Canonical Pipeline ================== Phase 3 adds a package-backed canonical import and export layer for approved workbook-shaped inputs. The pipeline is intentionally conservative: it converts known sheets into structured records, preserves source-attribution rows for validation, and emits diagnostics when values are incomplete or unsupported. Imported Sheets --------------- ``Species_Master`` Candidate species rows using the canonical master species column order from ``schemas/master_species_columns.csv``. Legacy workbook headers such as ``Species_ID`` and ``Botanical_Name`` are normalized to canonical names such as ``Species ID`` and ``Botanical Name``. ``Lookup_Tables`` Controlled vocabulary rows. P3 supports simple ``lookup_name,value`` shaped rows and conservative workbook-block extraction. ``Source_Attribution`` Evidence rows normalized into the P2 source-attribution validator. These rows support future review and joins; they are not treated as ecological facts by import alone. ``Bloom_Calendar`` Month columns are imported as candidate bloom-calendar rows. P3 does not calculate scores or infer missing bloom values. Validation Rules ---------------- Canonical import emits structured diagnostics for missing required species fields, invalid ``BCNPPD-0001`` species IDs, invalid evidence confidence values, duplicate species IDs, excluded source URLs, invalid source-attribution rows, and sheets that are not imported by the canonical pipeline. Warnings, such as ignored dashboard sheets, do not fail the CLI. Error diagnostics produce a nonzero CLI exit code. CLI --- Inspect a workbook without writing files: .. code-block:: shell bc-nppd import-canonical-workbook data/workbooks/native_plant_restoration_workbook_v1.0.0c.xlsx --json Export deterministic CSV tables to an ignored output directory: .. code-block:: shell bc-nppd export-canonical-workbook data/workbooks/native_plant_restoration_workbook_v1.0.0c.xlsx --out-dir outputs/p3 --json Exported tables are intended for local review under ``outputs/`` unless a later release task explicitly approves a public tracked derivative. FreshForge Shape ---------------- P3 records an optional FreshForge workflow shape but does not add FreshForge as a dependency. A future workflow provider can map these steps onto reproducible nodes: * ``workbook.inventory`` * ``workbook.import_tables`` * ``tables.validate`` * ``tables.export`` The example YAML in ``examples/p3_canonical_pipeline_freshforge.yaml`` is a planning artifact only. CI does not execute it.