from pathlib import Path

from benchmark import Scenario, build_command, build_scenarios, parse_threads, render_report, ScenarioResult


def test_build_default_matrix():
    scenarios = build_scenarios(
        ["tiny", "small", "medium"],
        ["fast", "standard"],
        [144],
        [None],
    )
    assert len(scenarios) == 6
    assert scenarios[0].name == "tiny-fast-dpi144-thauto-s64min-d0p3-b0p6-u1p5-r0-rb6"
    assert scenarios[-1].name == "medium-standard-dpi144-thauto-s64min-d0p3-b0p6-u1p5-r0-rb6"


def test_parse_threads():
    assert parse_threads(["auto", "4", "4"]) == [None, 4]


def test_command_contains_model_and_profile_flags(tmp_path):
    scenario = Scenario("small", "fast", 120, 4)
    command = build_command(
        scenario,
        data_dir=tmp_path / "data",
        output_dir=tmp_path / "output",
        device="cpu",
        device_id=0,
        warmup=1,
        rounds=2,
        pdf_mode="ocr",
    )
    assert command[1].endswith("ocr.py")
    assert command[command.index("--model-size") + 1] == "small"
    assert "--no-doc-orientation-classify" in command
    assert "--no-textline-orientation" in command
    assert command[command.index("--threads") + 1] == "4"
    assert command[command.index("--text-det-thresh") + 1] == "0.3"
    assert command[command.index("--text-recognition-batch-size") + 1] == "6"


def test_report_contains_comparison_table(tmp_path):
    scenario = Scenario("tiny", "fast", 144, None)
    result = ScenarioResult(
        scenario=scenario,
        status="completed",
        exit_code=0,
        wall_seconds=3.0,
        model_init_seconds=1.0,
        pure_ocr_seconds=2.0,
        image_inference_seconds=1.0,
        pdf_ocr_seconds=1.0,
        processed_files=2,
        image_files=1,
        pdf_files=1,
        ocr_pages=1,
        recognized_lines=10,
        mean_confidence=0.95,
        units_per_second=1.0,
        output_dir=str(tmp_path / "output"),
        log_file=str(tmp_path / "run.log"),
        command=["python", "ocr.py"],
    )
    from datetime import datetime

    report = render_report(
        [result],
        data_dir=Path("data"),
        device="cpu",
        pdf_mode="ocr",
        warmup=1,
        rounds=1,
        started_at=datetime.now().astimezone(),
        finished_at=datetime.now().astimezone(),
    )
    assert "# PP-OCRv6 参数测试报告" in report
    assert "tiny-fast-dpi144-thauto-s64min-d0p3-b0p6-u1p5-r0-rb6" in report
    assert "纯 OCR 耗时最短" in report