Skip to content

openWFN 0.8.0

Historical release notes. Version 0.8.0 made the 0.8 calculation, result and batch interfaces stable. It combined the two public alphas and tightened the Python API without expanding the scientific validation scope.

Install

python -m pip install --upgrade openwfn

For a reproducible environment, pin the exact version:

python -m pip install openwfn==0.8.0

These notes describe 0.8.0, not the current package. Check openwfn --version and pin the required version when repeating an older analysis.

Consistent Python results

The high-level orbitals(), population(), density(), and analyze_geometry() helpers now return ResultRecord, matching analyze(name). Each result carries the analysis name, status, units, warnings, timing, and source provenance in a consistent envelope.

analyze(name) is the preferred interface for automation. The specialized helpers remain available when discoverability is more useful than a single generic entry point.

Strict input and numerical validation

Public analysis options reject unknown spin channels, density kinds, population methods, and ESP kinds before computation. Molecular coordinates, basis data, matrices, orbital data, grids, and nested result values reject NaN and infinity instead of allowing invalid numbers into scientific output.

Public geometry commands and high-level Python helpers use one-based atom numbers. Internal model arrays and low-level Python functions use zero-based indices; the API reference marks that boundary explicitly.

Installed workflows and batching

The wheel contains the maintained water fixture and provenance record:

openwfn examples install ./openwfn-examples
openwfn ./openwfn-examples/water.fchk summary

Parallel batches retain the bounded, completion-driven queue introduced in 0.8.0a2. Finished inputs are persisted atomically and final manifests remain in deterministic input order. Plotting imports remain lazy for ordinary CLI and batch commands.

The repository benchmark can exercise 10,000 deterministic inputs on a Linux validation host. The maintained command for the Linux validation host is:

python scripts/benchmark_batch.py \
  --count 10000 \
  --workers "$(nproc)" \
  --workspace /tmp/openwfn-benchmark \
  --output openwfn-benchmark-10000.json

Scientific boundary

Capability labels remain authoritative. Methods documented as Validated retain their named evidence scope. Experimental methods are usable but do not carry the same validation claim, and Unsupported combinations are rejected. Gaussian formatted-checkpoint data remains the primary input for full wavefunction analysis.

See the 0.8.0a2 notes and 0.8.0a1 notes for the installation, batching, parser, model, and scientific changes accumulated in this stable release.