aeat.adapters.inbound.pdf._shared module

Shared strict+frozen records for casilla-bearing PDF imports.

ExtractedCasilla is the boundary-crossing record produced by declaracion and borrador extractors after they find a printed casilla value. It pairs the stable casilla identifier with the typed value and extraction provenance (page, optional bounding box, confidence) so downstream consumers can classify discrepancies and attach warnings without re-parsing the PDF.

Justificante PDFs do not carry per-casilla values; that receipt surface shares the package’s error and text/provenance helpers, but not this record.

class ExtractedCasilla(**data)[source]

Bases: BaseModel

One casilla ID + printed value extracted from a filing PDF.

Variables:

  • casilla_id – Stable casilla identifier (e.g. "01", "071"). Aligned to the canonical CasillaId constraint (max_length=64, pattern [A-Za-z0-9][A-Za-z0-9._:-]*).

  • printed_value – Typed value as printed on the PDF. None when the casilla was located but blank on the page. Monetary values are parsed into decimal.Decimal; non-monetary casillas may retain typed integers, strings, booleans, or dates.

  • source_page – 1-based page number the value was read from.

  • source_bbox – Optional (x0, y0, x1, y1) bounding box in pdfplumber coordinates. Populated by the bbox-anchored extraction primitive; None for label-regex / AcroForm extractions where coordinates are not captured.

  • extraction_confidence – Float in [0.0, 1.0]. 1.0 for AcroForm + unambiguous label-regex hits; lower when the extractor fell back to bbox-anchored recovery or had to disambiguate competing label candidates.

Parameters:
casilla_id: CasillaId
printed_value: Decimal | int | str | bool | date | None
source_page: int
source_bbox: tuple[float, float, float, float] | None
extraction_confidence: float