Skip to content

SEAD numerical analysis value lineage

The source-bound value export reconciles all 60 captured tbl_analysis_numerical_values rows, keyed by analysis_numerical_value_id, with 60 distinct parent tbl_analysis_values.analysis_value_id rows. Each parent joins through its analysis entity, physical sample and sample group to SEAD site 6486 (Skänninge Abbey). The rows span five physical samples and ten datasets. The report binds the exact admitted raw table digests, run, build, admission and normalized materialization identities.

The typed value and parent analysis_value text are separate source fields. They are numerically equal in 43 rows; 17 rows differ in their reported precision. For example, numerical row 21 reports 3.90122 while its parent reports 3.9012199999999995. Both source spellings remain in the export, alongside each complete raw row, source key, normalized observation ID, relation keys, null states, qualifier, variant flag and refusal reasons. All 60 typed and parent values are reported; no source null or parent parse failure occurs in this capture. The producer retains distinct states for future source-null or parse-failed rows.

The nine source value classes describe ancient DNA laboratory measurements, including read length, sequence coverage, endogenous content and read damage. Their two source value types are Measurement and Percentage. The export binds each class and type to its admitted raw row and normalized semantic relation. Neither type reports a unit ID for these classes. No physical unit is inferred from a class name. All 60 source observations remain refused as scientific events; this report does not admit aDNA, vegetation evidence or a dated event. The atlas currently publishes the parent analysis text without its typed numerical component; this export provides a public, keyed review of that source component.

Reproduce

From the repository root, choose an absent destination under artifacts/:

PYTHONDONTWRITEBYTECODE=1 OMP_NUM_THREADS=1 OPENBLAS_NUM_THREADS=1 \
  MKL_NUM_THREADS=1 NUMEXPR_NUM_THREADS=1 VECLIB_MAXIMUM_THREADS=1 \
  .venv/bin/python -m bijux_pollenomics_dev.source_completion.sead_analysis_numerical_values \
  --data-root data --output artifacts/sead-analysis-numerical-values/replay

The producer verifies admitted raw bytes, normalized part digests, source rows, parent and entity links, value states and refusals before writing. It refuses an existing output directory. Compare values.csv and report.json byte for byte with this directory after replay.