Improve performance of ShortNameComparator.

[jdcove2]

This is an updated version of #135 .

[drivenflywheel]

Do we think the performance improvements will be primarily speed, memory utilization, or both?

[ldhardy]

Is there a trick to recreate the profile you describe in #135 ? I've been trying to use the profiler on our main branch to observe the described behavior and I'm not having any luck. I guess there have to be a lot of children extracted - I tried dumping in all kinds of archive input to see if I can see something similar on the profiler, but I'm not observing what you were.

[jdcove2]

Yes, this optimization is designed for objects with a large number of children (i.e. 100,000 or more). Let me know if you would like me to generate an object for you.

[ldhardy]

Yes, this optimization is designed for objects with a large number of children (i.e. 100,000 or more). Let me know if you would like me to generate an object for you.

Nah, I believe you. It's not that big of a change really.

[jdcove2]

Based on the profiling results in the original PR, this change should improve both speed and memory usage as in the normal case of sorting numerical children it creates virtually no objects.

[ldhardy]

Yes, this optimization is designed for objects with a large number of children (i.e. 100,000 or more). Let me know if you would like me to generate an object for you.

I meant the algorithm is basically the same, but yes, I can see how it's a less heavy handed implementation.