Refactor make_downshift_arrays and enhance the performance of topology construction

[yuxuanzhuang]

Fixes #3721

Changes made in this Pull Request:

• refactor make_downshift_arrays
• lazy building Universe._topology.tt. _RA, _SR during construction and serialization.

PR Checklist

• Tests?
• Docs?
• CHANGELOG updated?
• Issue raised/referenced?

[pep8speaks]

Hello @yuxuanzhuang! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2022-07-25 08:23:04 UTC

[codecov]

Codecov Report

Merging #3724 (267dcb6) into develop (886cb60) will decrease coverage by 0.01%.
The diff coverage is 100.00%.

❗ Current head 267dcb6 differs from pull request most recent head b6c42c6. Consider uploading reports for the commit b6c42c6 to get more accurate results

@@             Coverage Diff             @@
##           develop    #3724      +/-   ##
===========================================
- Coverage    94.31%   94.30%   -0.02%     
===========================================
  Files          192      192              
  Lines        24757    24713      -44     
  Branches      3340     3328      -12     
===========================================
- Hits         23350    23306      -44     
  Misses        1359     1359              
  Partials        48       48              

Impacted Files	Coverage Δ
package/MDAnalysis/core/topology.py	99.44% <100.00%> (-0.56%)	⬇️
package/MDAnalysis/coordinates/PDB.py	94.85% <0.00%> (-0.28%)	⬇️
package/MDAnalysis/lib/util.py	90.84% <0.00%> (-0.14%)	⬇️
package/MDAnalysis/lib/distances.py	98.48% <0.00%> (-0.01%)	⬇️
package/MDAnalysis/core/groups.py	98.58% <0.00%> (ø)
package/MDAnalysis/topology/PDBParser.py	100.00% <0.00%> (ø)
package/MDAnalysis/topology/ExtendedPDBParser.py	100.00% <0.00%> (ø)
package/MDAnalysis/core/topologyattrs.py	96.65% <0.00%> (+0.06%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 886cb60...b6c42c6. Read the comment docs.

[yuxuanzhuang]

Performance Improvement

from MDAnalysis.tests.datafiles import PDB
import MDAnalysis as mda
from MDAnalysis.core.topology import (
    TransTable,
    make_downshift_arrays,
)

u = mda.Universe(PDB)
n_atoms = u._topology.tt.n_atoms
# 47681 atoms
n_residues = u._topology.tt.n_residues
# 11302 residues
n_segments = u._topology.tt.n_segments
# 1 segment
AR = u._topology.tt._AR
RS = u._topology.tt._RS

%timeit make_downshift_arrays(AR, n_residues)
# original
29.6 ms ± 254 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)
# now
12.1 ms ± 73.9 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

%timeit TransTable(n_atoms, n_residues, n_segments, AR, RS)
# original 
29.8 ms ± 329 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)
# now
17 µs ± 57.4 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)

tt = TransTable(n_atoms, n_residues, n_segments, AR, RS)
%timeit pickle.loads(pickle.dumps(tt))
# original
29.6 ms ± 1.27 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)
# now
69.7 µs ± 740 ns per loop (mean ± std. dev. of 7 runs, 10,000 loops each)

[richardjgowers]

thanks @yuxuanzhuang , this looks great

[IAlibay]

@richardjgowers this is approved but unmerged, is it good to go?

[orbeckst]

@richardjgowers I assigned the PR to you — please merge if all is good. Thanks!