subsequently destroying the memory.

A new method came to me when writing the second adjacency iterator that avoids the memory allocation like here:
https://github.com/SCOREC/EnGPar/blob/master/test/NgraphTestSuite.cpp#L432-L437

I'm not exactly sure that the other iterators could be converted this method or not, but if they can be we can avoid these memory allocations. Is there any benefit I'm missing from the old version? Is there a negative in this new method that I'm missing?

This would be a big change internally so there is no rush on decisions.

[cwsmith]

This seems fine to me. How does this new interface avoid allocating memory?

[diamog]

The pointers can be maintained the same way that vertex iterators are done. Convert an integer to a pointer and just edit the number as needed. The EdgeIterator and PinIterator are classes that require dynamic allocations and thus destruction to use in their current state.