tests/test_parser_output.py

from __future__ import annotations

import pytest

from parser_output import (
    extract_bounding_boxes,
    extract_tables,
    render_text,
    shape_parse_response,
    shape_parser_output,
)


def test_extracts_and_normalizes_bounding_boxes() -> None:
    raw = """
{"label":"title","bbox":[100,200,900,400]}
<bbox label="table">[0.1, 0.3, 0.8, 0.9]</bbox>
{"label":"invalid","bbox":[5,5,1,1]}
"""

    boxes = extract_bounding_boxes(raw)

    assert len(boxes) == 2
    assert boxes[0]["label"] == "title"
    assert boxes[0]["bbox"] == [0.1, 0.2, 0.9, 0.4]
    assert boxes[1]["bbox"] == [0.1, 0.3, 0.8, 0.9]


def test_sanitizes_tables() -> None:
    raw = """
<table onclick="steal()">
  <tr><th scope="col">Name</th><td><script>alert(1)</script>Ada</td></tr>
</table>
"""

    table = extract_tables(raw)

    assert "<table" in table
    assert 'scope="col"' in table
    assert "onclick" not in table
    assert "<script" not in table


def test_sanitizes_rendered_markdown() -> None:
    rendered = render_text(
        '# Heading\n\n<script>alert(1)</script>\n\n[unsafe](javascript:alert(1))'
    )

    assert "<h1>Heading</h1>" in rendered
    assert "<script" not in rendered
    assert "javascript:" not in rendered


def test_rendered_text_keeps_a_visible_marker_for_html_tables() -> None:
    rendered = render_text(
        "html:\n<table><tr><td>Quarterly revenue</td></tr></table>"
    )

    assert "html:" in rendered
    assert "<code>&lt;table&gt;...&lt;/table&gt;</code>" in rendered
    assert "Quarterly revenue" not in rendered
    assert "<table>" not in rendered


def test_table_only_output_has_nonempty_text_html() -> None:
    shaped = shape_parser_output("<table><tr><td>Ada</td></tr></table>")

    assert shaped["tables_html"] == "<table><tbody><tr><td>Ada</td></tr></tbody></table>"
    assert "&lt;table&gt;...&lt;/table&gt;" in shaped["text_html"]


def test_shapes_empty_output() -> None:
    assert shape_parser_output("") == {
        "raw_output": "",
        "output_chars": 0,
        "boxes": [],
        "tables_html": "",
        "text_html": "",
    }


def test_shapes_structured_parse_blocks() -> None:
    body = {
        "id": "parse-123",
        "pages": [
            {
                "index": 0,
                "type": "blocks",
                "blocks": [
                    {"type": "text", "text": {"content": "# Report"}},
                    {
                        "type": "table",
                        "table": {
                            "type": "html",
                            "title": "Revenue",
                            "html": (
                                '<table onclick="bad()"><tr><td>42</td></tr></table>'
                            ),
                            "bounding_box_normalized": {
                                "top_left_x": 0.1,
                                "top_left_y": 0.2,
                                "bottom_right_x": 0.8,
                                "bottom_right_y": 0.7,
                            },
                        },
                    },
                    {
                        "type": "image",
                        "image": {
                            "id": "img-0",
                            "description": "A bar chart",
                            "category": "other",
                            "bounding_box_normalized": {
                                "top_left_x": 0.2,
                                "top_left_y": 0.3,
                                "bottom_right_x": 0.9,
                                "bottom_right_y": 0.8,
                            },
                        },
                    },
                ],
            }
        ],
        "meta": {"billed_units": {"pages": 1}},
    }

    shaped = shape_parse_response(body)

    assert shaped["text_output"].startswith("# Report")
    assert "Image: A bar chart" in shaped["text_output"]
    assert "<h1>Report</h1>" in shaped["text_html"]
    assert "onclick" not in shaped["tables_html"]
    assert [box["label"] for box in shaped["boxes"]] == ["Revenue", "other"]
    assert shaped["boxes"][0]["bbox"] == [0.1, 0.2, 0.8, 0.7]
    assert '"id": "parse-123"' in shaped["raw_output"]
    assert shaped["response"] == body


def test_shapes_markdown_parse_page() -> None:
    shaped = shape_parse_response(
        {
            "id": "parse-456",
            "pages": [
                {
                    "index": 0,
                    "type": "markdown",
                    "markdown": {
                        "content": "## Heading\n\nBody",
                        "images": [],
                    },
                }
            ],
        }
    )

    assert shaped["text_output"] == "## Heading\n\nBody"
    assert "<h2>Heading</h2>" in shaped["text_html"]


@pytest.mark.parametrize(
    "body",
    [
        {},
        {"pages": []},
        {"pages": [{"index": 0, "type": "unknown"}]},
        {"pages": [{"index": 0, "type": "blocks", "blocks": [{}]}]},
    ],
)
def test_rejects_invalid_structured_parse_responses(body: dict) -> None:
    with pytest.raises((TypeError, ValueError)):
        shape_parse_response(body)