llm-ingestion-okf/tests/test_propose_sheet_sections.py

"""D3 round 3: section rows INSIDE a sheet's one table block, behind a flag.

The spreadsheet is the one file type no arm has ever moved. Its whole body
extracts as ONE continuous pipe-table block -- one heading and 101 rows -- so
every rule this module has proposes exactly one boundary for it, and the
operator's worksheet asks for one concept per numbered cost group. Those groups
are ROWS, not headings, which is why no heading rule can reach them and why
`--keep-table-heading` moved the first byte without moving the count.

The rule tested here is the opposite direction from `--table-grid`: that one
stops a converter's rule line from CLOSING a block, so one grid table proposes
one candidate instead of many; this one CUTS an open block at the rows that
label its sections. They read different grammars (a `+---+` rule line against a
pipe row's first cell) and compose in one order -- grid decides how far a block
extends, section rows decide where it is cut inside.

Generality is the whole point, so the fixtures below carry no word list and no
knowledge of any real document's labels: what makes a row a section row is that
its first cell is a bare numeric label and that it is one of a RUN of such rows.
The run is the same bounding device `CONTENTS_RUN` already uses in this module,
for the same reason: a single numbered row is a quantity, not a section.
"""

from __future__ import annotations

from pathlib import Path

import pytest

from llm_ingestion_okf import propose as okf_propose_segments

# One heading, one continuous table block, a preamble, a run of four numbered
# section rows, and a summing tail. The shape of a priced sheet, written from
# the grammar rather than copied from one.
SECTIONED_SHEET = """## Kostnadsoversikt {#sheet-1}

| Skjema |  |  |
|----|----|----|
| Skjemaet fylles ut i sin helhet. |  |  |
| Post |  | SUM |
| 01 | Felleskostnader |  |
| 11+12 | Rigging og drift |  |
| 02 | Bygning |  |
| 07 | Utendoers |  |
| Tilbudt fastpris eksklusive avgift |  | 100 |
"""

# The same block with the numbered run removed. Nothing here is a section row,
# so the flag must leave it exactly as it found it.
UNSECTIONED_SHEET = """## Kostnadsoversikt {#sheet-1}

| Skjema |  |  |
|----|----|----|
| Skjemaet fylles ut i sin helhet. |  |  |
| Post |  | SUM |
| Felleskostnader |  | 10 |
| Rigging og drift |  | 20 |
| Tilbudt fastpris eksklusive avgift |  | 100 |
"""

# A single row whose first cell is a number -- a computation basis, not a
# section. This is the known-negative the run threshold exists for, and it is
# the shape that actually occurs: a sheet states a quantity on its own row.
LONE_NUMERIC_ROW = """## Regningsarbeider {#sheet-2}

| Grunnlag |  |  |
|----|----|----|
| Beregningsgrunnlag massehaandtering | tonn | 16000 |
| Paaslag i prosent |  | 10 |
"""


def test_a_run_of_numbered_rows_cuts_the_block_into_sections() -> None:
    """The rule, red first: one candidate per section row, named from the row.

    Default: the whole sheet is ONE table block whose title is inherited from
    the orphaned heading. On: the preamble keeps that block and each numbered
    row opens its own candidate, running to the next section row or to the end
    of the block -- so the LAST section carries the sheet's tail, which is the
    honest consequence of a rule that cuts and never discards.
    """
    # The title carries NO `{#sheet-1}` since 2026-09-09: the converter anchor
    # is stripped where the title is formed, because a concept's id is reduced
    # from its title and the two must not name the same concept differently.
    # The fixture keeps the anchor -- it is what pandoc writes, and removing it
    # from the input would test the strip against text that never has it.
    off = okf_propose_segments.find_candidates(SECTIONED_SHEET)
    assert [(c.title, c.rule) for c in off] == [
        ("Kostnadsoversikt", okf_propose_segments.RULE_TABLE_BLOCK)
    ]

    on = okf_propose_segments.find_candidates(SECTIONED_SHEET, sheet_section_rows=True)
    assert [(c.number, c.title, c.rule) for c in on] == [
        (None, "Kostnadsoversikt", okf_propose_segments.RULE_TABLE_BLOCK),
        ("01", "01 Felleskostnader", okf_propose_segments.RULE_SHEET_SECTION),
        ("11+12", "11+12 Rigging og drift", okf_propose_segments.RULE_SHEET_SECTION),
        ("02", "02 Bygning", okf_propose_segments.RULE_SHEET_SECTION),
        ("07", "07 Utendoers", okf_propose_segments.RULE_SHEET_SECTION),
    ]
    # The preamble block ends where the first section opens, and the last
    # section reaches the end of the text.
    assert on[0].end == on[1].start
    assert on[1].end == on[2].start
    assert on[-1].end == len(SECTIONED_SHEET)
    # The tail row is inside the last section rather than in no concept at all.
    assert "Tilbudt fastpris" in SECTIONED_SHEET[on[-1].start : on[-1].end]


def test_a_section_row_survives_the_orphan_check() -> None:
    """A stated exception, because without it the rule cannot fire at all.

    The orphan check reads the lines UNDER a candidate's first line, which is
    right for a heading -- a heading with nothing under it names nothing. A
    section row carries its content in its own cells, so every one-row section
    is bodiless by that test and all but the last would be dropped. The
    exception is written once, on the rule, and it is what the count below
    proves.
    """
    on = okf_propose_segments.find_candidates(SECTIONED_SHEET, sheet_section_rows=True)
    single_row = [c for c in on if c.rule == okf_propose_segments.RULE_SHEET_SECTION][:3]
    assert len(single_row) == 3
    for candidate in single_row:
        assert len(SECTIONED_SHEET[candidate.start : candidate.end].splitlines()) == 1


def test_a_table_without_section_rows_is_untouched() -> None:
    """First known-negative: identical objects, not merely an equal count."""
    off = okf_propose_segments.find_candidates(UNSECTIONED_SHEET)
    on = okf_propose_segments.find_candidates(UNSECTIONED_SHEET, sheet_section_rows=True)
    assert len(off) == 1
    assert off == on


def test_a_lone_numbered_row_is_not_a_section() -> None:
    """Second known-negative, and the reason the rule counts a RUN.

    `16000` is a quantity. A rule that read one numbered cell as a section
    would cut a sheet at every stated basis, which is the ungated widening this
    module has already refused once for outline candidates.
    """
    off = okf_propose_segments.find_candidates(LONE_NUMERIC_ROW)
    on = okf_propose_segments.find_candidates(LONE_NUMERIC_ROW, sheet_section_rows=True)
    assert len(off) == 1
    assert off == on


def test_the_flag_composes_with_keep_table_heading() -> None:
    """Both on: the heading keeps the preamble, the sections stay separate.

    D1 decides where the FIRST concept starts; this rule decides where the
    block is cut. They read different things and neither disables the other,
    which is what makes the pair safe to expose as two flags rather than one.
    """
    on = okf_propose_segments.find_candidates(
        SECTIONED_SHEET, sheet_section_rows=True, keep_table_heading=True
    )
    assert [(c.number, c.rule) for c in on] == [
        (None, okf_propose_segments.RULE_HEADING),
        ("01", okf_propose_segments.RULE_SHEET_SECTION),
        ("11+12", okf_propose_segments.RULE_SHEET_SECTION),
        ("02", okf_propose_segments.RULE_SHEET_SECTION),
        ("07", okf_propose_segments.RULE_SHEET_SECTION),
    ]
    assert on[0].start == 0
    assert on[0].end == on[1].start


def test_the_rule_names_itself_in_derived() -> None:
    """A proposal an operator disagrees with is traceable to the rule."""
    plan = okf_propose_segments.build_plan(
        Path("ark.xlsx"),
        SECTIONED_SHEET,
        b"ark",
        okf_type="requirement",
        proposed_at="1970-01-01T00:00:00Z",
        sheet_section_rows=True,
    )
    sections = [e for e in plan["entries"] if e["title"].startswith("01")]
    assert len(sections) == 1
    assert sections[0]["derived"] == [
        okf_propose_segments.PROPOSED_MARKER,
        okf_propose_segments.RULE_SHEET_SECTION,
    ]
    # The label becomes the directory and is stripped from the stem, the same
    # way a section number does for a heading.
    assert sections[0]["path"] == "01/felleskostnader.md"


def test_sheet_section_rows_takes_no_argument(
    tmp_path: Path, capsys: pytest.CaptureFixture[str]
) -> None:
    """A boolean at the CLI. The rule's one number is a module constant."""
    source = tmp_path / "ark.md"
    source.write_text(SECTIONED_SHEET, encoding="utf-8")
    out = tmp_path / "ark.json"
    with pytest.raises(SystemExit) as exit_info:
        okf_propose_segments.main([str(source), "--out", str(out), "--sheet-section-rows", "3"])
    assert exit_info.value.code == 2
    assert "usage:" in capsys.readouterr().err