fix: NovaBio Flex2 - do not emit measurements that have no measured value CUST-12697

[ajcariaga16]

Summary

Converting a NovaBio BioProfile Flex2 SampleResults CSV still fails schema validation after #1261 whenever a row has no cell counts but either (a) a Cell Density Dilution value, or (b) the file has a unit row, which sets the cell density / PO2 / PCO2 unit strings on every sample. Reported by a customer (CUST-12475, follow-up to the pH-only failure fixed in #1261).

Root cause

_get_measurements() emitted a measurement for a detection type whenever any of its mapped properties was non-None:

if any(value is not None for value in kwargs.values()):

That includes *_unit fields and the data processing fields (cell_type_processing_method, cell_density_dilution_factor), none of which are measured values. A chemistry-only row then produced a cell-counting measurement with no processed data (or only a cell density dilution factor), and the cell-counting schema requires viability (cell counter) and viable cell density (cell counter):

Failed to validate allotrope model against schema: {'device control aggregate document': {'device control document': [{'detection type': 'cell-counting', ...}]}, ..., 'processed data aggregate document': {'processed data document': [{'data processing document': {'cell density dilution factor': {'value': 1.0, 'unit': '(unitless)'}}}]}} is not valid under any of the given schemas

An empty analyte list ([]) was also counted as a value, producing a metabolite measurement with no analyte document.

Fix

Add MEASURED_PROPERTIES_BY_DETECTION_TYPE listing the measured-value properties of each detection type, and only emit a measurement when at least one of them is present. This generalizes the pH-only guard from #1261 (REQUIRED_PROPERTY_BY_DETECTION_TYPE), which it replaces: pH remains the only measured property of ph-detection, since vessel temperature is the temperature at which pH was measured rather than a measurand.

Behavior for rows that do have measured values is unchanged. A row with partial cell counts (e.g. total density without viability) still fails validation as before; that is inconsistent data and dropping it silently seemed worse than a clear error.

Testing

• New fixture SampleResultsT000000001C2026-01-01_120000.csv (anonymized, derived from the customer file) with a unit row and three rows: full cell counts, chemistry-only, and dilution-setting-only. Without the fix it fails with the error above; expected JSON generated with --overwrite.
• New parametrized structure tests covering unit-only, dilution-only, empty-analyte and present-value cases, plus a check that measured properties are a subset of the mapped properties.
• The customer's 12-sample file converts successfully with this change.
• hatch run test_all.py3.10:pytest tests: 1538 passed. ruff, black --check and mypy clean on the changed files.

🤖 Generated with Claude Code