fix(explorer): absolutize spliced stat-mech file paths to the source's directory

[alongd]

Stacked on #200 → #199 → #198 → #197. Base is i020-generated-modelchem, so this diff shows only its own change; GitHub retargets as each parent merges. Review in order.

What this fixes

T3's hybrid writer references a QM transition state's stat-mech data by a relative path — transitionState('TS2', 'qm/TS2.py'), with the qm/ tree vendored beside the hybrid. That convention is deliberate and keeps a run directory portable. But input_file.py splices that line verbatim into a generated Arkane input in a different directory, and Arkane then cannot find the file:

FileNotFoundError: [Errno 2] No such file or directory: 'qm/TS2.py'
  at arkane/statmech.py:372

The mechanism is worth stating precisely, because the obvious reading is wrong. arkane/input.py:690 does job.path = os.path.join(directory, job.path) — Arkane does rebase, against the input file's directory. That is exactly the problem: the input file is the generated explorer input, not the hybrid, so a path meaningful relative to the hybrid is rebased against a directory with no qm/ tree in it.

The change

Each spliced species() / transitionState() external file path — Arkane's <name>(label, path) form — is rewritten from relative to absolute, anchored at the source file's own directory, where the vendored qm/ tree actually lives. Adds _stat_mech_file_path_node, an edit in the existing AST walk, and a stat_mech_paths_absolutized record on ExplorerInputSummary.

The hybrid writer's relative convention is untouched — only the spliced copy is rewritten, so hybrids stay portable. A string literal remains a string literal, so neither the inbound source guard nor the outbound self-check is weakened, and there is no chdir.

Position-dependence sweep

Since this is the second defect caused by the same verbatim splice moving text into a new context, the other position-dependent references were enumerated rather than assumed:

• species() / transitionState() external paths — the defect; fixed for both call names.
• Log('logs/TS2/...') inside qm/TS2.py — a different level. statmech.py:319 rebases those against the species file's own directory, which is now correct, so they resolve. Confirmed by the run reading them.
• database() carries library names, not paths.

Testing

• Driven on the real artifact. The emitted line is now transitionState('TS2', '/abs/path/to/qm/TS2.py'); Arkane loads the transition state and its logs and proceeds to the master-equation solve. No cluster, no quantum chemistry.
• test_real_hybrid_transition_state_path_is_absolutized_to_an_existing_file asserts os.path.isfile(emitted_path) — that it resolves to a file that exists, not merely that it is absolute.
• The real qm/TS2.py is committed beside the existing hybrid fixture so that assertion is meaningful.
• Mutation: neutralising the absolutization branch turns it red; restored, green.
• tests/test_pdep: 1930 passed, 133 s.

Known next failure, not chased

The run now gets all the way to the physics and stops there: numpy.linalg.LinAlgError: Singular matrix at rmgpy/pdep/msc.pyx:147, in the Modified Strong Collision solve. That is a numerical/scientific issue rather than a plumbing one, and it is tracked separately. With this change, the round-0 → round-1 handoff no longer fails on any file-contract defect.

[Copilot]

Pull request overview

This PR fixes explorer handoffs by converting relative stat-mechanics paths to absolute source-relative paths when generating inputs.

Changes:

• Absolutizes species() and transitionState() external paths.
• Records rewritten paths in ExplorerInputSummary.
• Adds regression coverage and a real stat-mechanics fixture.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Summary
t3/pdep/explorer/input_file.py	Rewrites and records stat-mechanics paths. Moderate issue: positional-label species(label, path) forms are rewritten but not fully supported by species registration and bath-gas handling.
tests/test_pdep/test_explorer_input_file.py	Tests path absolutization and resolution to an existing file.
tests/data/pdep_hybrid/cho2_round0/qm/TS2.py	Adds the referenced stat-mechanics fixture.

Suppressed comments (1)

t3/pdep/explorer/input_file.py:367

• This new branch explicitly promises to absolutize both species() and transitionState() file references, but the regression coverage only drives the latter. Please add an end-to-end case for the positional species('S', 'qm/S.py') form (and assert its summary record/path), so a species-specific argument-handling regression cannot pass unnoticed.

        if call_name in ('species', 'transitionState'):
            path_node = _stat_mech_file_path_node(call)
            if path_node is not None and not os.path.isabs(path_node.value):
                absolute_path = os.path.join(source_dir, path_node.value)

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.