0.26.0 / 2023-07-31
- (GH#308) Fix CI Pipelines. Dropped Python3.6. Bring CI pipelines into harmony with local dev experience. Lint and format entire code base. Accelerate CI pipelines. Update setup.py to correctly define extras packages. Breaking change due to dropped support for Python3.6. No code functionality was altered.
Dropped support for dead Python 3.6. Minimum supported Python is now 3.7.
Updated CONTRIBUTING.md to contain detailed instructions for developers on how to contribute.
Fixed broken code that failed to prepend the “v” to version numbers.
Updated CI to run without conda and using only packages defined in setup.py. CI is now much faster and runs the same way for local developers and GitHub Actions.
Added test.sh and format.sh to devtools/scripts for easy execution of formatting and testing.
Formatted all code with black. Sorted imports with isort.
Added pre-commit to repo so code formatting, linting, and testing will all run as part of regular git workflow.
- (GH#310) Modernize DevOps Tooling
Added /scripts directory to root of project that contains scripts for testing, formatting code, and building docs.
Updated build system from setuptools to modern pyproject.toml specification using poetry for the build backend.
Removed complicated versioning code in favor of single source of truth in pyproject.toml. Using standard library importlib for looking up package version in __init__.py file.
Added build_docs.sh script to /scrips and removed Makefile from /docs. Flattened /docs file structure.
Removed travis-ci code from devtools
Removed LGTM code (they no longer exist as a project).
Bring all package directories under black, isort, and autoflake control.
(GH#305) Initialize Molecule.extras as empty dictionary.
(GH#311) Update docs location from RTD to GH pages. Resolve escape char warnings. Update changelog.
(GH#311) Clear up NumPy “Conversion of an array with ndim > 0 to a scalar is deprecated” in util.measure_coordinates called by Molecule.measure.
(GH#314) Import pydantic.v1 from pydantic v2 so that QCElemental can work with any >=1.8.2 pydantic until QCElemental is updated for v2.
0.25.1 / 2022-10-31
(GH#297) Rearrange imports for compatibility with Pint v0.20. No new restrictions on pint version.
0.25.0 / 2022-06-13
(GH#285) Standardized default on
None. May break strict logic.
(GH#289, GH#290) Transition from some early documentation tools (class
qcarchive_sphinx_theme) to externally maintained ones (project https://github.com/mansenfranzen/autodoc_pydantic and
sphinx_rtd_theme). Expand API docs.
(GH#286) Sphinx autodocumentation with typing of
qcelemental.testing.compare_recursiveno longer warns about circular import.
0.24.0 / 2021-11-18
native_filesfield of a dictionary of file names (or generic
'input') and text (not binary) contents that the CMS program may have generated but which haven’t necessarily been harvested into QCSchema. Contents controlled by a new native_files protocol analogous to stdout protocol.
TorsionDriveInputnow accepts a list of
Moleculeobjects as the
initial_moleculeto seed the torsiondrive with multiple conformations.
0.23.0 / 2021-09-23
AtomicResultProperties.dict()no longer forces arrays to JSON flat lists but now allows NumPy arrays. That is,
AtomicResultPropertiesnow behaves like every other QCElemental model. Expected to be disruptive to QCFractal.
(GH#280) Examples of QCSchema from test cases are now saved at branch https://github.com/MolSSI/QCElemental/tree/qcschema-examples . These have passed validation as models by Pydantic and as JSON by schema generated from Pydantic models.
(GH#277) Documentation is now served from https://molssi.github.io/QCElemental/ and built by https://github.com/MolSSI/QCElemental/blob/master/.github/workflows/CI.yml . https://qcelemental.readthedocs.io/en/latest/ will soon be retired.
(GH#274) The molecule
from_stringparser when no dtype specified learned to return the most specialized error message among the dtypes, not the full input string.
Molecule.to_string(..., dtype="nwchem")learned to handle ghosts (
real=False) correctly. It also now prints the user label, which is used downstream for custom basis sets and shows up in a NWChem output file. QCEngine will be edited to process the label, but other uses may need modification.
Molecule.alignlearned a new keyword
generic_ghosts=Trueso that it can act on molecules that have centers with content Gh, not Gh(He).
Molecule.to_string(..., dtype="gamess")learned to handle ghosts (
real=False) correctly for
coord=unique. Note that QCEngine uses
coord=prinaxis, so actual ghosts are still NOT interpretable by downstream GAMESS.
0.22.0 / 2021-08-26
Moleculelearned to create instances with geometry rounded to other than 8 decimal places through
Molecule(..., geometry_noise=<13>)to optionally override
qcel.models.molecule.GEOMETRY_NOISE = 8. This should be used sparingly, as it will make more molecules unique in the QCA database. But it is sometimes necessary for accurate finite difference steps and to preserve intrinsic symmetry upon geometry rotation. Previous route was to reset the qcel module variable for the duration of instance creation.
Molecule.scramblelearned to return a fuller copy of self than previously. Now has aligned atom_labels, real, and mass_numbers as well as incidentals like Identifiers. Fragmentation still not addressed.
Molecule.to_string(dtype="gamess")learned to write symmetry information to the prinaxis output if passed in through field fix_symmetry. This is provisional, as we’d like the field to be uniform across qcprogs.
(GH#271) Testing function
compare_values()on arrays corrected the RMS maximum o-e value displayed and added a relative value.
0.21.0 / 2021-06-30
(GH#267) Serialization learned msgpack mode that, in contrast to msgpack-ext, doesn’t embed NumPy objects.
(GH#266) Testing function
compare_values()learned to print RMS and MAX statistics for arrays.
(GH#265) Fix search path construction. @sheepforce
(GH#266) Bump minimum pydantic to 1.8.2 to avoid security hole – https://github.com/samuelcolvin/pydantic/security/advisories/GHSA-5jqp-qgf6-3pvh .
0.20.0 / 2021-05-16
PhysicalConstantsContextlearned to write a Fortran header. @awvwgk
(GH#261) Documentation became type-aware and grew more links.
0.19.0 / 2021-03-04
(GH#253) Fixed incompatibility with Pydantic >=1.8.
0.18.0 / 2021-02-16
(GH#237) Exports models to JSON Schema with
(GH#237) Build bank of JSON examples from Pydantic models defined in tests. Test that bank against exported schema with
pytest --validate qcelemental/.
(GH#237) Many model descriptions edited, dimensions added to array properties,
BasisSetobject not just string, several properties added to match QCSchema, several limitations on number and uniqueness imposed.
0.17.0 / 2020-10-01
molparse.to_stringMRChem dtype developed.
0.16.0 / 2020-08-19
compare_recursivelearned new keyword
-computedto pass the comparison. For
compare_recursive, the leniency can be restricted to specific leaves of the iterable by passing a list of allowed leaves.
Molecule.aligntold the full truth in its documentation that the
mol_alignargument can take a float.
0.15.1 / 2020-06-25
(GH#228) Fix testing bug for installed module, which was missing a dummy directory.
0.15.0 / 2020-06-25
(GH#182) Added experimental protocol for controlling autocorrection attemps. (That is, when a calculation throws a known error that QCEngine thinks it can tweak the input and rerun.) Currently in trial for NWChem.
which_importto distinguish between ordinary and namespace packages.
(GH#227) Add non-default
to_elementthat fails when isotope information is given.
(GH#227) Allow nonphysical masses to pass validation in
molparse.from_schema(..., nonphysical=True). Also allowed in forming
(GH#227) Fixed deception described in issue 225 where
qcel.models.Molecule(..., symbols=["O18"])accepted “O18” but did not influence the isotope, as user might have expected. That now raises
NotAnElementError, and an example of correctly setting isotope/masses has been added. This error now caught at
0.14.0 / 2020-03-06
0.13.1 / 2020-02-05
(GH#209) Added the option to Hill-order molecular formulas.
(GH#208) Fixes a Molecule hashing issue due to order of operations changes in
Molecule.from_data. The order of operations changed in
Molecule.from_dataand occasionally resulted in different hashes for Molecules undergoing orient operations. This issue was introduced in 0.13.0 and is unlikely to have any serious negative effects as this did not affect hash integrity.
0.13.0 / 2020-01-29
(GH#179, GH#181) QCElemental works with Python 3.8 at the expense of needing a new 0.10 pint (rather than generic install). Pint 0.10 has optional NumPy dependency of >=1.12.0, so QCElemental that requires both NumPy and pint needs this constraint.
0.12.0 / 2019-11-13
Moleculescan now be correctly compared with
molparse.to_stringQ-Chem dtype developed. Psi4 dtype now includes label and doesn’t have extraneous info for single fragment systems.
(GH#162) New protocol
stdoutfield (which generally contains the primary logfile, whether a program writes it to file or stdout) is returned.
(GH#165) The code base is now Black formatted.
0.11.1 / 2019-10-28
- (GH#152) Patches
Molecule.from_datato read XYZ+ format and correctly handle keyword arguments.
Molecule.to_fileto write XYZ+ format as the default for XYZ and XYZ+ files.
- (GH#152) Patches
0.11.0 / 2019-10-24
which_importto correctly handle submodules.
0.10.0 / 2019-10-16
0.9.0 / 2019-10-01
(GH#139) A wavefunction property have been added to the
ResultModel. Adds the ability for Engine and other programs to store and fetch wavefunction data.
Protocolshave been added to
QCInputSpecificationwhich allows data to pre-pruned by different specifications. Main intention is to reduce wavefunction data which may be re-computed cheaply rather than storing all of it. This does change the input model, so requires a minor version bump.
Result’s documentations have been brought up to the standards of other models.
0.8.0 / 2019-09-13
(GH#125) Add back a consistency check that had been optimized out.
0.7.0 / 2019-08-23
(GH#118) Model string representations should be more user friendly and descriptive without overload the output.
return_datanow returns molecule keywords for GAMESS and NWChem. The
(GH#120) Auto documentation tech is now built into the
ProtoModeland does not need an external function.
0.6.1 / 2019-08-19
(GH#114) The Numpy einsum calls reference the top level functions and not core C functions. This fixes an issue which can result in NumPy version dependencies.
0.6.0 / 2019-08-14
(GH#85, GH#87) Msgpack is a new serialization option for Models. Serialization defaults to msgpack when available (
conda install msgpack-python [-c conda-forge]), falling back to JSON otherwise. This results in substantial speedups for both serialization and deserialization actions and should be a transparent replacement for users within Elemental itself.
(GH#78) Molecular alignments can now be aligned on the derivatives of vector components.
(GH#81) Testing is now operated both on the minimal supported and the latest released versions of dependencies.
(GH#82) Molecule fragment grouping is now disabled by default to match expected behavior.
(GH#84) Testing without internet connection should now pass since PubChem testing is skipped with no connection.
(GH#85) Molecule switches from lists to numpy arrays for internal storage of per-atom fields.
(GH#86) Molecule performance and memory enhancements through reduced validation times and LRU caching of common validations.
(GH#87) Molecule objects built from Schema are run through validators for consistency.
0.5.0 / 2019-07-16
(GH#76) Adds a built-in
Molecule.to_filefunction for easy serialization into
(GH#74) Atom and fragment ordering are preserved when invoking
0.4.2 / 2019-06-13
0.4.1 / 2019-05-31
return_datathat contains aspects of the
models.Moleculenot expressible in the string. Implemented for dtypes xyz, cfour, psi4.
Datumgained an attribute
numericthat reflects whether arithmetic on
Datums that aren’t numeric can now be created by initializing with
0.4.0 / 2019-05-13
models.Moleculeconnectivity to default to None rather than an empty list. WARNING this change alters the hashes produced from the
models.ResultPropertiessupports CCSD and CCSD(T) properties.
(GH#56) Algorithms Kabsch
util.linear_sum_assignment, and Uno
util.unoadded. Utilities to generate random 3D rotations
util.random_rotation_matrixand reindex a NumPy array into smaller blocks
(GH#56) Molecular alignment taking into account displacement, rotation, atom exchange, and mirror symmetry for superimposable and differing geometries was added in
molutil.B787(basis NumPy function) and
models.Molecule.align(far more convenient). Suitable for QM-sized molecules. Requires addition package
whichthat provide a path or boolean result for locating modules or commands, respectively. These were migrated from QCEngine along with
parse_versionto colocate the import utilities.
(GH#61) Add molecular visualization to the
models.Moleculeobject through the optional 3dMol.js framework.
relative_geoms='align'that lets Molecules pass if geometries within a translation and rotation of each other.
forgivethat is a list of paths that may differ without failing the comparison.
dtype='terachem'for writing the separate XYZ file required by TeraChem. Angstroms or Bohr allowed, though the latter requires extra in input file.
util.whichadded the Python interpreter path to the default search
*to parameter list of many functions requiring subsequent to be keyword only. Code relying heavily on positional arguments may get broken.
envto use an alternate search
ModuleNotFoundError(for both functions) when not located. It error will have a generic error message which can be extended by
raise_msg. It is strongly encouraged to add specific remedies (like how to install) through this parameter. This is the third exit pattern possible from the “which” functions, of which path/None is the default, True/error happens when
raise_error=True, and True/False happens otherwise when
(GH#65) Testing functions
return_handlerthat lets other printing, logging, and pass/fail behavior to be interjected.
0.3.3 / 2019-03-12
(GH#49) Precompute some mass number and mass lookups and store on
qcel.periodic_table. Also move static
re.compileexpressions out of fns on to module. Mol validation .127s –> .005s.
0.3.2 / 2019-03-11
(GH#48) If Molecule object has passed through molparse validation because it was created with a molparse constructor (e.g.,
from_string), save some time by not passing it through again at
(GH#48) Fixed a
Molecule.get_fragmentbug where ghosted fragments still asserted charge/multiplicity to the validator, which was rightly confused.
0.3.1 / 2019-03-07
(GH#37) Documentation now pulls from the custom QC Archive Sphinx Theme, but can fall back to the standard RTD theme. This allows all docs across QCA to appear consistent with each other.
(GH#41) Conda-build recipe removed to avoid possible confusion for everyone who isn’t a Conda-Forge recipe maintainer. Tests now rely on the
(GH#44) Molecule objects are now always validated against a more rigorous model and fragment multiplicities are fixed at the correct times, even when no multiplicities are provided. Molecule defaults to
0.3.0 / 2019-02-27
dtype=2in keeping with GH:MolSSI/QCSchema#60 with internal
molparse.from_schemarecognizes external fields (existing functionality), internal fields (dtype=2), and mixed.
(GH#33) Pydantic molecule model now contains schema_name and schema_version=2 information.
(GH#35) Models now have an
extrafield for extra attributes, no additional base keys are allowed.
qcel.Datumto Pydantic model. Changes: (a) comment, doi, glossary fields must be accessed by keyword, (b)
dict()and instead of only label, units, data fields in dict, now comment, doi, glossary present _if_ non-default, (c) complex values no longer list-ified by
(GH#36) Changelog and Models documentation.
0.2.6 / 2019-02-18
(GH#32) Updates compliance with Pydantic v0.20.
0.2.5 / 2019-02-13
(GH#31) Lints the code base preparing for a release and minor test improvements.
dihedralmeasurement code for incorrect phase in certain quadrants.
0.2.4 / 2019-02-08
(GH#28) Reduces loading time from ~1 second to 200 ms by delaying
pintimport and ensuring git tags are only computed once.
0.2.3 / 2019-01-29
(GH#24) Update models to be compatible with QCFractal and MongoDB objects in the QCArchive Ecosystem. Also enhances the
jsonfunction to accept
as_dictkeyword, permitting a return as a dictionary of Pydantic-serialized python (primitive) objects, instead of a string.
0.2.2 / 2019-01-28
0.2.1 / 2019-01-27
(GH#20) Moves several Molecule parsing functions to the molparse module.
0.2.0 / 2019-01-25
now requires Python 3.6+
now requires Pydantic
util.unnpthat recursively list-ifies ndarray in a dict now handles lists and flattens.
0.1.3 / 2018-12-14
(GH#12) Adds “connectivity” validation and storage consistent with QCSchema.
0.1.2 / 2018-11-3
0.1.1 / 2018-10-30
(GH#7) Adds outer schema_name/schema_version to
to_schema(..., dtype=1)output so is inverse to
(GH#8) Tests pass for installed module now that comparison tests are xfail.
0.1.0a / 2018-10-24
This is the first alpha release of QCElemental containing the primary three components.
(GH#6) Updated molparse to write new Molecule QCSchema fields in keeping with GH:MolSSI/QCSchema#44
Periodic Table data from NIST SRD144 (c. pre-2015?) collected into
qcelemental.periodictableinstance, with accessors
to_A(and redundant accessors
Decimalformats. Also includes functionality to write a corresponding “C” header.
Physical Constants data from NIST SRD121 (CODATA 2014) collected into
qcelemental.constantsinstance, with access through
Decimalresult). Also includes functionality to write a corresponding “C” header.
from_schemaconstructors parse and rearrange (if necessary) and validate molecule topology inputs from the QC and EFP domains into a QCSchema-like data structure. Current deficiencies from QCSchema are non-contiguous fragments and “provenance” fields. Accessors
to_schemaare highly customizable.
A pint context has been built around the NIST physical constants data so that
qcelemental.constants.conversion_factor(from_unit, to_unit)uses the QCElemental values in its conversions. Resulting
floatis within uncertainty range of NIST constants but won’t be exact for conversions involving multiple fundamental dimensions or
wavelength -> energy != 1 / (energy -> wavelength).