paceENSDF stores the measured beta endpoint energy of each transition
(ENSDF column 7). The parser instead derived branch endpoints as
base-endpoint minus level energy with the base taken from an arbitrary
first-listed branch, so Co60 produced negative branch endpoints and a
hard ValueError (bs_pnpi Co60 exited 1 regardless of cutoff).
- BranchInfo carries the measured endpoint_MeV
- overall endpoint = max measured endpoint over all branches
- decay_info_to_config prefers measured endpoints; the subtraction is
retained only as a legacy fallback for hand-built DecayInfo objects
- Co60/Tc99/Cs137 endpoint tests + CLI dry-run/export tests (D3)
Remove the duplicated inline paceENSDF config builder in the CLI.
It renormalized branch intensities before filtering, diverging from
the library decay_info_to_config path that library callers use
(audit finding D5). CLI paceENSDF configs now come from
create_config_from_source exactly once per run.
Regression test spies on create_config_from_source during a CLI run.
- main(argv) signature enables in-process CLI invocation and testing
- parser defaults for --e-step/--neutrino-mass/--intensity-cutoff/
--csv-separator/--normalize-to-one now None = 'not provided'
- JSON flow applies file values by default and CLI flags as overrides
(previously these five options were silently discarded)
- paceENSDF flow resolves None flags to literal defaults
- new tests/integration/test_cli_flows.py: JSON propagation,
degenerate-mode non-zero exit with no file (JSON + paceENSDF)
- adapted two neutrino-mass default tests to None-parsing semantics
CSVEmitter.export_to_csv raised ValueError only in intensity mode;
degenerate exports with --normalize-to-one silently wrote all-zero
CSV files. The guard now runs before the mode split and no file is
written in either mode.
When config.branches carries exactly one branch (e.g. after
--intensity-cutoff filtering), the branch endpoint/transition define
the spectrum, energy grid, analyzer unified branch view and CSV
header. The stale top-level endpoint inherited from the parent decay
no longer leaks into physics (proven wrong on Cs137: spectrum
extended to 1176 keV while the lone survivor ends at 514 keV).
- Add neutrino mass parameter to SpectrumConfig and CLI interface
- Update PhaseSpace component to handle neutrino mass effects properly
- Update CSV headers to show basic endpoint and all transition types
- Add --csv-verbosity option to CLI with 'full' and 'concise' options
- Fix comparison plots to properly show linear scale variations
- Add comprehensive tests for neutrino mass functionality
- Fixed bug: decay_info_to_config() was dropping branches, causing
create_config_from_source('paceENSDF') to return single-branch config
- Added BranchConfig import to nuclear_data.py
- Marked B1 (paceENSDF integration) as complete in TODO.md
- Updated README.md with paceENSDF details in Implementation Status
- Updated TODO.md Current Status with v0.3.1 notes
paceENSDF returns intensity at state[9] (percent), not state[8] (ENSDF
normalization field, ~1.0 for all branches). This was causing ~50/50
split for Tc99 instead of the correct 99.9984%/0.0016%.
Branch intensity, endpoint, and transition are constant per branch and
already documented in the CSV header. No need to repeat them as columns
for every energy row.
- Cross out B3 in TODO.md with completion details
- Export BranchConfig in __init__.py public API
- Update README to mention per-branch transition types and intensity cutoff
- Add --intensity-cutoff CLI option to filter negligible branches
- Fetch DecayInfo from paceENSDF and construct SpectrumConfig with
branches including per-branch transition types
- Normalize branch intensities to sum to 1.0, filter by cutoff
- Display branch info in dry-run output and logging
- Add intensity_cutoff to SpectrumConfig
- Add transition_type field to BranchInfo and extract per-branch
forbiddenness codes from paceENSDF (not defaulting all to 'A')
- Fix _multi_branch_call to use np.maximum instead of np.where mask
so branches smoothly approach zero at endpoint
- Rewrite _plot_multi_branch_spectra and _plot_multi_branch_debug Panel 1
to plot raw (un-normalized) branch contributions so total = sum of branches
- Remove independent unit-area normalization that made small branches
appear larger than total
- Add commit hash (6 chars) and UTC timestamp to all plots via text boxes
- Add nuclear data header (nuclide, endpoint, transition, corrections) to plots
- Two display modes:
- DEBUG (-vv): 4-panel debug view with all correction factors
- Normal (default): single spectrum plot with nuclear data header
- CLI: --plot now passes show_components=(args.verbose >= 2)
- Remove --mode, --transition-type, and --decay-index CLI flags
(parameters now come from ENSDF data or custom JSON input)
- Clarify energy units in all CLI help text (MeV throughout, keV for detector params)
- Enhance logging verbosity:
- -v (INFO): shows source, component initialization, workflow steps
- -vv (DEBUG): shows detailed internals of all spectrum components
(parameters, evaluation counts, input/output ranges)
- Improve CSV headers with element notation (e.g. Tc99 -> Ru99)
- Add CONVENTIONS.md documenting energy unit and logging conventions
- Update TODO.md with completion status
- Remove empty beta_spectrum/components/recoil.py (0-line stub)
- Remove recoil.py from README.md project tree and components table
- Update TODO.md: replace specific recoil section with generic 'Additional
Corrections' deferred item, update Current Status to 0.3.0 with actual
implementation list, mark A4 as complete, set A3 as next step
- Add proper type annotations to all source files
- Fix return types for numpy operations (np.asarray with dtype=np.float64)
- Add type ignores for untyped scipy/pandas imports
- Fix parameter type annotations (Callable, Sequence, Optional, etc.)
- Add TYPE_CHECKING imports for forward references
- Fix fitter.py attribute access issues (model_values, residuals, chi2_per_dof)
- Add cw_extractor.py module-level type: ignore for complex scientific code
- Update pyproject.toml to exclude notebooks from mypy
- Add mypy disable_error_code configuration for test files
- Run black and ruff for formatting and linting
- Add proper type annotations to all source files
- Fix return types for numpy operations (np.asarray with dtype=np.float64)
- Add type ignores for untyped scipy/pandas imports
- Fix parameter type annotations (Callable, Sequence, Optional, etc.)
- Add TYPE_CHECKING imports for forward references
- Fix fitter.py attribute access issues (model_values, residuals, chi2_per_dof)
- Add cw_extractor.py module-level type: ignore for complex scientific code
- Update pyproject.toml to exclude notebooks from mypy
- Run black and ruff for formatting and linting
- Add nuclear_data.py module for paceENSDF integration and JSON input
- Add DecayInfo and BranchInfo dataclasses for parsed nuclear data
- Implement get_decay_info_from_paceENSDF() to retrieve decay parameters
- Implement load_json_input() and json_to_config() for custom JSON input
- Add create_config_from_source() unified interface for both sources
- Add bs_pnpi CLI with --nuclide, --input, --output, --plot, and detector options
- Add 23 comprehensive tests for nuclear_data module
- Add sample JSON input file
- Update __init__.py and pyproject.toml with new exports and entry point
- Add paceENSDF as optional dependency
Add CurveFitter with scipy least_squares, configurable loss functions
(soft_l1, huber, cauchy), profile likelihood, confidence intervals, and
goodness-of-fit metrics. Add CWExtractor for extracting shape factors
from beta spectra via Kurie analysis, parametrization fitting, and
g_V/g_A extraction. Includes 32 new tests (18 fitter + 14 CW extractor).
Bump version to 0.3.0.
The resummation was previously applied to all energy points when the
flag was set, causing resummed and standard modes to differ everywhere.
Per spec, it should only replace C(beta)*ln(W0-W) with the resummed
power-law form when (W0-W) < delta_cut (~1e-3).
Adds delta_cut parameter (default 1e-3) and corrects the resummation
switch test to verify identical results far from endpoint.
- Add Z (nuclear charge) and A (mass number) parameters to RadiativeCorrection
- Implement O(Z*alpha^2) finite nuclear size correction per Sirlin 1987:
- Energy-dependent part: ln(m_p) - (5/12)*ln(2W) + 43/18
- Nuclear-structure-dependent part: Delta_F(Z,A) from modified Gaussian model
- Add small-beta Taylor expansion for numerical stability at threshold (beta < 1e-6)
- Extract _safe_beta() and _safe_tanh_inv_beta_over_beta() helper methods
- Remove unused log_wright_bessel import
- Add 14 new tests: Z-dependence, numerical stability, A parameter coverage
- Update docs/06-radiative-corrections.md with new API usage examples
All 70 tests pass. 14 new tests in 3 new test classes.
unzhakov