midas-defect 0.2.0__tar.gz → 0.3.0__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- {midas_defect-0.2.0 → midas_defect-0.3.0}/PKG-INFO +2 -2
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/__init__.py +1 -1
- midas_defect-0.3.0/midas_defect/peak_tracker.py +260 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/raster.py +243 -5
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/seed_index.py +51 -7
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect.egg-info/PKG-INFO +2 -2
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect.egg-info/SOURCES.txt +2 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect.egg-info/requires.txt +1 -1
- {midas_defect-0.2.0 → midas_defect-0.3.0}/pyproject.toml +3 -3
- midas_defect-0.3.0/tests/test_peak_tracker.py +246 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_raster.py +125 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_seed_index.py +32 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/LICENSE +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/README.md +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/asterism/__init__.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/asterism/direction.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/asterism/eigenvalue_spectrum.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/asterism/family_asterism.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/asterism/local_decomposition.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/asterism/second_moment.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/asterism_fit.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/attribution.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/bootstrap/__init__.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/bootstrap/aggregators.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/bootstrap/decorators.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/bootstrap/parallel.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/bootstrap/samplers.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/bragg_diffuse.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/burgers_population.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/cli/__init__.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/cli/asterism.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/cli/polytype.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/cli/rods.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/completeness.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/contrast_factor.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/contrast_factor_hex.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/cpfem/__init__.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/cpfem/damask_io.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/cpfem/elastic_sc.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/cpfem/fepx_io.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/cpfem/prisms_io.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/data_io.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/debye_waller/__init__.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/debye_waller/per_grain_wilson.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/defect_tests.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/delta_pdf.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/detector_qc.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/distributions/__init__.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/distributions/divergence.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/distributions/friedel.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/distributions/mackenzie.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/domains.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/energy/__init__.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/energy/balance_closure.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/energy/per_grain.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/energy/volume_weighted.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/examples/__init__.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/examples/_figures.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/examples/copland_fullres_validate.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/examples/demk_fcc_end_to_end.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/examples/raster_batch_cli.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/examples/tutorial_size_strain_mosaic.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/forward_sim.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/geometry.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/gnd/__init__.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/gnd/nye_tensor.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/gnd/scalar_gnd.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/gnd/ssd_decomposition.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/honesty.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/huang.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/indexing.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/ingest.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/intensity_budget.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/lattice.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/line_profile/__init__.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/line_profile/modified_wh.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/line_profile/per_grain_reflections.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/line_profile/warren_alpha.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/line_profile/warren_beta.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/peakshape.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/phases/__init__.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/phases/_systems.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/phases/bcc.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/phases/fcc.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/phases/hcp.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/polytype/__init__.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/polytype/_width.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/polytype/activated_axis.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/polytype/aggregate_thickness.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/polytype/cell_index.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/polytype/doublet_survey.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/polytype/fault_balance.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/polytype/finite_stack.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/polytype/ladder.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/polytype/lamella_thickness.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/polytype/modulation_tilt.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/polytype/modulation_type.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/polytype/satellite_doublet.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/polytype/satellite_excess.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/polytype/satellite_intensity.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/provenance.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/reports/__init__.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/reports/figures.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/reports/inventory.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/residual_decomposition.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/rod_detect.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/rod_profile.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/rows.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/schmid/__init__.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/schmid/per_grain.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/schmid/per_system.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/schmid/stratification.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/segmentation_bias.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/selfcal.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/spatial/__init__.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/spatial/autocorrelation.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/spatial/hall_petch.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/spatial/stress_gradient.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/strain/__init__.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/strain/eigenvalue_spectrum.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/strain/projection.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/strain/von_mises.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/stress/__init__.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/stress/cubic_anisotropic.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/stress/invariants.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/stress/resolved_shear.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/stress/voigt.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/subgrain.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/synthetic.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/thermodynamics/__init__.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/thermodynamics/mecking_kocks.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/thermodynamics/taylor_inversion.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/types.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/variants/__init__.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/variants/common_reference.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/variants/kmeans_fz.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/variants/matched_pairs.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/viz.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect/williamson_hall.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect.egg-info/dependency_links.txt +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect.egg-info/entry_points.txt +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/midas_defect.egg-info/top_level.txt +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/setup.cfg +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_ab_from_dscatter_is_refuted.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_asterism.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_asterism_fit.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_bootstrap.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_bragg_diffuse.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_burgers_population.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_cell_index.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_cli_polytype.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_completeness.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_contrast_factor.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_contrast_factor_hex.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_contrast_factor_ungar1999_table2.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_convention_scan.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_cpfem.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_debye_waller.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_defect_tests.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_delta_pdf.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_demk_experimental_modules.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_demk_g1592_9r_regression.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_demk_real_data_regression.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_detector_angle_maps.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_detector_qc.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_diffuse_to_bragg.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_distributions.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_elastic_sc.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_energy.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_family_asterism.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_fault_balance.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_find_domains.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_forward_sim.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_friedel_pairing.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_geometry.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_geometry_honest_2026.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_gnd.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_grains_reader_widths.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_honesty_decoy_raw.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_huang.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_index_from_cloud_cell.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_ingest.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_intensity_budget.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_lattice.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_line_profile.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_manual_calling_contract.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_modulation_tilt.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_orientation_convention.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_p1_honesty_2026.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_peakshape.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_phases.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_polytype.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_polytype_productized_2026.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_polytype_width.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_qlab_to_pixel.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_real_data_regression.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_reports.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_residual_decomposition.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_rod_detect.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_rod_path_conventions.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_rod_profile.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_rows.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_satellite_doublet.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_satellite_ladder.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_schmid.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_segmentation_bias.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_selfcal.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_space_group_forwarded.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_spatial.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_strain.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_stress.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_subgrain.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_synthetic_regression.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_targeted_recovery.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_thermodynamics.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_types.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_variants.py +0 -0
- {midas_defect-0.2.0 → midas_defect-0.3.0}/tests/test_williamson_hall.py +0 -0
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
Metadata-Version: 2.4
|
|
2
2
|
Name: midas-defect
|
|
3
|
-
Version: 0.
|
|
3
|
+
Version: 0.3.0
|
|
4
4
|
Summary: Differentiable, phase-agnostic diffuse-scattering defect metrology on top of standard FF-HEDM: intensity-budget closure, per-grain dislocation density, fault-rod and selection-rule tests.
|
|
5
5
|
Author-email: Hemant Sharma <hsharma@anl.gov>
|
|
6
6
|
License-Expression: BSD-3-Clause
|
|
@@ -21,7 +21,7 @@ Requires-Dist: tifffile>=2023.0
|
|
|
21
21
|
Requires-Dist: scikit-image>=0.20
|
|
22
22
|
Requires-Dist: zarr>=2.14
|
|
23
23
|
Requires-Dist: midas-transforms>=0.1
|
|
24
|
-
Requires-Dist: midas-hkls>=0.
|
|
24
|
+
Requires-Dist: midas-hkls>=0.14.0
|
|
25
25
|
Requires-Dist: midas-stress>=0.1
|
|
26
26
|
Requires-Dist: midas-diffract>=0.1
|
|
27
27
|
Requires-Dist: midas-index>=0.1
|
|
@@ -0,0 +1,260 @@
|
|
|
1
|
+
"""Model-free peak tracking across a raster: no hkl, no indexing, no cell.
|
|
2
|
+
|
|
3
|
+
Why detector-space tracking, not hkl matching: each raster position's crystal domain has its
|
|
4
|
+
OWN orientation (U), so "hkl=(2,0,2)" at position A and the same label at position B are not
|
|
5
|
+
the same physical reflection unless the two domains happen to share orientation. But NEIGHBORING
|
|
6
|
+
raster positions very likely share the same physical grain with only a slowly drifting
|
|
7
|
+
orientation (mosaic), so a peak's DETECTOR position (row, col, frame) moves smoothly from one
|
|
8
|
+
raster point to its neighbor. Tracking that trajectory directly -- flood-fill from a seed pixel,
|
|
9
|
+
re-centering a small search window at each step from the previous position's own centroid -- is
|
|
10
|
+
a raw-data-first measurement that sidesteps the whole indexing/index-asymmetry discussion (an
|
|
11
|
+
a/b-splitting artifact already documented for one project) entirely: no cell, no lattice, no
|
|
12
|
+
domain. Where the peak's signal drops out (grain boundary, moved off a family's Bragg condition,
|
|
13
|
+
etc.) the walk simply stops there and is not propagated further -- that is real information (a
|
|
14
|
+
discontinuity), not a bug to paper over.
|
|
15
|
+
"""
|
|
16
|
+
from __future__ import annotations
|
|
17
|
+
|
|
18
|
+
from typing import Callable, Dict, Optional
|
|
19
|
+
|
|
20
|
+
import numpy as np
|
|
21
|
+
|
|
22
|
+
__all__ = ["windowed_centroid", "track_peak_over_raster", "plot_peak_maps",
|
|
23
|
+
"snap_to_local_max", "STATUS_NO_SIGNAL", "STATUS_REJECTED", "STATUS_TRACKED"]
|
|
24
|
+
|
|
25
|
+
STATUS_NO_SIGNAL = 0 # windowed_centroid found nothing above the SNR floor
|
|
26
|
+
STATUS_REJECTED = 1 # a candidate was found, but the continuity guard rejected it
|
|
27
|
+
STATUS_TRACKED = 2 # accepted and propagated to neighbors
|
|
28
|
+
|
|
29
|
+
_FIELDS = ("centroid_row", "centroid_col", "centroid_frame", "intensity",
|
|
30
|
+
"sigma_row", "sigma_col", "sigma_frame", "peak_max", "background")
|
|
31
|
+
|
|
32
|
+
|
|
33
|
+
def windowed_centroid(frames: np.ndarray, row0: int, col0: int, frame0: int, *,
|
|
34
|
+
win_rc: int = 15, win_f: int = 4, min_snr: float = 5.0
|
|
35
|
+
) -> Optional[Dict[str, float]]:
|
|
36
|
+
"""Intensity-weighted 3-D centroid and width in a box around (row0, col0, frame0).
|
|
37
|
+
|
|
38
|
+
Background is the box's own median (robust to a few bright pixels); a peak is accepted
|
|
39
|
+
only if its max exceeds `min_snr` times the local background's MAD-based noise estimate --
|
|
40
|
+
otherwise this position has genuinely lost the peak, and the caller must not propagate a
|
|
41
|
+
centroid onward from it.
|
|
42
|
+
|
|
43
|
+
Returns a dict (centroid_row, centroid_col, centroid_frame, intensity, sigma_row, sigma_col,
|
|
44
|
+
sigma_frame, peak_max, background) or None if the peak is not found in this window.
|
|
45
|
+
"""
|
|
46
|
+
n_f, n_r, n_c = frames.shape
|
|
47
|
+
r0, r1 = max(0, row0 - win_rc), min(n_r, row0 + win_rc + 1)
|
|
48
|
+
c0, c1 = max(0, col0 - win_rc), min(n_c, col0 + win_rc + 1)
|
|
49
|
+
f0, f1 = max(0, frame0 - win_f), min(n_f, frame0 + win_f + 1)
|
|
50
|
+
box = frames[f0:f1, r0:r1, c0:c1].astype(np.float64)
|
|
51
|
+
if box.size == 0:
|
|
52
|
+
return None
|
|
53
|
+
|
|
54
|
+
bg = float(np.median(box))
|
|
55
|
+
mad = float(np.median(np.abs(box - bg))) + 1e-9
|
|
56
|
+
noise = 1.4826 * mad
|
|
57
|
+
peak_max = float(box.max())
|
|
58
|
+
if noise <= 0 or (peak_max - bg) < min_snr * noise:
|
|
59
|
+
return None
|
|
60
|
+
|
|
61
|
+
w = np.clip(box - bg, 0, None)
|
|
62
|
+
total = w.sum()
|
|
63
|
+
if total <= 0:
|
|
64
|
+
return None
|
|
65
|
+
|
|
66
|
+
ff, rr, cc = np.meshgrid(np.arange(f0, f1), np.arange(r0, r1), np.arange(c0, c1),
|
|
67
|
+
indexing="ij")
|
|
68
|
+
cen_f = float((w * ff).sum() / total)
|
|
69
|
+
cen_r = float((w * rr).sum() / total)
|
|
70
|
+
cen_c = float((w * cc).sum() / total)
|
|
71
|
+
sig_f = float(np.sqrt((w * (ff - cen_f) ** 2).sum() / total))
|
|
72
|
+
sig_r = float(np.sqrt((w * (rr - cen_r) ** 2).sum() / total))
|
|
73
|
+
sig_c = float(np.sqrt((w * (cc - cen_c) ** 2).sum() / total))
|
|
74
|
+
|
|
75
|
+
return dict(centroid_row=cen_r, centroid_col=cen_c, centroid_frame=cen_f,
|
|
76
|
+
intensity=float(total), sigma_row=sig_r, sigma_col=sig_c, sigma_frame=sig_f,
|
|
77
|
+
peak_max=peak_max, background=bg)
|
|
78
|
+
|
|
79
|
+
|
|
80
|
+
def snap_to_local_max(proj: np.ndarray, row: int, col: int, radius: int = 12):
|
|
81
|
+
"""Snap an approximate (row, col) to the true local maximum of `proj` within `radius`.
|
|
82
|
+
|
|
83
|
+
Used to turn an imprecise click (or any coarse guess) on a max-over-frames projection into
|
|
84
|
+
a usable seed pixel -- clicking exactly on a 1-pixel-wide spot is unrealistic.
|
|
85
|
+
"""
|
|
86
|
+
n_r, n_c = proj.shape
|
|
87
|
+
r0, r1 = max(0, row - radius), min(n_r, row + radius + 1)
|
|
88
|
+
c0, c1 = max(0, col - radius), min(n_c, col + radius + 1)
|
|
89
|
+
box = proj[r0:r1, c0:c1]
|
|
90
|
+
rr, cc = np.unravel_index(np.argmax(box), box.shape)
|
|
91
|
+
return r0 + rr, c0 + cc
|
|
92
|
+
|
|
93
|
+
|
|
94
|
+
def track_peak_over_raster(loader: Callable[[int], np.ndarray], n_rows: int, n_cols: int,
|
|
95
|
+
seed_pos: int, seed_row: float, seed_col: float, seed_frame: float,
|
|
96
|
+
*, win_rc: int = 15, win_f: int = 4, min_snr: float = 5.0,
|
|
97
|
+
max_step_px: float = 8.0, max_step_frames: float = 2.0,
|
|
98
|
+
max_intensity_ratio: float = 4.0, log: Callable = print
|
|
99
|
+
) -> Dict[str, np.ndarray]:
|
|
100
|
+
"""Flood-fill a peak's detector trajectory over a row-major (n_rows x n_cols) raster.
|
|
101
|
+
|
|
102
|
+
`loader(p)` returns position p's (already dead-frame-cleaned) raw frame stack.
|
|
103
|
+
`seed_pos` is the 0-indexed position the user picked the peak from; `seed_row/col/frame`
|
|
104
|
+
is that peak's approximate location there (need not be exact -- the first window re-centers).
|
|
105
|
+
|
|
106
|
+
**Continuity guard, added after checking a real run for jumps between raster-grid
|
|
107
|
+
neighbors:** a first version with only an SNR cut let the tracker HIJACK a different,
|
|
108
|
+
unrelated bright peak whenever one drifted within the `win_rc` search box -- 391/1171
|
|
109
|
+
grid-adjacent pairs jumped >5 px, several 20-32 px with a simultaneous 3-20x intensity
|
|
110
|
+
change, concentrated where the search entered a busier (powder-rich) region. A 4 um raster
|
|
111
|
+
step should not move a real peak far on the detector, so a candidate is accepted only if
|
|
112
|
+
BOTH its centroid shift from the seed (`max_step_px`, `max_step_frames`) AND its intensity
|
|
113
|
+
ratio to the seed's own last intensity (`max_intensity_ratio`) stay bounded; otherwise this
|
|
114
|
+
branch stops here rather than silently propagating a wrong lock forward looking locally
|
|
115
|
+
smooth.
|
|
116
|
+
|
|
117
|
+
A rejected candidate is not simply discarded: it is recorded with ``status`` ==
|
|
118
|
+
``STATUS_REJECTED`` and its field values are still stored, so a caller can show "a peak was
|
|
119
|
+
found here but it does not connect continuously" rather than an indistinguishable blank.
|
|
120
|
+
Neighbors are not enqueued from a rejected step, same as from a no-signal one -- only an
|
|
121
|
+
accepted step propagates the walk, since a rejected candidate's own position is exactly
|
|
122
|
+
what should not be trusted as a re-centering seed.
|
|
123
|
+
|
|
124
|
+
Returns a dict of (n_rows, n_cols) float arrays (NaN where `status` ==
|
|
125
|
+
:data:`STATUS_NO_SIGNAL`): the fields in :data:`_FIELDS`, plus an integer ``status`` array
|
|
126
|
+
(:data:`STATUS_NO_SIGNAL` / :data:`STATUS_REJECTED` / :data:`STATUS_TRACKED`) and a boolean
|
|
127
|
+
``tracked`` array (``status == STATUS_TRACKED``, kept for callers that only care about the
|
|
128
|
+
accepted walk).
|
|
129
|
+
"""
|
|
130
|
+
grids = {f: np.full((n_rows, n_cols), np.nan) for f in _FIELDS}
|
|
131
|
+
status = np.full((n_rows, n_cols), STATUS_NO_SIGNAL, dtype=np.int8)
|
|
132
|
+
visited = np.zeros((n_rows, n_cols), dtype=bool)
|
|
133
|
+
|
|
134
|
+
queue = [(seed_pos, seed_row, seed_col, seed_frame, None)]
|
|
135
|
+
n_done = 0
|
|
136
|
+
n_rejected = 0
|
|
137
|
+
while queue:
|
|
138
|
+
p, sr, sc, sf, seed_intensity = queue.pop(0)
|
|
139
|
+
r, c = p // n_cols, p % n_cols
|
|
140
|
+
if visited[r, c]:
|
|
141
|
+
continue
|
|
142
|
+
visited[r, c] = True
|
|
143
|
+
|
|
144
|
+
frames = loader(p)
|
|
145
|
+
result = windowed_centroid(frames, int(round(sr)), int(round(sc)), int(round(sf)),
|
|
146
|
+
win_rc=win_rc, win_f=win_f, min_snr=min_snr)
|
|
147
|
+
n_done += 1
|
|
148
|
+
if n_done % 25 == 0:
|
|
149
|
+
log(f" tracked {n_done} positions so far ({int((status == STATUS_TRACKED).sum())} "
|
|
150
|
+
f"with signal, {n_rejected} rejected as likely hijacks)...")
|
|
151
|
+
|
|
152
|
+
if result is None:
|
|
153
|
+
continue # lost here -- do not enqueue neighbors from a failed centroid
|
|
154
|
+
|
|
155
|
+
accepted = True
|
|
156
|
+
if seed_intensity is not None:
|
|
157
|
+
step_px = np.hypot(result["centroid_row"] - sr, result["centroid_col"] - sc)
|
|
158
|
+
step_f = abs(result["centroid_frame"] - sf)
|
|
159
|
+
ratio = max(result["intensity"], seed_intensity) / min(result["intensity"], seed_intensity)
|
|
160
|
+
if step_px > max_step_px or step_f > max_step_frames or ratio > max_intensity_ratio:
|
|
161
|
+
accepted = False
|
|
162
|
+
|
|
163
|
+
for f in _FIELDS:
|
|
164
|
+
grids[f][r, c] = result[f]
|
|
165
|
+
|
|
166
|
+
if not accepted:
|
|
167
|
+
n_rejected += 1
|
|
168
|
+
status[r, c] = STATUS_REJECTED
|
|
169
|
+
continue # likely hijacked a different, unrelated peak -- stop this branch here
|
|
170
|
+
|
|
171
|
+
status[r, c] = STATUS_TRACKED
|
|
172
|
+
for dr, dc in ((1, 0), (-1, 0), (0, 1), (0, -1)):
|
|
173
|
+
nr, nc = r + dr, c + dc
|
|
174
|
+
if 0 <= nr < n_rows and 0 <= nc < n_cols and not visited[nr, nc]:
|
|
175
|
+
queue.append((nr * n_cols + nc, result["centroid_row"], result["centroid_col"],
|
|
176
|
+
result["centroid_frame"], result["intensity"]))
|
|
177
|
+
|
|
178
|
+
grids["status"] = status
|
|
179
|
+
grids["tracked"] = status == STATUS_TRACKED
|
|
180
|
+
n_tracked = int(grids["tracked"].sum())
|
|
181
|
+
log(f"done: {n_tracked} / {n_rows * n_cols} tracked, {n_rejected} rejected as likely "
|
|
182
|
+
f"hijacks, {n_rows * n_cols - n_tracked - n_rejected} no signal")
|
|
183
|
+
return grids
|
|
184
|
+
|
|
185
|
+
|
|
186
|
+
def plot_peak_maps(data, *, fields=("centroid_row", "centroid_col", "centroid_frame",
|
|
187
|
+
"intensity", "sigma_row", "sigma_col"),
|
|
188
|
+
cmap: str = "viridis", title_prefix: str = "", show: bool = True):
|
|
189
|
+
"""Render a grid of per-field raster maps from a :func:`track_peak_over_raster` result.
|
|
190
|
+
|
|
191
|
+
`data` is the dict returned by :func:`track_peak_over_raster` (or an ``np.load`` of an
|
|
192
|
+
``.npz`` saved from it -- anything indexable by the field names plus ``status``).
|
|
193
|
+
|
|
194
|
+
Three-state per pixel, using ``status``: STATUS_NO_SIGNAL pixels are left blank (no color,
|
|
195
|
+
via ``cmap.set_bad``); STATUS_TRACKED pixels are colored normally by the field value;
|
|
196
|
+
STATUS_REJECTED pixels are colored by their (still-real) field value but additionally
|
|
197
|
+
outlined with a red square marker, so "a discontinuous candidate was found here" stays
|
|
198
|
+
visible rather than reading identically to "nothing here" or to a trusted walk step.
|
|
199
|
+
|
|
200
|
+
`fields` selects which panels to draw and in what order -- pass any subset/order of
|
|
201
|
+
``centroid_row``, ``centroid_col``, ``centroid_frame``, ``intensity``, ``sigma_row``,
|
|
202
|
+
``sigma_col``, ``sigma_frame``, ``peak_max``, ``background``. `cmap` is any matplotlib
|
|
203
|
+
colormap name, applied to every panel.
|
|
204
|
+
|
|
205
|
+
Returns the ``(fig, axes)`` matplotlib objects (matplotlib is imported lazily so the
|
|
206
|
+
package can be installed without it).
|
|
207
|
+
|
|
208
|
+
Accepts a result from a version of :func:`track_peak_over_raster` predating ``status``
|
|
209
|
+
(only ``tracked``, a plain bool grid) -- degrades to the old two-state rendering (no
|
|
210
|
+
STATUS_REJECTED outline) rather than raising, since existing saved ``.npz`` results should
|
|
211
|
+
not need re-tracking just to be plotted with the new function.
|
|
212
|
+
"""
|
|
213
|
+
import matplotlib.pyplot as plt
|
|
214
|
+
|
|
215
|
+
tracked = np.asarray(data["tracked"])
|
|
216
|
+
if "status" in data:
|
|
217
|
+
status = np.asarray(data["status"])
|
|
218
|
+
else:
|
|
219
|
+
status = np.where(tracked, STATUS_TRACKED, STATUS_NO_SIGNAL)
|
|
220
|
+
n_total = status.size
|
|
221
|
+
n_tracked = int(tracked.sum())
|
|
222
|
+
|
|
223
|
+
labels = {"centroid_row": "centroid row (px)", "centroid_col": "centroid col (px)",
|
|
224
|
+
"centroid_frame": "centroid frame (omega step)",
|
|
225
|
+
"intensity": "integrated intensity", "sigma_row": "width, row (px)",
|
|
226
|
+
"sigma_col": "width, col (px)", "sigma_frame": "width, omega (frames)",
|
|
227
|
+
"peak_max": "peak max (counts)", "background": "local background (counts)"}
|
|
228
|
+
|
|
229
|
+
n = len(fields)
|
|
230
|
+
ncols = 3 if n > 2 else n
|
|
231
|
+
nrows = -(-n // ncols) # ceil
|
|
232
|
+
fig, axes = plt.subplots(nrows, ncols, figsize=(4.3 * ncols, 3.6 * nrows), squeeze=False)
|
|
233
|
+
axes_flat = axes.flat
|
|
234
|
+
|
|
235
|
+
rej_r, rej_c = np.where(status == STATUS_REJECTED)
|
|
236
|
+
|
|
237
|
+
for ax, key in zip(axes_flat, fields):
|
|
238
|
+
cmap_obj = plt.get_cmap(cmap).copy()
|
|
239
|
+
cmap_obj.set_bad("#dddddd")
|
|
240
|
+
arr = np.where(status >= STATUS_REJECTED, np.asarray(data[key]), np.nan)
|
|
241
|
+
im = ax.imshow(arr, origin="lower", cmap=cmap_obj)
|
|
242
|
+
if rej_r.size:
|
|
243
|
+
ax.scatter(rej_c, rej_r, s=16, facecolors="none", edgecolors="red",
|
|
244
|
+
linewidths=0.9, marker="s")
|
|
245
|
+
ax.set_title(labels.get(key, key), fontsize=10)
|
|
246
|
+
ax.set_xlabel("col (raster)")
|
|
247
|
+
ax.set_ylabel("row (raster)")
|
|
248
|
+
fig.colorbar(im, ax=ax, fraction=0.046)
|
|
249
|
+
for ax in list(axes_flat)[n:]:
|
|
250
|
+
ax.axis("off")
|
|
251
|
+
|
|
252
|
+
seed_p = data["seed_p"] if "seed_p" in data else None
|
|
253
|
+
seed_str = f"seed p={int(seed_p)} -- " if seed_p is not None else ""
|
|
254
|
+
fig.suptitle(f"{title_prefix}{seed_str}{n_tracked}/{n_total} tracked, "
|
|
255
|
+
f"{int((status == STATUS_REJECTED).sum())} rejected "
|
|
256
|
+
f"(red outline = rejected, found but discontinuous)")
|
|
257
|
+
fig.tight_layout()
|
|
258
|
+
if show:
|
|
259
|
+
plt.show()
|
|
260
|
+
return fig, axes
|
|
@@ -113,10 +113,12 @@ from .geometry import (Geometry, detector_angle_maps, ewald_crossing_omegas, pix
|
|
|
113
113
|
from .honesty import decoy_test, feature_in_raw, inflated_cell
|
|
114
114
|
from .ingest import build_mask, detect_powder_rings, find_blobs_3d, flag_powder, subtract_background
|
|
115
115
|
from .rows import hkl_box_from_geometry
|
|
116
|
+
from .seed_index import bootstrap_orientation_uncertainty
|
|
116
117
|
|
|
117
118
|
__all__ = ["PositionResult", "reduce_one_position", "reduce_raster_block",
|
|
118
119
|
"assemble_raster_results", "omega_sign_check", "OmegaSignCheck",
|
|
119
|
-
"predict_reflections", "raster_wide_asymmetry_sign_test"
|
|
120
|
+
"predict_reflections", "raster_wide_asymmetry_sign_test",
|
|
121
|
+
"plot_domain_overlay", "plot_orientation_envelope", "plot_lattice_envelope"]
|
|
120
122
|
|
|
121
123
|
|
|
122
124
|
# ---------------------------------------------------------------------------
|
|
@@ -202,6 +204,49 @@ class PositionResult:
|
|
|
202
204
|
split_pct: Optional[Tuple[float, float, float]] # (delta, sigma, z) from split_with_error, or None
|
|
203
205
|
notes: List[str] = field(default_factory=list)
|
|
204
206
|
quotable: bool = False
|
|
207
|
+
#: SKETCH, 2026-09-13. **`/verify` REFUTED the claim that this envelope's width is "a real,
|
|
208
|
+
#: differentiating signal of orientation determinability"** (claim b82502d6b5ad: physics,
|
|
209
|
+
#: statistics, and artifact lenses all REFUTED; independent reproduction SURVIVED -- the
|
|
210
|
+
#: numbers themselves are computed correctly, the interpretation was not). Do not present
|
|
211
|
+
#: a sparse-domain envelope as evidence that domain's orientation is "real but uncertain"
|
|
212
|
+
#: without also checking `decoy` for that domain -- a domain the pipeline's own honesty
|
|
213
|
+
#: gate does not trust can show a wide envelope for reasons having nothing to do with
|
|
214
|
+
#: orientation physics (see `bootstrap_orientation_uncertainty`'s docstring for the full
|
|
215
|
+
#: mechanism: generic 1/sqrt(n) bootstrap scaling, non-independent resamples from a tiny
|
|
216
|
+
#: combinatorial pool below ``keep_n``, or -- the bug actually fixed by this claim -- a
|
|
217
|
+
#: silently-tetragonal orientation fit for a domain that is not tetragonal).
|
|
218
|
+
#:
|
|
219
|
+
#: One entry per domain (aligned with ``domains.domains``), from
|
|
220
|
+
#: ``seed_index.bootstrap_orientation_uncertainty``, or ``None`` where not computed
|
|
221
|
+
#: (default: not computed at all, see ``reduce_one_position``'s ``orientation_uncertainty``
|
|
222
|
+
#: flag -- this is NOT free, each entry re-runs the orientation fit
|
|
223
|
+
#: ``n_orientation_boot`` times). Keys: ``pair_angle_mean_deg`` / ``pair_angle_p95_deg`` /
|
|
224
|
+
#: ``pair_angle_max_deg`` (pairwise misorientation across the bootstrap U's), plus
|
|
225
|
+
#: ``U_mean``/``U_bootstraps``/``n_distinct_subsets``/``independent_resamples`` -- READ
|
|
226
|
+
#: ``independent_resamples`` before trusting a spread from a sparse domain.
|
|
227
|
+
#:
|
|
228
|
+
#: ``max`` several times ``mean`` is NOT by itself evidence of a bimodal/non-unimodal
|
|
229
|
+
#: spread -- checked directly on the cleanest synthetic domain available (321 reflections,
|
|
230
|
+
#: textbook single orientation): max/mean still came out ~3.7x, purely from comparing the
|
|
231
|
+
#: MAX of many bootstrap pairs against their MEAN (an order-statistics effect, not a
|
|
232
|
+
#: property of the underlying distribution's shape). To check for a genuinely multi-modal
|
|
233
|
+
#: spread, inspect ``U_bootstraps`` directly rather than reading a ratio off the summary
|
|
234
|
+
#: numbers.
|
|
235
|
+
orientation_envelope: List[Optional[dict]] = field(default_factory=list)
|
|
236
|
+
#: SKETCH, 2026-09-13, not yet independently /verify'd. (n_kept, 6) array of every
|
|
237
|
+
#: bootstrap resample's full expanded cell -- (a, b, c, alpha, beta, gamma) -- from
|
|
238
|
+
#: `midas_hkls.refine_cell_joint`'s own resampling loop, or None when no domain reached
|
|
239
|
+
#: `refine_cell_joint` (``refined_cell is None``) or too few resamples succeeded.
|
|
240
|
+
#: `split_pct`'s sigma is this array's std on columns 0/1; this is the same data before
|
|
241
|
+
#: being collapsed to that one number -- read individual a/b/c ranges from columns
|
|
242
|
+
#: 0/1/2, not just their difference.
|
|
243
|
+
cell_bootstrap_samples: Optional[np.ndarray] = None
|
|
244
|
+
#: (a, b, c, alpha, beta, gamma) bootstrap sigma, i.e. ``cell_bootstrap_samples.std(axis=0,
|
|
245
|
+
#: ddof=1)`` -- the compact, per-parameter summary of the same data, kept separately
|
|
246
|
+
#: because it (not the full array) is what belongs in a raster-wide map: cheap to carry
|
|
247
|
+
#: through JSON for hundreds of positions, where the full bootstrap array is not (see
|
|
248
|
+
#: ``to_dict()``, which drops ``cell_bootstrap_samples`` but keeps this).
|
|
249
|
+
cell_sigma_bootstrap: Optional[Tuple[float, ...]] = None
|
|
205
250
|
|
|
206
251
|
def summary(self) -> str:
|
|
207
252
|
lines = [f"position {self.point}: {self.n_spots} spots, {self.omega_sign}",
|
|
@@ -238,6 +283,20 @@ class PositionResult:
|
|
|
238
283
|
d["omega_sign"] = dict(chosen_sign=self.omega_sign.chosen_sign,
|
|
239
284
|
n_explained=self.omega_sign.n_explained,
|
|
240
285
|
decisive=self.omega_sign.decisive)
|
|
286
|
+
# orientation_envelope/cell_bootstrap_samples carry raw bootstrap arrays (U_bootstraps:
|
|
287
|
+
# n_domains * n_orientation_boot * 3x3 floats; cell_bootstrap_samples: n_bootstrap * 6
|
|
288
|
+
# floats) meant for one position's own live plotting (05), not for a whole raster's
|
|
289
|
+
# worth of JSON files (06/reduce_raster_block) -- keep only the compact summaries a
|
|
290
|
+
# raster map actually needs; U_mean/cell_bootstrap_samples stay in-memory-only.
|
|
291
|
+
d["orientation_envelope"] = [
|
|
292
|
+
None if oe is None else
|
|
293
|
+
dict(U_mean=oe["U_mean"].tolist(), pair_angle_mean_deg=oe["pair_angle_mean_deg"],
|
|
294
|
+
pair_angle_p95_deg=oe["pair_angle_p95_deg"], pair_angle_max_deg=oe["pair_angle_max_deg"],
|
|
295
|
+
n_boot=oe["n_boot"], keep_n=oe["keep_n"], n_distinct_subsets=oe["n_distinct_subsets"],
|
|
296
|
+
independent_resamples=oe["independent_resamples"])
|
|
297
|
+
for oe in self.orientation_envelope
|
|
298
|
+
]
|
|
299
|
+
d["cell_bootstrap_samples"] = None
|
|
241
300
|
return d
|
|
242
301
|
|
|
243
302
|
def save(self, path) -> None:
|
|
@@ -328,6 +387,124 @@ def predict_reflections(U: np.ndarray, B: np.ndarray, hkl: np.ndarray, geom: Geo
|
|
|
328
387
|
hkl[kept_idx].astype(int), kept_idx)
|
|
329
388
|
|
|
330
389
|
|
|
390
|
+
# ---------------------------------------------------------------------------
|
|
391
|
+
# Reporting plots -- SKETCH, 2026-09-13. matplotlib is the package's optional `viz`
|
|
392
|
+
# extra, imported lazily inside each function so it stays optional at module import time.
|
|
393
|
+
# ---------------------------------------------------------------------------
|
|
394
|
+
|
|
395
|
+
def plot_domain_overlay(frames: np.ndarray, res: "PositionResult", geom: Geometry, *, ax=None):
|
|
396
|
+
"""The max-intensity projection over ``frames``, with every domain's own claimed
|
|
397
|
+
reflections predicted back onto it (same discipline DFXM calls "look at the curves
|
|
398
|
+
before the summary"). One color per domain, labelled with its `decoy_test` verdict --
|
|
399
|
+
a domain whose own honesty gate failed should look suspect here too, not just in a table.
|
|
400
|
+
"""
|
|
401
|
+
import matplotlib.pyplot as plt
|
|
402
|
+
fig = None
|
|
403
|
+
if ax is None:
|
|
404
|
+
fig, ax = plt.subplots(figsize=(7.5, 6.6))
|
|
405
|
+
colors = plt.cm.tab10(np.linspace(0, 1, max(len(res.domains.domains), 1)))
|
|
406
|
+
im = ax.imshow(np.log10(np.clip(frames.max(axis=0), 1, None)), cmap="magma")
|
|
407
|
+
if fig is not None:
|
|
408
|
+
fig.colorbar(im, ax=ax, fraction=0.046, label="log10(max over frames)")
|
|
409
|
+
for i, dom in enumerate(res.domains.domains):
|
|
410
|
+
pr, pc, _pw, _phkl, _kept = predict_reflections(dom.U, dom.lat.B, dom.hkl, geom,
|
|
411
|
+
res.omega_sign.chosen_sign)
|
|
412
|
+
ax.scatter(pc, pr, s=32, facecolors="none", edgecolors=[colors[i]], linewidths=1.3,
|
|
413
|
+
label=f"domain {i} ({len(pr)} claimed, decoy={res.decoy[i]['verdict']})")
|
|
414
|
+
ax.set_title(f"position {res.point}: {len(res.domains.domains)} domain(s) found")
|
|
415
|
+
ax.set_xlabel("column"); ax.set_ylabel("row")
|
|
416
|
+
if res.domains.domains:
|
|
417
|
+
ax.legend(fontsize=8, frameon=False, loc="lower right")
|
|
418
|
+
return fig
|
|
419
|
+
|
|
420
|
+
|
|
421
|
+
def plot_orientation_envelope(res: "PositionResult"):
|
|
422
|
+
"""Per domain, a histogram of each bootstrap ``U``'s misorientation from the bootstrap
|
|
423
|
+
mean -- the visual form of ``PositionResult.orientation_envelope``.
|
|
424
|
+
|
|
425
|
+
**Read this plot, not just the summary numbers, before trusting a sparse domain's
|
|
426
|
+
envelope.** `/verify` REFUTED an early version of this envelope as a general
|
|
427
|
+
"orientation determinability signal" (claim ``b82502d6b5ad``) partly because a domain
|
|
428
|
+
resampled from too few reflections draws from a tiny combinatorial pool
|
|
429
|
+
(``n_distinct_subsets``) and can show a visibly LUMPY, non-smooth, even bimodal
|
|
430
|
+
histogram rather than a genuine continuous spread -- exactly the kind of thing a summary
|
|
431
|
+
mean/p95/max cannot show but a histogram does immediately. Domains with
|
|
432
|
+
``orientation_envelope[i] is None`` (fewer than 4 claimed reflections) are skipped.
|
|
433
|
+
"""
|
|
434
|
+
import matplotlib.pyplot as plt
|
|
435
|
+
n_dom = len(res.domains.domains)
|
|
436
|
+
fig, axes = plt.subplots(1, max(n_dom, 1), figsize=(5.5 * max(n_dom, 1), 4.2), squeeze=False)
|
|
437
|
+
axes = axes[0]
|
|
438
|
+
colors = plt.cm.tab10(np.linspace(0, 1, max(n_dom, 1)))
|
|
439
|
+
for i in range(n_dom):
|
|
440
|
+
ax = axes[i]
|
|
441
|
+
oe = res.orientation_envelope[i] if i < len(res.orientation_envelope) else None
|
|
442
|
+
dom = res.domains.domains[i]
|
|
443
|
+
if oe is None:
|
|
444
|
+
ax.text(0.5, 0.5, "not computed\n(<4 claimed reflections, or\n"
|
|
445
|
+
"orientation_uncertainty=False)", ha="center", va="center",
|
|
446
|
+
transform=ax.transAxes, fontsize=9)
|
|
447
|
+
ax.set_title(f"domain {i}"); ax.set_xticks([]); ax.set_yticks([])
|
|
448
|
+
continue
|
|
449
|
+
U_mean = oe["U_mean"]
|
|
450
|
+
angles = []
|
|
451
|
+
for Ub in oe["U_bootstraps"]:
|
|
452
|
+
delta = Ub @ np.asarray(U_mean).T
|
|
453
|
+
c = np.clip((np.trace(delta) - 1) / 2, -1, 1)
|
|
454
|
+
angles.append(np.degrees(np.arccos(c)))
|
|
455
|
+
ax.hist(angles, bins=min(12, max(len(angles) // 2, 3)), color=colors[i], edgecolor="k", alpha=0.85)
|
|
456
|
+
ax.axvline(oe["pair_angle_mean_deg"], color="k", ls="--", lw=1,
|
|
457
|
+
label=f"mean pairwise = {oe['pair_angle_mean_deg']:.3f} deg")
|
|
458
|
+
ax.set_xlabel("misorientation vs bootstrap mean U (deg)")
|
|
459
|
+
ax.set_ylabel(f"count (of {oe['n_boot']} resamples)")
|
|
460
|
+
indep = "independent" if oe["independent_resamples"] else "NOT independent (too few reflections)"
|
|
461
|
+
ax.set_title(f"domain {i}: {len(dom.hkl)} claimed refl., keep_n={oe['keep_n']}\n"
|
|
462
|
+
f"decoy={res.decoy[i]['verdict']}, resamples {indep}", fontsize=9)
|
|
463
|
+
ax.legend(fontsize=8, frameon=False)
|
|
464
|
+
fig.suptitle(f"Orientation envelope -- position {res.point} "
|
|
465
|
+
f"(/verify claim b82502d6b5ad: REFUTED as a general determinability signal; "
|
|
466
|
+
f"mechanism correctness-fixed, read decoy_test alongside this)", fontsize=9)
|
|
467
|
+
fig.tight_layout()
|
|
468
|
+
return fig
|
|
469
|
+
|
|
470
|
+
|
|
471
|
+
def plot_lattice_envelope(res: "PositionResult"):
|
|
472
|
+
"""``a``, ``b``, ``c`` bootstrap distributions, individually -- NOT just their difference.
|
|
473
|
+
|
|
474
|
+
Needs ``res.cell_bootstrap_samples``, which is only populated in-memory by
|
|
475
|
+
:func:`reduce_one_position` (never persisted through :meth:`PositionResult.to_dict`,
|
|
476
|
+
since it is one position's full raw bootstrap array, not something a whole raster's
|
|
477
|
+
JSON files should carry). Call this on a live ``reduce_one_position`` result, not on
|
|
478
|
+
something read back via :func:`assemble_raster_results`.
|
|
479
|
+
"""
|
|
480
|
+
import matplotlib.pyplot as plt
|
|
481
|
+
if res.cell_bootstrap_samples is None:
|
|
482
|
+
raise ValueError(
|
|
483
|
+
"res.cell_bootstrap_samples is None -- either no domain reached refine_cell_joint "
|
|
484
|
+
"(res.refined_cell is None), too few bootstrap resamples succeeded, or this "
|
|
485
|
+
"PositionResult came from assemble_raster_results (which does not carry it; "
|
|
486
|
+
"see this function's docstring)."
|
|
487
|
+
)
|
|
488
|
+
samples = res.cell_bootstrap_samples
|
|
489
|
+
a_fit, b_fit, c_fit = res.refined_cell[0], res.refined_cell[1], res.refined_cell[2]
|
|
490
|
+
fig, axes = plt.subplots(1, 3, figsize=(14, 4.2))
|
|
491
|
+
for j, (lab, fit_val) in enumerate(zip(["a", "b", "c"], [a_fit, b_fit, c_fit])):
|
|
492
|
+
ax = axes[j]
|
|
493
|
+
vals = samples[:, j]
|
|
494
|
+
ax.hist(vals, bins=15, color="steelblue", edgecolor="k", alpha=0.85)
|
|
495
|
+
ax.axvline(fit_val, color="crimson", lw=1.5, label=f"joint fit: {fit_val:.5f} A")
|
|
496
|
+
lo, hi = np.percentile(vals, [2.5, 97.5])
|
|
497
|
+
ax.axvspan(lo, hi, color="crimson", alpha=0.08, label=f"95% range: [{lo:.5f}, {hi:.5f}]")
|
|
498
|
+
ax.set_xlabel(f"{lab} (A)")
|
|
499
|
+
ax.set_ylabel(f"count (of {len(vals)} bootstrap resamples)")
|
|
500
|
+
ax.set_title(f"{lab}: {fit_val:.5f} +/- {vals.std(ddof=1):.5f} A", fontsize=10)
|
|
501
|
+
ax.legend(fontsize=7, frameon=False)
|
|
502
|
+
fig.suptitle(f"Lattice envelope -- position {res.point}, joint fit over "
|
|
503
|
+
f"{len(res.domains.domains)} domain(s) (quotable={res.quotable})", fontsize=10)
|
|
504
|
+
fig.tight_layout()
|
|
505
|
+
return fig
|
|
506
|
+
|
|
507
|
+
|
|
331
508
|
def reduce_one_position(
|
|
332
509
|
frames: np.ndarray, geom: Geometry, *,
|
|
333
510
|
a: float, c: float, space_group_number: int,
|
|
@@ -343,6 +520,8 @@ def reduce_one_position(
|
|
|
343
520
|
seed_from_nominal: bool = False,
|
|
344
521
|
ingest_kwargs: Optional[dict] = None,
|
|
345
522
|
find_domains_kwargs: Optional[dict] = None,
|
|
523
|
+
orientation_uncertainty: bool = False,
|
|
524
|
+
n_orientation_boot: int = 25,
|
|
346
525
|
) -> PositionResult:
|
|
347
526
|
"""One raster position, start to finish: frames -> a gated a/b splitting, if one is possible.
|
|
348
527
|
|
|
@@ -364,6 +543,18 @@ def reduce_one_position(
|
|
|
364
543
|
Nothing here decides FOR you whether a splitting is real. ``PositionResult.quotable`` is
|
|
365
544
|
only set once every gate in ``phase-3-refine.md``'s order has been checked; read
|
|
366
545
|
``PositionResult.notes`` for which one, if any, said no.
|
|
546
|
+
|
|
547
|
+
``orientation_uncertainty`` (SKETCH, 2026-09-13, off by default): also bootstrap each
|
|
548
|
+
domain's OWN orientation fit (``seed_index.bootstrap_orientation_uncertainty``,
|
|
549
|
+
``n_orientation_boot`` resamples each) and report the result in
|
|
550
|
+
``PositionResult.orientation_envelope`` -- the same "don't report a bare point estimate"
|
|
551
|
+
discipline ``split_with_error`` already applies to the cell, extended to ``U``. This is
|
|
552
|
+
NOT free: it re-runs an Adam refinement ``n_orientation_boot`` times per domain, on top of
|
|
553
|
+
everything else this function already does, so it defaults to off for raster-wide batch
|
|
554
|
+
runs where that cost multiplies by every position. Independent of the a/b gate chain --
|
|
555
|
+
computed for every domain FOUND, not only ones that reach ``refine_cell_joint``, since
|
|
556
|
+
orientation determinability is a property of a domain's own indexing fit, not of whether
|
|
557
|
+
its a/b split is separable.
|
|
367
558
|
"""
|
|
368
559
|
ik = dict(ingest_kwargs or {})
|
|
369
560
|
fk = dict(find_domains_kwargs or {})
|
|
@@ -444,6 +635,7 @@ def reduce_one_position(
|
|
|
444
635
|
decoy_list: List[dict] = []
|
|
445
636
|
feature_list: List[dict] = []
|
|
446
637
|
gate_list: List[dict] = []
|
|
638
|
+
orientation_list: List[Optional[dict]] = []
|
|
447
639
|
fit_domains = []
|
|
448
640
|
|
|
449
641
|
hkl_box = _hkl_box(hmax, hmax, lmax, space_group_number)
|
|
@@ -453,6 +645,38 @@ def reduce_one_position(
|
|
|
453
645
|
for dom in domains.domains:
|
|
454
646
|
claim_idx = np.flatnonzero(dom.claim)
|
|
455
647
|
|
|
648
|
+
# SKETCH, 2026-09-13. /verify REFUTED an early version of this on real La3Ni2O7
|
|
649
|
+
# 2601_LT p364 data (claim b82502d6b5ad, physics+statistics+artifact lenses; numbers
|
|
650
|
+
# reproduced exactly, REPRODUCTION survived -- the interpretation did not):
|
|
651
|
+
# - physics: the FIRST version called bootstrap_orientation_uncertainty with only
|
|
652
|
+
# (a, c), silently assuming h and k share one axis (tetragonal) -- wrong for any
|
|
653
|
+
# domain this project's own a/b work has already found is NOT tetragonal. Fixed:
|
|
654
|
+
# dom.lat.b is now passed explicitly.
|
|
655
|
+
# - statistics/artifact: a sparse domain's "wider envelope" is not automatically a
|
|
656
|
+
# meaningful determinability signal -- it can be (a) the generic 1/sqrt(n) scaling
|
|
657
|
+
# any bootstrap M-estimator shows regardless of data quality, compounded by (b) too
|
|
658
|
+
# few centroids to draw independent resamples at all (see
|
|
659
|
+
# `independent_resamples`/`n_distinct_subsets` in the returned dict), and (c) the
|
|
660
|
+
# domain itself may be one the pipeline's OWN honesty gates (decoy_test) do not
|
|
661
|
+
# trust -- check `res.decoy[i]['verdict']` alongside this envelope, not instead of
|
|
662
|
+
# it, before reading a wide spread as "this domain's orientation is real but
|
|
663
|
+
# uncertain" rather than "this domain's fit may not be trustworthy at all."
|
|
664
|
+
# Off by default (orientation_uncertainty=False) -- see reduce_one_position's
|
|
665
|
+
# docstring for the cost tradeoff. Computed for every domain FOUND, independent of the
|
|
666
|
+
# a/b gate chain below (a domain's orientation determinability is not the same
|
|
667
|
+
# question as whether its a/b split is separable) -- but NOT independent of whether
|
|
668
|
+
# the domain's own fit passes `decoy_test`; read both together.
|
|
669
|
+
if orientation_uncertainty:
|
|
670
|
+
centroids = list(zip((tuple(int(x) for x in row) for row in dom.hkl), q[claim_idx]))
|
|
671
|
+
if len(centroids) >= 4:
|
|
672
|
+
orientation_list.append(bootstrap_orientation_uncertainty(
|
|
673
|
+
centroids, dom.U, a=float(dom.lat.a), b=float(dom.lat.b), c=float(dom.lat.c),
|
|
674
|
+
n_boot=n_orientation_boot))
|
|
675
|
+
else:
|
|
676
|
+
orientation_list.append(None) # bootstrap_orientation_uncertainty floors keep_n at 4
|
|
677
|
+
else:
|
|
678
|
+
orientation_list.append(None)
|
|
679
|
+
|
|
456
680
|
# completeness: predict this domain's own cell/orientation over the WHOLE box, and
|
|
457
681
|
# separately over its own claimed hkl (to derive the window from its own residuals --
|
|
458
682
|
# never guessed, per `window_from_residuals`).
|
|
@@ -568,6 +792,8 @@ def reduce_one_position(
|
|
|
568
792
|
|
|
569
793
|
refined_cell = None
|
|
570
794
|
split_pct = None
|
|
795
|
+
cell_bootstrap_samples = None
|
|
796
|
+
cell_sigma_bootstrap = None
|
|
571
797
|
if fit_domains:
|
|
572
798
|
from midas_hkls import DomainData, refine_cell_joint, split_with_error
|
|
573
799
|
dd = [DomainData(hkl=d.hkl, g=q[np.flatnonzero(d.claim)], label=f"domain{i}")
|
|
@@ -577,10 +803,12 @@ def reduce_one_position(
|
|
|
577
803
|
n_bootstrap=n_bootstrap)
|
|
578
804
|
refined_cell = fit.cell
|
|
579
805
|
split_pct = split_with_error(fit)
|
|
806
|
+
cell_bootstrap_samples = fit.cell_bootstrap_samples # (n_kept, 6): a,b,c,alpha,beta,gamma
|
|
807
|
+
cell_sigma_bootstrap = fit.cell_sigma_bootstrap
|
|
580
808
|
else:
|
|
581
809
|
notes.append("no domain passed the a/b gate chain; no cell was jointly refined")
|
|
582
810
|
|
|
583
|
-
fit_passed_decoy = all(decoy_list[i]["verdict"]
|
|
811
|
+
fit_passed_decoy = all(decoy_list[i]["verdict"] == "informative"
|
|
584
812
|
for i, g in enumerate(gate_list) if g.get("passed_to_refine"))
|
|
585
813
|
quotable = (
|
|
586
814
|
bool(fit_domains)
|
|
@@ -589,8 +817,16 @@ def reduce_one_position(
|
|
|
589
817
|
and np.isfinite(split_pct[1])
|
|
590
818
|
)
|
|
591
819
|
if fit_domains and not fit_passed_decoy:
|
|
592
|
-
|
|
593
|
-
|
|
820
|
+
failing_verdicts = sorted({decoy_list[i]["verdict"]
|
|
821
|
+
for i, g in enumerate(gate_list) if g.get("passed_to_refine")}
|
|
822
|
+
- {"informative"})
|
|
823
|
+
if "real_fails" in failing_verdicts:
|
|
824
|
+
notes.append("a domain that passed the a/b gate chain did not pass the decoy test "
|
|
825
|
+
"(the real fit itself did not clear the acceptance threshold) -- "
|
|
826
|
+
"not quotable")
|
|
827
|
+
if "uninformative" in failing_verdicts:
|
|
828
|
+
notes.append("a domain that passed the a/b gate chain did not pass the decoy test "
|
|
829
|
+
"(a deliberately wrong cell scored as well) -- not quotable")
|
|
594
830
|
if split_pct is not None and not np.isfinite(split_pct[1]):
|
|
595
831
|
notes.append("a and b are equal by the refined symmetry: no splitting is being measured")
|
|
596
832
|
|
|
@@ -599,7 +835,9 @@ def reduce_one_position(
|
|
|
599
835
|
completeness=completeness_list, recovery=recovery_list,
|
|
600
836
|
decoy=decoy_list, feature_in_raw=feature_list, ab_gates=gate_list,
|
|
601
837
|
refined_cell=refined_cell, split_pct=split_pct, notes=notes,
|
|
602
|
-
quotable=bool(quotable)
|
|
838
|
+
quotable=bool(quotable), orientation_envelope=orientation_list,
|
|
839
|
+
cell_bootstrap_samples=cell_bootstrap_samples,
|
|
840
|
+
cell_sigma_bootstrap=cell_sigma_bootstrap)
|
|
603
841
|
|
|
604
842
|
|
|
605
843
|
# ---------------------------------------------------------------------------
|