Summary

This PR speeds up the instantiation of Structure objects by preventing hash collisions in the lru_cache of get_el_sp and increasing its maxsize. The issue is that currently Element objects are hashed to the same value as the integer atomic numbers (e.g., Element[H] maps to the same hash as int(1)). This forces the lru_hash to perform an expensive __eq__ comparison between the two, which reduces the performance of instantiating many Structure objects. Also here we increase the maxsize of get_el_sp's lru_cache to 1024 for further performance improvements.

This reduces time taken to instantiate 100,000 Structure objects from 31 seconds to 8.7s (avoid hash collisions) to 6.1s (also increase maxsize to 1024). Test snippet:

from pymatgen.core import Structure
import numpy as np
from time import time

t = time()
for _ in range(100_000):
    struc = Structure(
        coords = np.random.rand(10, 3),
        species = np.random.randint(1, 100, size=(10,)),
        lattice = np.random.rand(3, 3),
        coords_are_cartesian = False,
    )
print("Time taken to create 100,000 structures:", time() - t)

Unless I'm missing some deep reason for why the hash of Element objects should be the same as their atomic number hashes, this should not lead to any different behavior of the code.

Major changes:

• Update ElementBase __hash__ function to map to a different value than the atomic number
• Increase the maxsize of get_el_sp's lru_cache to 1024

Checklist

• [ x ] Google format doc strings added. Check with ruff.
• [ x ] Type annotations included. Check with mypy.
• Tests added for new features/fixes.
• [ x ] If applicable, new classes/functions/modules have duecredit @due.dcite decorators to reference relevant papers by DOI (example)

Tip: Install pre-commit hooks to auto-check types and linting before every commit:

pip install -U pre-commit
pre-commit install