Skip to content

Arkane matching method name#814

Open
calvinp0 wants to merge 6 commits intomainfrom
arkane_match
Open

Arkane matching method name#814
calvinp0 wants to merge 6 commits intomainfrom
arkane_match

Conversation

@calvinp0
Copy link
Member

@calvinp0 calvinp0 commented Jan 6, 2026

RMG-database for AEC has the suffix of years for methods and this can cause issues for ARC when it attempts to find a method/basis if the user does not provide the year. This PR is meant to allow for the following:

  • omit the year and if a non year version is not availble, ARC should report it and it will attempt to match latest year. (CBS-QB3 in RMG-database has a 2023 version and a non year version)
  • provide the year in the input.yml for the sp_level and it will then attempt to match that specific year
  • If a year is not available, it will be report
  • If the user writes '23' instead of '2023' it will error.

This pull request introduces support for specifying an optional 4-digit year suffix for levels of theory, allowing users to differentiate between method variants (e.g., b97d3 vs b97d32023). The changes include enhancements to the Level class, updates to the ARC main workflow to handle the new year parameter, improved input validation and coercion, and expanded test coverage to ensure correct behavior and error handling.

Level of Theory Year Support

  • Added an optional year parameter to the Level class, including validation to ensure it is a 4-digit integer, and incorporated it into string representations (__str__ and simple methods). (arc/level.py) [1] [2] [3] [4] [5]
  • Updated the build method of Level to include the year key in its dictionary representation. (arc/level.py)

ARC Main Workflow Enhancements

  • Introduced the level_of_theory_year parameter to the ARC class, with input validation and coercion logic to handle integer or string inputs, and updated documentation/comments accordingly. (arc/main.py) [1] [2] [3] [4] [5]
  • Modified process_level_of_theory to correctly apply and validate the year suffix, including error handling for conflicting year specifications and ensuring level_of_theory_year is only used with level_of_theory. (arc/main.py) [1] [2]

Testing and Documentation

  • Added and expanded unit tests for year validation, input coercion, correct application of year to levels of theory, and error cases for conflicting or missing year specifications. (arc/level_test.py, arc/main_test.py, arc/statmech/arkane_test.py) [1] [2] [3] [4]
  • Updated documentation to describe the new year parameter and its usage in input files. (docs/source/advanced.rst)

Arkane Integration and Helper Functions

  • Enhanced Arkane adapter and helper functions to handle year suffixes, including normalization, parsing, and selection of the appropriate Arkane database entry based on year, with preference for no-year entries and fallback to the latest available year. (arc/statmech/arkane_test.py) [1] [2] [3]

These changes collectively improve ARC's flexibility and robustness in handling method variants distinguished by year, and ensure consistent integration with Arkane databases and user inputs.

Copy link

Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

This PR adds support for optional 4-digit year suffixes to distinguish method variants in the Arkane database (e.g., b97d3 vs b97d32023). The implementation allows users to omit the year (ARC matches the latest available), specify an explicit year, or receive warnings when a requested year is unavailable.

Key changes:

  • Added year parameter to the Level class with validation requiring 4-digit integers
  • Implemented flexible method/basis matching logic with year-aware searching in Arkane database
  • Enhanced error reporting to list available years when a requested year is not found

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 7 comments.

Show a summary per file
File Description
docs/source/advanced.rst Documents the new year parameter with usage examples
arc/level.py Adds year attribute with 4-digit validation and updates string representations
arc/level_test.py Adds test for year validation (rejects 2-digit, accepts 4-digit)
arc/statmech/arkane.py Core implementation: adds normalization functions, year-aware matching logic, and improved error messages; also adds spinMultiplicity to non-SMILES species template and thread limit environment variables
arc/statmech/arkane_test.py Adds test for year-not-found warning behavior and year in _level_to_str

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

@calvinp0 calvinp0 force-pushed the arkane_match branch 2 times, most recently from 7955fd4 to 67e2cfe Compare January 7, 2026 13:55
@calvinp0 calvinp0 force-pushed the arkane_match branch 2 times, most recently from 1926671 to 773ab92 Compare January 22, 2026 11:27
@calvinp0 calvinp0 requested review from Lilachn91 and Copilot January 22, 2026 17:47
Copy link

Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Copilot reviewed 10 out of 11 changed files in this pull request and generated 11 comments.


💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Comment on lines 157 to 158
and are not valid QC methods. Do not include year suffixes in ``level_of_theory``; instead, set
``arkane_level_of_theory`` with a ``year`` value if you need a specific correction year.
Copy link

Copilot AI Jan 22, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The documentation incorrectly advises users to set year in arkane_level_of_theory, but based on the PR description and implementation, the year parameter should be set directly on sp_level (or other level arguments like freq_level, opt_level). The arkane_level_of_theory is typically derived from sp_level unless explicitly overridden. Users should be advised to set the year parameter on sp_level or other level_of_theory parameters, not on arkane_level_of_theory specifically.

Suggested change
and are not valid QC methods. Do not include year suffixes in ``level_of_theory``; instead, set
``arkane_level_of_theory`` with a ``year`` value if you need a specific correction year.
and are not valid QC methods. Do not include year suffixes in ``level_of_theory``; instead, specify a
``year`` key on ``sp_level`` (and/or other job-specific level dictionaries such as ``opt_level``,
``freq_level``, or ``scan_level``) if you need a specific correction year.

Copilot uses AI. Check for mistakes.
Comment on lines 89 to 91
set ``arkane_level_of_theory`` with a ``year`` value instead.
If ``year`` is omitted, ARC will prefer the no-year Arkane entry for that method/basis; if none exists,
ARC will fall back to the latest available year in the Arkane database.
Copy link

Copilot AI Jan 22, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The documentation incorrectly advises users to set year in arkane_level_of_theory. Based on the PR implementation and examples shown earlier in the documentation (lines 98-112), the year parameter should be set directly on sp_level or other level parameters (opt_level, freq_level, etc.), not specifically on arkane_level_of_theory. The arkane_level_of_theory is typically auto-derived from sp_level unless explicitly overridden.

Suggested change
set ``arkane_level_of_theory`` with a ``year`` value instead.
If ``year`` is omitted, ARC will prefer the no-year Arkane entry for that method/basis; if none exists,
ARC will fall back to the latest available year in the Arkane database.
set a ``year`` field on the relevant level parameter (e.g., ``sp_level``, ``opt_level``, ``freq_level``,
etc.) rather than on ``level_of_theory`` or ``arkane_level_of_theory``. If ``year`` is omitted, ARC will
prefer the no-year Arkane entry for that method/basis; if none exists, ARC will fall back to the latest
available year in the Arkane database.

Copilot uses AI. Check for mistakes.
f"available years: {_format_years(years)}. "
f"Specify a year to select a matching entry."
)
return _level_to_str(sp_level)
Copy link

Copilot AI Jan 22, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Inconsistent behavior when Arkane AEC entry is not found: For composite methods (line 978), the function returns _level_to_str(sp_level) even when no matching entry exists in the database. However, for non-composite methods with freq_scale_factor (line 1005), it returns None. This inconsistency could lead to Arkane attempting to use a level of theory that lacks proper corrections. Both cases should handle missing entries consistently, likely by returning None after logging the warning.

Suggested change
return _level_to_str(sp_level)
# No matching AEC level in Arkane DB for this composite method
return None

Copilot uses AI. Check for mistakes.
# Iodine SOC calculated as a weighted average of the electronic spin splittings of the lowest energy state.
# The splittings are obtained from Huber, K.P.; Herzberg, G., Molecular Spectra and Molecular Structure. IV.
# Constants of Diatomic Molecules, Van Nostrand Reinhold Co., 1979
# Spin orbit correction for F, Si, Cl, Br and B taken form https://cccbdb.nist.gov/elecspin.asp
Copy link

Copilot AI Jan 22, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Spelling error: 'form' should be 'from'

Suggested change
# Spin orbit correction for F, Si, Cl, Br and B taken form https://cccbdb.nist.gov/elecspin.asp
# Spin orbit correction for F, Si, Cl, Br and B taken from https://cccbdb.nist.gov/elecspin.asp

Copilot uses AI. Check for mistakes.
Comment on lines 779 to 782
def _available_years_for_level(level: "Level",
file_path: str,
section_start: str,
section_end: str) -> list[Optional[int]]:
Copy link

Copilot AI Jan 22, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Type hint inconsistency: This function uses lowercase list[Optional[int]] for the return type, but the file imports and uses List from typing module elsewhere (see line 9 and line 683). For consistency with the rest of the codebase, use List[Optional[int]] instead of list[Optional[int]].

Copilot uses AI. Check for mistakes.
return sorted(years, key=lambda y: (-1 if y is None else y))


def _format_years(years: list[Optional[int]]) -> str:
Copy link

Copilot AI Jan 22, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Type hint inconsistency: This function parameter uses lowercase list[Optional[int]], but the file imports and uses List from typing module elsewhere (see line 9 and line 683). For consistency with the rest of the codebase, use List[Optional[int]] instead of list[Optional[int]].

Suggested change
def _format_years(years: list[Optional[int]]) -> str:
def _format_years(years: List[Optional[int]]) -> str:

Copilot uses AI. Check for mistakes.

target_method_norm = _normalize_method(level.method)
target_base, method_year = _split_method_year(target_method_norm)
target_year = getattr(level, 'year', None) if getattr(level, 'year', None) is not None else method_year
Copy link

Copilot AI Jan 22, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The code allows year suffixes in method names (e.g., 'b97d32023') and extracts them for matching, but the documentation explicitly advises against including year suffixes in method names. Consider adding validation to warn or raise an error if a user provides both a year suffix in the method name AND an explicit year parameter with conflicting values, to prevent user confusion and potential bugs.

Suggested change
target_year = getattr(level, 'year', None) if getattr(level, 'year', None) is not None else method_year
explicit_year = getattr(level, 'year', None)
if explicit_year is not None and method_year is not None and explicit_year != method_year:
raise InputError(
f"Conflicting year specifications for level '{level}': "
f"explicit year={explicit_year}, method suffix year={method_year}. "
"Please remove the year suffix from the method name or update the 'year' attribute to match."
)
target_year = explicit_year if explicit_year is not None else method_year

Copilot uses AI. Check for mistakes.
Comment on lines 936 to 937
If multiple entries only differ by year, the one with the *latest* year
is chosen (year=0 if no year in that entry).
Copy link

Copilot AI Jan 22, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The docstring description at lines 936-937 is misleading. It states that "the one with the latest year is chosen" but this only applies when the user does NOT specify a year AND no no-year entry exists. When a user specifies a year explicitly via the year parameter, only that exact year will be matched. The docstring should clarify that the latest-year fallback only occurs when no year is specified and no no-year entry exists.

Suggested change
If multiple entries only differ by year, the one with the *latest* year
is chosen (year=0 if no year in that entry).
When a year is explicitly specified in the Level, only entries with that exact
year are matched. If no year is specified and an entry without a year exists,
that entry is used. Only when no year is specified and no no-year entry exists,
if multiple entries differ only by year, the one with the *latest* year is
chosen (treating entries with no year as year=0).

Copilot uses AI. Check for mistakes.
Comment on lines 947 to 949
qm_corr_files = _get_qm_corrections_files()
if not qm_corr_files:
return None
Copy link

Copilot AI Jan 22, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

When no quantum corrections files are found (qm_corr_files is empty), the function silently returns None without any warning or error message. Consider adding a logger.warning() to inform users that no quantum corrections database files were found, which would help with debugging configuration issues.

Copilot uses AI. Check for mistakes.
@calvinp0 calvinp0 force-pushed the arkane_match branch 2 times, most recently from aeabe6c to 8dee498 Compare January 25, 2026 11:41
@alongd
Copy link
Member

alongd commented Feb 2, 2026

@calvinp0, can you please see which of the copilot comments are relevant?

@calvinp0 calvinp0 force-pushed the arkane_match branch 2 times, most recently from 611490b to 43ca9df Compare February 12, 2026 09:25
Added the input parameter year so that the year of the level of theory can be defined (if required).

Added a test where it ensures that the year provided is 4 digits and not 2.
When a user provides in the input file (for example)

```yaml
level_of_theory: wb97xd/def2tzvp
```

They can also provide

```yaml
level_of_theory_year: 2023
```

If they want to be specific about the year LoT (Or provide None - therefore ensuring no match to any year)

Also we check that there is a conflict between LoTY and sp_level year, although that would not be permissable anyway.

Added tests to ensure sp_level respects the year and opt_level does not need.
Added test o coerce the year from string
added test for explicit year in sp_level
added tests for year in arkane_level_theory year
Intention of the addtional logic and function is to match the users year in LoTY along with the LoT to the correct quantum correction from RMG-database. However, if the user omits the LoTY - then it will be assumed that the latest year available in the quantum corrections is the one to be used.

We also provide feedback now if the year req. was not present in the quantum corrections
Documentation now provides information about the year matching
Introduces a function to generate unique project names for parallel test execution using xdist. This prevents collisions when cleaning up project directories, ensuring reliable test execution.
allows for `pip install .` or `pip install -e .`
Copy link
Member

@alongd alongd left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks! I added a couple of comments, both on the same topic

e.g., 'ZINDO/2', 'DLPNO-MP2-F12/D'
For these cases, use the dictionary-type job-specific level of theory arguments
instead (e.g., ``opt_level``).
level_of_theory_year (int or str, optional): An optional 4-digit year suffix to apply to the single-point
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this should not be an arg of ARC. Instead, ARC has individual Level objects for sp_level, opt_liver, etc., and each one has a year attr as you implemented in the prev commit. I recommend removing this from main

for Arkane database matching only and are not valid QC methods. If you need a specific correction year,
specify either:

- ``level_of_theory_year: 2023`` (applies to the SP side of ``level_of_theory``), or
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

can we instead only give the year in the Level argument, and then use it just for Arkane and not for the ESS calcs?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants