Is it noticably more efficient to cache direct-memory arrays such as ts.nodes_time?

[hyanwong]

I'm quite often looking up node times in a tree sequence, which can be accessed using ts.nodes_time. If I'm e.g. accessing it 4 or 5 times, am I better saving it as a variable, or is ts.nodes_time almost as efficient:

parent_times = ts.nodes_time[ts.edges_parent]
child_times = ts.nodes_time[ts.edges_child]

# Or is this a lot more efficient?
nodes_time = ts.nodes_time
parent_times = nodes_time[ts.edges_parent]
child_times = nodes_time[ts.edges_child]

[hyanwong]

ts.nodes_time provides zero copy access, but there are a few function calls involved, including an attribute accessor, so it is likely that the second incantation will be more efficient, but not by much, unless it's inside a tight loop.

[jeromekelleher]

There wouldn't be much in it here.

It does make a substantial difference when accessing table columns, which superficially looks the same but involves full copies.