[SchrodingersCattt]

Summary

Explicit --style ball_stick on conventional ionic / inorganic unit cells draws coordination sticks on the same wrapped lattice-site layout as #243: centres stay in [0, 1] (crystallographic sites plus the symmetric face, edge, and corner images at 0 and 1). There is no one-sided coordination shell at fractional -0.5.

Each home site still has its first-neighbor coordination (CN=6 along ±a/2 for rocksalt NaCl). A neighbor with no sphere inside the closed cell is a stick ending at that image coordinate. Those exterior ends are closed under inversion through the cell centre and under cubic axis permutation, so a blob on one corner fails the regression. Atom spheres stay larger than the sticks.

Ball and space-filling styles do not stroke those contacts. When --style is omitted, ionic and inorganic lattices still use ball and unit_cell.

Based on the lattice-site layout now on main from #243 (bc2515f).

Closes #238

Acceptance

ASE bulk('NaCl', 'rocksalt', a=5.6402, cubic=True) written as a VASP POSCAR:

mat-vis render POSCAR_NaCl -o nacl_bs.png \
  --backend cpu --orthogonal --background '#FFFFFF' --style ball_stick --json

Centered rocksalt NaCl ball-and-stick unit cell

Tests

• tests/render/test_ionic_coordination_sticks.py checks CN=6, equal axis counts, centres inside the closed cell, and inversion plus cubic symmetry of both centres and stick endpoints.
• The two-atom ionic NaCl benchmark uses the 12-atom symmetric 0/1 layout and keeps Na–Cl contacts visible.
• The ball unit-cell checks still require the 27-site figure and no bond primitives when the representation is ball.

_{To show artifacts inline, enable in settings.}

  

[SchrodingersCattt]

Review @ 927a95fa (pending CI)

Looks right for #238.

• Ionic contacts stay in scene["bonds"] (suppressed_ionic_bond_count == 0); ball/space_filling still don’t stroke them.
• Half-cell MIC tie in _bond_endpoints keeps both ±a/2 arms of the octahedron instead of doubling one image.
• coordination_stick_radius + lattice ball radii give spheres larger than sticks; omitted --style → ball for ionic/inorganic matches the viewer.
• Regression suite in tests/render/test_ionic_coordination_sticks.py (CN=6, equal axes, omitted-style default) is solid.
• PR body keywords are safe (Closes #238; “does not address #239”).

Merge note vs open #243: both touch scene/core.py and tests/test_inorganic_defaults.py. #244 still expects CIF NaCl draw_atoms == 4 (spanning-bond shell); #243 rewrites that path to wrapped lattice sites (12). Prefer merge #243 first, then rebase #244 onto main so the inorganic defaults assertion follows the new site contract.

Waiting on GHA 9/9. Will not squash until you say so.

[SchrodingersCattt]

User feedback (needs-changes): the acceptance PNG is still a one-sided corner blob (structure clustered bottom-left; top-right of the cell empty). That is not acceptable.

Wanted instead:

1. Box-centered / lattice-site layout (same contract as Keep ionic lattice ball centres inside the conventional cell #243: wrapped sites + symmetric 0/1 replicas; centres in [0, 1]), or
2. If anything protrudes outside the wire, protrusions at all symmetry-equivalent places — not one corner only.

CN=6 octahedral sticks, sphere>stick radii, and omitted---style→ball stay. Follow-up sent to the cloud agent: rebase onto #243’s lattice_site_display path and re-render.

[SchrodingersCattt]

Re-review @ a712c652 (pending CI) — addresses the one-sided corner blob.

• Acceptance PNG is now box-centered: wrapped sites + symmetric 0/1 replicas; sticks to exterior image ends appear on every symmetry-equivalent face.
• Tests assert centres in [0, 1], inversion + cubic closure for both centres and stick endpoints, CN=6, and no framework/bonded-image shell.
• Branch includes #243’s lattice-site commit; land #243 first, then this.
• Keywords OK (Closes #238; dependency on #243 stated without auto-close phrasing).

Waiting on GHA 9/9. Will not squash until you say so.

[SchrodingersCattt]

Closing without merge: owner review found the change rocksalt-specific (NaCl octahedral CN=6 tests/acceptance/comments; stick geometry comments hard-coded around rocksalt). Redo as a general ionic lattice ball_stick coordination-sticks PR on main (post-#243).