Pass optimized ASE atoms through run_thermo - #52
Merged
Conversation
dftbplus_thermo optimized the geometry, obtained the ASE Atoms via geoopt.read(), then discarded it and called run_thermo(coord_file="geo_opt.xyz") — a file DFTB+ never writes (it emits geo_opt.gen). Add an optional atoms argument to run_thermo that builds the Atom list, cell and pbc directly from an ASE Atoms object, and wire dftbplus_thermo to use it. File-based inputs are unchanged when atoms is not given.
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #52 +/- ##
==========================================
+ Coverage 93.83% 94.27% +0.44%
==========================================
Files 21 21
Lines 1200 1206 +6
==========================================
+ Hits 1126 1137 +11
+ Misses 74 69 -5
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
galjos
added a commit
that referenced
this pull request
Jun 26, 2026
Closes #53. Stacked on #52 (retargets to main once that merges) — it depends on #52 because the pipeline now runs on the in-memory optimized atoms rather than re-reading a file from the working directory. The DFTB+ steps read and write fixed filenames (`geo_opt.gen`, `hessian.out`, `vibrations.tag`, `modes_in.hsd`, ...) in the current directory, so running more than one molecule in the same directory clobbers the previous molecule's geometry/Hessian/frequencies. This blocks any batch/screening loop (#33). ## Change - Add a `directory` argument to `dftbplus_thermo`; the pipeline runs there via a `_run_in_directory` context manager that creates the directory and restores the previous working directory afterwards (including on error). - Backward-compatible: defaults to the current directory, so existing single-molecule workflows are unchanged. - chdir-based, so correct for sequential and multi-process batches; per-thread parallelism within one process would need the calculators to thread a directory through (follow-up). ## Tests - Unit tests for `_run_in_directory`: isolates into the target dir and restores the cwd, restores on exception, and is a no-op for `None`. - Verified locally with the real DFTB+ binaries: two molecules run into separate directories, the cwd is untouched, and results are independent and physically sensible (H2O 45.05, NH3 45.97 cal/(mol*K)). 148 passed, 0 skipped (suite run with DFTB+ binaries); pylint 8.04/10.
galjos
added a commit
that referenced
this pull request
Jun 26, 2026
Closes #53. Re-targets the work from #54 onto main: #54 was stacked on the #52 branch and got merged into that branch instead of main (which was squash-merged separately), so the change never landed. These are the same commits, cleanly rebased onto current main. The DFTB+ steps read and write fixed filenames (`geo_opt.gen`, `hessian.out`, `vibrations.tag`, `modes_in.hsd`, ...) in the current directory, so running more than one molecule in the same directory clobbers the previous molecule's geometry/Hessian/frequencies. This blocks any batch/screening loop (#33). ## Change - Add a `directory` argument to `dftbplus_thermo`; the pipeline runs there via a `_run_in_directory` context manager that creates the directory and restores the previous working directory afterwards (including on error). - Backward-compatible: defaults to the current directory, so existing single-molecule workflows are unchanged. - chdir-based, so correct for sequential and multi-process batches; per-thread parallelism within one process would need the calculators to thread a directory through (follow-up). ## Tests - Unit tests for `_run_in_directory`: isolates into the target dir and restores the cwd, restores on exception, no-op for `None`. - A mocked-pipeline test drives `dftbplus_thermo(..., directory=...)` without the binaries, covering the directory wiring in CI (api.py coverage 87% -> 93%). - Verified locally with the real DFTB+ binaries: two molecules run into separate directories, the cwd is untouched, results independent and physically sensible (H2O 45.05, NH3 45.97 cal/(mol*K)). 144 passed, 5 skipped (no binaries); pylint 8.04/10.
galjos
added a commit
that referenced
this pull request
Jun 29, 2026
Implements the core screening framework from #33 (first slice). Builds on the per-job directory isolation (#55) and ASE-atoms passthrough (#52). ## What it does A \`screen()\` entry point (exported as \`ThermoScreening.screen\`) and a \`thermo screen\` CLI subcommand that, for each molecule, run the DFTB+ geometry-optimisation/Hessian/modes pipeline in its own working directory and collect the key thermochemistry into \`results.csv\` and \`results.json\`. - **Input (both):** a directory of \`.xyz\`/\`.gen\` structures (run at a global \`--charge\`), or a CSV manifest with \`path\` plus optional \`name\`/\`charge\` columns (per-molecule charge — e.g. anthraquinone at -1/-2). - **Output (both):** a CSV summary table and a JSON file (name, path, charge, status, E/H/G in Hartree, S/Cv in cal/(mol*K), error). - **Error isolation:** a molecule that fails is recorded with \`status=error\` and does not abort the run. - **Per-job isolation:** each molecule runs in \`<directory>/<name>\`. ``` # Python from ThermoScreening import screen screen("molecules.csv", out="results") # CLI thermo screen ./structures/ -o results --charge 0 thermo screen molecules.csv -o results ``` ## Tests - Job loading from a directory and a CSV manifest (per-row + default charge, relative path resolution), and rejection of empty/unknown inputs. - \`screen()\` with the pipeline mocked: collects results, writes CSV + JSON, runs each molecule in its own directory, and isolates a failing molecule. - CLI routing for the \`screen\` subcommand and the failure-count return code. - Verified locally with the real DFTB+ binaries over a directory of CH4/H2O/NH3 plus a deliberately broken structure: the three succeed (entropies 44.44 / 45.05 / 45.97 cal/(mol*K), matching literature), the broken one is isolated as an error. 152 passed, 5 skipped (no binaries); pylint 8.09/10. ## Deferred to follow-ups (not in this PR) - Resume/checkpoint (this PR always recomputes, per the agreed scope). - Parallel execution (chdir-based isolation is correct for sequential and multi-process batches; per-thread parallelism would need the calculators to thread a directory through).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #34.
dftbplus_thermoalready optimizes the geometry and gets the result as an ASE Atoms object (optimized_atoms = geoopt.read()), but then discarded it and calledrun_thermo(coord_file="geo_opt.xyz")— a file DFTB+ never writes (it emitsgeo_opt.gen). The disk round-trip was both wasteful and fragile.Change
atomsargument torun_thermo; when given, it builds the Atom list, cell and pbc directly from the ASE Atoms object (_atoms_from_ase) instead of readingcoord_file.dftbplus_thermoto passoptimized_atomsdirectly.atomsis not provided, so existing workflows keep working.Tests
Equivalence test on the real
geo_opt.gen+frequency.txtfixtures: the ASE-atoms path and the file path produce identical total energy, entropy, and Gibbs free energy.145 passed, 0 skipped (suite run with DFTB+ binaries); pylint 7.99/10.