Speedup identify_connections

[daico007]

The current Topology.identify_connections method is quite slow for big system, this PR made several changes to the gmso/utils/connectivity.py to improve the performance of said method (change the comparison of site to just their index, use IndexedSet instead of list, etc).

[codecov]

Codecov Report

Patch coverage: 100.00% and no project coverage change.

Comparison is base (727d32a) 91.99% compared to head (57aa644) 91.99%.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #741   +/-   ##
=======================================
  Coverage   91.99%   91.99%           
=======================================
  Files          67       67           
  Lines        6460     6460           
=======================================
  Hits         5943     5943           
  Misses        517      517           

Impacted Files	Coverage Δ
gmso/utils/connectivity.py	98.38% <100.00%> (ø)

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[chrisjonesBSU]

Speeding this up will be great. Thanks for doing this!

Have you bench marked it yet to get an idea of the performance improvement?

[daico007]

I haven't but I can sweep up something

[CalCraven]

Ran this code locally on a system.

import gmso
import sys
sys.path.append("..")
from basic_testing import gmso_water_graph, box_water, box_polymer
import timeit
import numpy as np
setupStr = """
import gmso;
import sys;
sys.path.append("..");
from basic_testing import gmso_water_graph, box_polymer;
n_mols=100;
poly_box = box_polymer(n_mols);
gmso_top = gmso_water_graph(poly_box);
"""
outList = timeit.repeat("gmso_top.identify_connections()", setup=setupStr, number=3, repeat=3)
print(f"Identify {gmso_top.n_angles + gmso_top.n_dihedrals} Connection in"
      + f" {np.mean(outList):.2f} +- {1.96*np.std(outList)/np.sqrt(3):.3f} seconds"
)

Output with this PR

Identify 1491 Connection in 0.92 +- 0.002 seconds

Output with current GMSO main

Identify 1491 Connection in 1.24 +- 0.002 seconds

Which is a a 25% speedup for this system

[CalCraven]

LGTM. Did a little plotting as well to verify locally, and I think these changes seem even better at larger number of connections.

[chrisjonesBSU]

LGTM. The performance improvement looks awesome, thanks @daico007!

[daico007]

I will merge this after all test passes, thanks for doing the profiling Cal!