I would prefer to keep it like we have it now. A barrier size of 1000 mean doesn't necessarily mean, that we will get a lot of traversers. In fact, it will mostly be way fewer in a good traversal executed on a well-modeled graph. Thus I'd say let's better keep the memory footprint low. However, I might underestimate the cost of resizing, so this is purely based on my opinion / gut feeling, not on any kind of benchmarks.
Cheers, Daniel On Thu, Mar 1, 2018 at 4:36 PM, Stephen Mallette <spmalle...@gmail.com> wrote: > I came across this JIRA: > > https://issues.apache.org/jira/browse/TINKERPOP-1503 > > which wonders about the initial size of TraverserSet instances for a given > barrier size. Anyone have any opinion about how to size that? >