Pre-compute node indices to greatly improve performance.

[estasney]

Hello,

Thanks for publishing this add-on. When I first tried it, I thought it had crashed as the UI became unresponsive for 5-10 minutes.

Looking at the code, I noticed a similar pattern:

([node.id_ for node in sorted_nodes]).index(some_vertex)

By itself, this isn't too expensive a computation. However, it was being recreated several times in each for loop. On my not too complex model, this was computed 250,000+ times.

Since we only really need to do this once, creating a dictionary of:

{node.id_: index}

ahead of time, allows the export to run < 1 second.

I tested the output jbeam file with md5sum and the outputs were equivalent.

Note - This assumes that node.id_ is unique. I.e this will return different values if they are not unique:

# Index Method
node = 1
sorted_nodes = [1, 1, 3, 1]
>>> sorted_nodes.index(node)
>>> 0

# Dictionary Method
node = 1
sorted_node_idx = {x: i for i, node in enumerate(sorted_nodes)}
# {1: 3, 3: 2}
>>> sorted_node_idx[node]
>>> 3

[50thomatoes50]

Thank you very much for the fix.