tests/test_pdf_mapping.py

"""Mappatura span→box: il box porta il testo originale che copre.

Serve al mascheramento per-carattere del PDF (il box deve conoscere il testo da
mascherare). Test puro, senza PDF reale (geometria sintetica)."""

from __future__ import annotations

from core.contracts import LayerPriority, Span
from core.contracts.extraction import PageGeometry, WordBox
from pdf.mapping import map_spans_to_boxes


def _word(text, x0, y0, x1, y1, cs, ce):
    return WordBox(text=text, page=0, x0=x0, y0=y0, x1=x1, y1=y1,
                   char_start=cs, char_end=ce)


def _span(start, end, text, etype="PERSONA"):
    return Span(start, end, text, etype, 0.95, LayerPriority.RULES)


def _boxes(page, spans):
    return map_spans_to_boxes(
        spans, (page,),
        severity_of=lambda t: "alta",
        label_of=lambda t: f"[{t}]",
    )


def test_box_carries_covered_text_single_row():
    # "Mario Rossi" su una riga → un box col testo originale coperto.
    page = PageGeometry(page=0, width=600, height=800, words=(
        _word("Mario", 72, 100, 120, 112, 0, 5),
        _word("Rossi", 124, 100, 175, 112, 6, 11),
    ))
    boxes = _boxes(page, [_span(0, 11, "Mario Rossi")])
    assert len(boxes) == 1
    assert boxes[0].text == "Mario Rossi"


def test_box_text_splits_per_row_for_multiline_entity():
    # Entità a capo: due box, ciascuno col testo della propria riga.
    page = PageGeometry(page=0, width=600, height=800, words=(
        _word("Mario", 72, 100, 120, 112, 0, 5),
        _word("Rossi", 72, 130, 123, 142, 6, 11),  # riga sotto
    ))
    boxes = _boxes(page, [_span(0, 11, "Mario Rossi")])
    texts = sorted(b.text for b in boxes)
    assert texts == ["Mario", "Rossi"]