2026Technology-Competition/docs/superpowers/plans/2026-08-08-milestone2-excel-parser.md

Phase1 里程碑2 Parser Agent - Excel 解析 实施计划

For agentic workers: REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (- [ ]) syntax for tracking.

Goal: 实现 ExcelParser，将要件定义 Excel（3 种模式：表格型/自由记述型/混合型）解析为 ExcelParseResult（tables + comments），为后代 Agent 提供唯一数据源。

Architecture: 在已完成的 data_models.py 基础上，按 design §3.5 拆分为职责单一模块并由 ExcelParser 编排。自由记述型的 LLM 结构化依赖 InferenceEngine（后续里程碑），本里程碑交付可独立测试的文本分段与占位表（extraction_method="llm_from_free_text"）。

Tech Stack: Python ≥3.11、openpyxl ≥3.1、pytest 8。

Global Constraints

• 仅读取 .xlsx；.xls 由 excel_reader.open_workbook 抛出 ValueError。
• Sheet 类型判定关键词为日文（design §3.5.1），用于匹配真实日文要件定义（技术必要保留）。
• 每个单元格生成 Provenance；source_uri 格式 file.xlsx#SheetName!CellRef（design §9.4.5，build_source_uri 提供）。
• 合并单元格下行填充（forward_fill，design §3.5.3）。
• 取消线/背景色/批注保留在 CellFormatting / CellComment（design §3.5.4 / §3.5.5）。
• 每个单元格使用 data_models 的 CellValue（字段 value / provenance / formatting / comment）。
• 交流语言统一中文（注释/提交信息/本计划正文）；标识符与技术名保留英文/日文关键词。
• 每个任务结束前 pytest 全绿并提交。

---

Task 1: ExcelParser 基础结构与来源标注

Files:

• Create: src/genesis/parsers/__init__.py
• Create: src/genesis/parsers/provenance.py
• Create: src/genesis/parsers/excel_reader.py
• Create: tests/excel_helpers.py
• Create: tests/test_excel_reader.py

Interfaces:

• Produces:
  • genesis.parsers.provenance.build_source_uri(file_name: str, sheet_name: str, cell_ref: str) -> str
  • genesis.parsers.excel_reader.open_workbook(path: str | Path) → openpyxl.Workbook（仅 .xlsx）
  • genesis.parsers.excel_reader.sheet_matrix(ws) -> list[list[Any]]

测试辅助 tests/excel_helpers.py 供全部任务复用。

• Step 1: 写失败测试

tests/excel_helpers.py:

from openpyxl import Workbook


def new_workbook(sheets: dict[str, list[list]]) -> Workbook:
    """生成临时 Workbook：key=Sheet 名，value=grid（cell 值）。"""
    wb = Workbook()
    wb.remove(wb.active)
    for name, grid in sheets.items():
        ws = wb.create_sheet(name)
        for r, row in enumerate(grid, start=1):
            for c, value in enumerate(row, start=1):
                ws.cell(row=r, column=c, value=value)
    return wb


def save_workbook(tmp_path, wb: Workbook) -> str:
    """落盘到 tmp_path 并返回路径字符串。"""
    path = tmp_path / "source.xlsx"
    wb.save(path)
    return str(path)

tests/test_excel_reader.py:

import pytest

from genesis.parsers.excel_reader import open_workbook, sheet_matrix
from genesis.parsers.provenance import build_source_uri

from tests.excel_helpers import new_workbook, save_workbook


def test_build_source_uri_format():
    assert build_source_uri("要求.xlsx", "機能一覧", "A5") == "要求.xlsx#機能一覧!A5"


def test_open_workbook_and_sheet_matrix(tmp_path):
    wb = new_workbook({"機能一覧": [["機能ID", "機能名"], ["F001", "社員登録"]]})
    path = save_workbook(tmp_path, wb)
    ws = open_workbook(path)["機能一覧"]
    assert sheet_matrix(ws) == [["機能ID", "機能名"], ["F001", "社員登録"]]


def test_open_workbook_rejects_xls(tmp_path):
    bad = tmp_path / "old.xls"
    bad.write_bytes(b"not really xls")
    with pytest.raises(ValueError):
        open_workbook(bad)

• Step 2: 运行确认失败

Run: python -m pytest tests/test_excel_reader.py -v Expected: FAIL（ModuleNotFoundError: No module named 'genesis.parsers'）

• Step 3: 实现基础模块

src/genesis/parsers/__init__.py:

"""Parser Agent：输入资料解析层。"""

src/genesis/parsers/provenance.py:

def build_source_uri(file_name: str, sheet_name: str, cell_ref: str) -> str:
    """单元格来源 URI：file.xlsx#SheetName!CellRef"""
    return f"{file_name}#{sheet_name}!{cell_ref}"

src/genesis/parsers/excel_reader.py:

from __future__ import annotations

from pathlib import Path
from typing import Any

from openpyxl import load_workbook
from openpyxl.worksheet.worksheet import Worksheet


def open_workbook(path: str | Path):
    """普通模式打开 .xlsx（保留公式/样式/批注），.xls 报错。"""
    path = Path(path)
    if path.suffix.lower() != ".xlsx":
        raise ValueError(f"不支持的 Excel 格式: {path.suffix}")
    return load_workbook(path)


def sheet_matrix(ws: Worksheet) -> list[list[Any]]:
    """整表矩形值（含 None），保留到 max_column。"""
    return [[cell.value for cell in row] for row in ws.iter_rows()]

• Step 4: 运行确认通过

Run: python -m pytest tests/test_excel_reader.py -v Expected: PASS（3 passed）

• Step 5: Commit

git add src/genesis/parsers tests/excel_helpers.py tests/test_excel_reader.py
git commit -m "feat: ExcelParser 基础（workbook/sheet 读取 + source_uri）"

注：计划中 Step 3 与 Step 4 之间未设独立 Step；Step 3 写完后直接执行 Step 4 运行验证。

---

Task 2: SheetDetector（自动识别 Sheet 类型）

Files:

• Create: src/genesis/parsers/sheet_detector.py
• Create: tests/test_sheet_detector.py

Interfaces:

• Consumes: data_models.SheetType

• Produces: detect_sheet_type(sheet_name: str, matrix: list[list[Any]]) -> SheetType （优先级：Sheet 名关键词 → 表头关键词 → GENERIC）

• Step 1: 写失败测试

tests/test_sheet_detector.py:

from genesis.data_models import SheetType
from genesis.parsers.sheet_detector import detect_sheet_type


def test_detect_by_sheet_name():
    assert detect_sheet_type("機能一覧", []) == SheetType.FUNCTION
    assert detect_sheet_type("画面一覧", []) == SheetType.SCREEN
    assert detect_sheet_type("帳票一覧", []) == SheetType.REPORT
    assert detect_sheet_type("DB定義", []) == SheetType.DATABASE
    assert detect_sheet_type("IF定義", []) == SheetType.INTERFACE
    assert detect_sheet_type("バッチ一覧", []) == SheetType.BATCH
    assert detect_sheet_type("コード管理", []) == SheetType.MASTER


def test_detect_by_header_keyword():
    matrix = [["帳票ID", "帳票名"], ["TB1", "月次"]]
    assert detect_sheet_type("補充シート", matrix) == SheetType.REPORT


def test_unknown_is_generic():
    assert detect_sheet_type("メモ", [["随筆"]]) == SheetType.GENERIC

• Step 2: 运行确认失败

Run: python -m pytest tests/test_sheet_detector.py -v Expected: FAIL（ImportError: cannot import name 'detect_sheet_type'）

• Step 3: 实现 sheet_detector.py

src/genesis/parsers/sheet_detector.py:

from __future__ import annotations

from typing import Any

from genesis.data_models import SheetType

# Sheet 名关键词（design §3.5.1，顺序即优先级）
NAME_KEYWORDS: list[tuple[str, SheetType]] = [
    ("機能", SheetType.FUNCTION),
    ("画面", SheetType.SCREEN),
    ("帳票", SheetType.REPORT),
    ("テーブル", SheetType.DATABASE),
    ("DB", SheetType.DATABASE),
    ("インターフェース", SheetType.INTERFACE),
    ("IF", SheetType.INTERFACE),
    ("バッチ", SheetType.BATCH),
    ("ジョブ", SheetType.BATCH),
    ("マスタ", SheetType.MASTER),
]

# 表头关键词
HEADER_KEYWORDS: list[tuple[str, SheetType]] = [
    ("機能ID", SheetType.FUNCTION),
    ("画面ID", SheetType.SCREEN),
    ("帳票ID", SheetType.REPORT),
    ("テーブルID", SheetType.DATABASE),
    ("IF名", SheetType.INTERFACE),
    ("バッチID", SheetType.BATCH),
]


def _name_hit(sheet_name: str) -> SheetType | None:
    for kw, st in NAME_KEYWORDS:
        if kw in sheet_name:
            return st
    return None


def _header_hit(matrix: list[list[Any]]) -> SheetType | None:
    for row in matrix[:3]:
        for cell in row:
            if isinstance(cell, str):
                for kw, st in HEADER_KEYWORDS:
                    if kw in cell:
                        return st
    return None


def detect_sheet_type(sheet_name: str, matrix: list[list[Any]]) -> SheetType:
    return _name_hit(sheet_name) or _header_hit(matrix) or SheetType.GENERIC

• Step 4: 运行确认通过

Run: python -m pytest tests/test_sheet_detector.py -v Expected: PASS（3 passed）

• Step 5: Commit

git add src/genesis/parsers/sheet_detector.py tests/test_sheet_detector.py
git commit -m "feat: SheetDetector 类型识别（名称/表头关键词）"

---

Task 3: Sheet 性质判定（表格型/自由记述型/混合型）

Files:

• Create: src/genesis/parsers/sheet_nature.py
• Create: tests/test_sheet_nature.py

Interfaces:

• Produces:
  • SheetNature(Enum)：TABLE / FREE_TEXT / MIXED
  • classify_sheet(matrix: list[list[Any]]) -> SheetNature
  • find_header_row(matrix: list[list[Any]]) -> int（首个连续非空 ≥2 的行的下标；无则 -1）

判定启发式（design §3.5.2 落地）：

• 空行占比 > 30%（非空行数 / 总行数 < 0.7）→ FREE_TEXT

• 仅 A 列使用（max_cols <= 1）→ FREE_TEXT

• 无表头行（find_header_row == -1）→ FREE_TEXT

• 表头行后存在以「・」/「■」开头的碎片行 → MIXED

• 其余 → TABLE

• Step 1: 写失败测试

tests/test_sheet_nature.py:

from genesis.parsers.sheet_nature import SheetNature, classify_sheet, find_header_row


def test_table_detection():
    m = [["機能ID", "機能名"], ["A001", "社員登録"], ["A002", "退職処理"]]
    assert classify_sheet(m) == SheetNature.TABLE


def test_free_text_single_col():
    m = [["新入社員を登録できる。"], ["氏名・所属・入社日を入力する。"]]
    assert classify_sheet(m) == SheetNature.FREE_TEXT


def test_free_text_many_empty_rows():
    m = [["要求A"], [], [], [], ["要求B"]]
    assert classify_sheet(m) == SheetNature.FREE_TEXT


def test_header_row_index():
    m = [["機能ID", "名前"], ["1", "田中"]]
    assert find_header_row(m) == 0
    assert find_header_row([["自由テキスト"]]) == -1

• Step 2: 运行确认失败

Run: python -m pytest tests/test_sheet_nature.py -v Expected: FAIL（导入错误）

• Step 3: 实现 sheet_nature.py

src/genesis/parsers/sheet_nature.py:

from __future__ import annotations

from enum import Enum
from typing import Any


class SheetNature(Enum):
    TABLE = "table"
    FREE_TEXT = "free_text"
    MIXED = "mixed"


def _non_empty(row: list[Any]) -> list[Any]:
    return [c for c in row if c is not None and str(c).strip() != ""]


def find_header_row(matrix: list[list[Any]]) -> int:
    for i, row in enumerate(matrix):
        if len(_non_empty(row)) >= 2:
            return i
    return -1


def _free_text_like(matrix: list[list[Any]]) -> bool:
    if not matrix:
        return True
    max_cols = max((len(row) for row in matrix), default=0)
    if max_cols <= 1:
        return True
    non_empty_rows = [r for r in matrix if _non_empty(r)]
    if len(non_empty_rows) / len(matrix) < 0.7:
        return True
    return find_header_row(matrix) == -1


def classify_sheet(matrix: list[list[Any]]) -> SheetNature:
    if _free_text_like(matrix):
        return SheetNature.FREE_TEXT
    header_row = find_header_row(matrix)
    if header_row >= 0:
        for row in matrix[header_row + 1:]:
            if any(str(c).strip().startswith(("・", "■")) for c in _non_empty(row)):
                return SheetNature.MIXED
    return SheetNature.TABLE

• Step 4: 运行确认通过

Run: python -m pytest tests/test_sheet_nature.py -v Expected: PASS（4 passed）

• Step 5: Commit

git add src/genesis/parsers/sheet_nature.py tests/test_sheet_nature.py
git commit -m "feat: Sheet 性质判定（table/free_text/mixed）"

---

Task 4: MergeHandler 与 TableExtractor

Files:

• Create: src/genesis/parsers/merge_fill.py
• Create: src/genesis/parsers/table_extractor.py
• Create: tests/test_table_extractor.py

Interfaces:

• Consumes: data_models（CellValue、ExcelTable、Provenance、SheetType）、build_source_uri
• Produces:
  • merge_fill.forward_fill(matrix, merged_ranges) -> list[list[Any]]
  • table_extractor.extract_table(sheet_name, matrix, file_name, detected_type) -> ExcelTable
  • table_extractor.column_letter(index) -> str（1→A、27→AA）

合并单元格展开：范围 (min_row, min_col, max_row, max_col)（1-based）；将范围内全部单元格填为主格（左上角）值。

• Step 1: 写失败测试

tests/test_table_extractor.py:

from genesis.data_models import CellValue, ExcelTable, SheetType
from genesis.parsers.merge_fill import forward_fill
from genesis.parsers.table_extractor import column_letter, extract_table


def test_forward_fill_vertical():
    matrix = [
        ["機能ID", "機能名", "備考"],
        ["A001", "", ""],
        ["", "", "メモ"],
    ]
    filled = forward_fill(matrix, [(1, 2, 2, 2)])  # B1:B2 纵向合并
    assert filled[1][1] == "機能名"
    assert filled[2][2] == "メモ"


def test_forward_fill_horizontal():
    matrix = [["A", "B"], ["x", ""]]
    filled = forward_fill(matrix, [(2, 1, 2, 2)])  # A2:B2 横向合并
    assert filled[1][1] == "x"
    assert filled[1][0] == "x"


def test_column_letter():
    assert column_letter(1) == "A"
    assert column_letter(27) == "AA"


def test_extract_table_basic():
    matrix = [["ID", "名前"], ["1", "田中"], ["2", "佐藤"]]
    table = extract_table("社員一覧", matrix, "f.xlsx", SheetType.FUNCTION)
    assert isinstance(table, ExcelTable)
    assert table.headers == ["ID", "名前"]
    assert table.extraction_method == "openpyxl"
    assert len(table.rows) == 2
    first = table.rows[0]
    assert isinstance(first["ID"], CellValue)
    assert first["ID"].value == "1"
    assert first["ID"].provenance.sheet_name == "社員一覧"
    assert first["ID"].provenance.column == "A"

• Step 2: 运行确认失败

Run: python -m pytest tests/test_table_extractor.py -v Expected: FAIL（导入错误）

• Step 3: 实现

src/genesis/parsers/merge_fill.py:

from __future__ import annotations

from typing import Any


def forward_fill(
    matrix: list[list[Any]],
    merged_ranges: list[tuple[int, int, int, int]],
) -> list[list[Any]]:
    """合并单元格：用左上角主格值填充范围内全部单元格。"""
    out = [list(row) for row in matrix]
    for (min_row, min_col, max_row, max_col) in merged_ranges:
        if not out or min_row > len(out) or min_col > len(out[min_row - 1]):
            continue
        main_value = out[min_row - 1][min_col - 1]
        for r in range(min_row, min(max_row, len(out)) + 1):
            row = out[r - 1]
            for c in range(min_col, min(max_col, len(row)) + 1):
                row[c - 1] = main_value
    return out

src/genesis/parsers/table_extractor.py:

from __future__ import annotations

from typing import Any

from genesis.data_models import CellValue, ExcelTable, Provenance, SheetType


def column_letter(index: int) -> str:
    """1 → A、27 → AA。"""
    letters = ""
    while index > 0:
        index, rem = divmod(index - 1, 26)
        letters = chr(65 + rem) + letters
    return letters


def extract_table(
    sheet_name: str,
    matrix: list[list[Any]],
    file_name: str,
    detected_type: SheetType,
    header_row: int = 0,
) -> ExcelTable:
    """从矩阵提取表格：首行视为表头，其后为数据行。"""
    if not matrix:
        return ExcelTable(
            name=sheet_name, detected_type=detected_type,
            extraction_method="openpyxl", headers=[], rows=[],
        )
    headers = [str(c) if c is not None else "" for c in matrix[header_row]]
    rows = []
    for r in range(header_row + 1, len(matrix)):
        row_dict = {}
        for c, h in enumerate(headers):
            raw = matrix[r][c] if c < len(matrix[r]) else None
            row_dict[h] = CellValue(
                value=raw,
                provenance=Provenance(
                    file_name=file_name,
                    sheet_name=sheet_name,
                    row=r - header_row,   # 数据行号从 1 开始
                    column=column_letter(c + 1),
                    column_header=h,
                ),
            )
        rows.append(row_dict)
    return ExcelTable(
        name=sheet_name, detected_type=detected_type,
        extraction_method="openpyxl", headers=headers, rows=rows,
    )

• Step 4: 运行确认通过

Run: python -m pytest tests/test_table_extractor.py -v Expected: PASS

• Step 5: Commit

git add src/genesis/parsers/merge_fill.py src/genesis/parsers/table_extractor.py tests/test_table_extractor.py
git commit -m "feat: MergeHandler + TableExtractor（forward_fill / 表格提取）"

---

Task 5: FormattingDetector（取消线/背景色/批注）

Files:

• Create: src/genesis/parsers/formatting_detector.py
• Create: tests/test_formatting_detector.py

Interfaces:

• Consumes: openpyxl Worksheet/Cell、data_models.CellFormatting / CellComment、build_source_uri

• Produces:

  • cell_formatting(cell) -> CellFormatting | None（strike / 字体色 / 背景色，纯默认则 None）
  • cell_comment(cell, file_name) -> CellComment | None
  • collect_comments(ws, file_name) -> list[CellComment]

• Step 1: 写失败测试

tests/test_formatting_detector.py:

from openpyxl import Workbook
from openpyxl.comments import Comment

from genesis.data_models import CellComment, CellFormatting
from genesis.parsers.formatting_detector import (
    cell_comment, cell_formatting, collect_comments,
)


def make_wb():
    wb = Workbook()
    ws = wb.active
    ws.title = "機能一覧"
    ws["A1"] = "F001"
    ws["A1"].font.strike = True
    ws["A2"] = "F002"
    ws["A2"].comment = Comment("要確認", "reviewer")
    return wb


def test_cell_formatting_detects_strike():
    fmt = cell_formatting(make_wb().active["A1"])
    assert fmt is not None
    assert fmt.strikethrough is True


def test_cell_formatting_none_when_plain():
    wb = Workbook()
    ws = wb.active
    ws["A1"] = "x"
    assert cell_formatting(ws["A1"]) is None


def test_cell_comment_returns_obj():
    wb = make_wb()
    cm = cell_comment(wb.active["A2"], "f.xlsx")
    assert isinstance(cm, CellComment)
    assert cm.author == "reviewer"
    assert cm.text == "要確認"
    assert cm.source_uri == "f.xlsx#機能一覧!A2"


def test_collect_comments():
    comments = collect_comments(make_wb().active, "f.xlsx")
    assert len(comments) == 1

• Step 2: 运行确认失败

Run: python -m pytest tests/test_formatting_detector.py -v Expected: FAIL（导入错误）

• Step 3: 实现

src/genesis/parsers/formatting_detector.py:

from __future__ import annotations

from genesis.data_models import CellComment, CellFormatting
from genesis.parsers.provenance import build_source_uri


def cell_formatting(cell) -> CellFormatting | None:
    strike = bool(cell.font.strike)
    font_color = None
    if cell.font.color and str(cell.font.color.rgb) not in ("00000000", "FF000000"):
        font_color = str(cell.font.color.rgb)
    bg_color = None
    fill = cell.fill
    if fill and fill.fgColor and str(fill.fgColor.rgb) not in ("00000000", "FF000000"):
        bg_color = str(fill.fgColor.rgb)
    if strike or font_color or bg_color:
        return CellFormatting(
            strikethrough=strike, font_color=font_color, bg_color=bg_color,
        )
    return None


def cell_comment(cell, file_name: str) -> CellComment | None:
    if cell.comment is None:
        return None
    return CellComment(
        author=cell.comment.author or "",
        text=cell.comment.text or "",
        source_uri=build_source_uri(file_name, cell.parent.title, cell.coordinate),
    )


def collect_comments(ws, file_name: str) -> list[CellComment]:
    result = []
    for row in ws.iter_rows():
        for cell in row:
            cm = cell_comment(cell, file_name)
            if cm is not None:
                result.append(cm)
    return result

• Step 4: 运行确认通过

Run: python -m pytest tests/test_formatting_detector.py -v Expected: PASS（4 passed）

• Step 5: Commit

git add src/genesis/parsers/formatting_detector.py tests/test_formatting_detector.py
git commit -m "feat: FormattingDetector（取消线/背景色/批注）"

---

Task 6: FreeTextExtractor（自由记述型 Sheet 结构占位）

Files:

• Create: src/genesis/parsers/free_text_extractor.py
• Create: tests/test_free_text_extractor.py

Interfaces:

• Consumes: data_models（CellValue、ExcelTable、Provenance、SheetType）
• Produces:
  • extract_text_blocks(matrix) -> list[str]（全空行分段，行内单元格以空格连接）
  • build_free_text_table(sheet_name, blocks, file_name, detected_type=SheetType.GENERIC) -> ExcelTable（extraction_method="llm_from_free_text"，headers=["text"]）

LLM：本里程碑不调用 LLM（InferenceEngine 未实现）；占位表为后续 LLM 结构化保留入口。

• Step 1: 写失败测试

tests/test_free_text_extractor.py:

from genesis.data_models import CellValue, ExcelTable, SheetType
from genesis.parsers.free_text_extractor import build_free_text_table, extract_text_blocks


def test_extract_blocks_splits_on_empty_rows():
    m = [["新入社員を登録。"], [], ["テスト要件：入力。"], [], []]
    assert extract_text_blocks(m) == ["新入社員を登録。", "テスト要件：入力。"]


def test_build_free_text_table():
    table = build_free_text_table("機能要件", ["A", "B"], "f.xlsx", SheetType.FUNCTION)
    assert isinstance(table, ExcelTable)
    assert table.detected_type == SheetType.FUNCTION
    assert table.extraction_method == "llm_from_free_text"
    assert len(table.rows) == 2
    assert isinstance(table.rows[0]["text"], CellValue)
    assert table.rows[0]["text"].value == "A"

• Step 2: 运行确认失败

Run: python -m pytest tests/test_free_text_extractor.py -v Expected: FAIL（导入错误）

• Step 3: 实现

src/genesis/parsers/free_text_extractor.py:

from __future__ import annotations

from typing import Any

from genesis.data_models import CellValue, ExcelTable, Provenance, SheetType


def extract_text_blocks(matrix: list[list[Any]]) -> list[str]:
    """按全空行分段；行内非空单元格以「 」连接。"""
    blocks: list[str] = []
    current: list[str] = []
    for row in matrix:
        cells = [str(c).strip() for c in row if c is not None and str(c).strip() != ""]
        if not cells:
            if current:
                blocks.append(" ".join(current))
                current = []
            continue
        current.append(" ".join(cells))
    if current:
        blocks.append(" ".join(current))
    return blocks


def build_free_text_table(
    sheet_name: str,
    blocks: list[str],
    file_name: str,
    detected_type: SheetType = SheetType.GENERIC,
) -> ExcelTable:
    rows = []
    for i, text in enumerate(blocks, start=1):
        rows.append({
            "text": CellValue(
                value=text,
                provenance=Provenance(
                    file_name=file_name,
                    sheet_name=sheet_name,
                    row=i,
                    column="A",
                    column_header="text",
                ),
            ),
        })
    return ExcelTable(
        name=sheet_name,
        detected_type=detected_type,
        extraction_method="llm_from_free_text",
        headers=["text"],
        rows=rows,
    )

• Step 4: 运行确认通过

Run: python -m pytest tests/test_free_text_extractor.py -v Expected: PASS（2 passed）

• Step 5: Commit

git add src/genesis/parsers/free_text_extractor.py tests/test_free_text_extractor.py
git commit -m "feat: FreeTextExtractor（文本分段 + 占位结构化表）"

---

Task 7: ExcelParser 编排器

Files:

• Create: src/genesis/parsers/excel_parser.py
• Create: tests/test_excel_parser.py

Interfaces:

• Produces:
  • @dataclass ExcelParseResult：file_name: str、tables: list[ExcelTable]、comments: list[CellComment]、skipped: list[str]
  • class ExcelParser: parse(path) -> ExcelParseResult

parse 编排：

1. open_workbook(path)
2. 每 Sheet：matrix = sheet_matrix(ws)；空 → skipped
3. detected_type = detect_sheet_type(ws.title, matrix)
4. nature = classify_sheet(matrix)
5. FREE_TEXT → extract_text_blocks + build_free_text_table(detected_type=detected_type)
6. 否则 forward_fill(matrix, merged) + extract_table
7. collect_comments(ws, file_name) 汇总
8. 返回 ExcelParseResult

• Step 1: 写失败测试

tests/test_excel_parser.py:

from genesis.data_models import SheetType
from genesis.parsers.excel_parser import ExcelParseResult, ExcelParser

from tests.excel_helpers import new_workbook, save_workbook


def test_parse_table_sheets(tmp_path):
    wb = new_workbook({
        "機能一覧": [["機能ID", "機能名"], ["A001", "社員登録"]],
        "バッチ一覧": [["バッチID", "処理名"], ["B1", "夜間集計"]],
    })
    path = save_workbook(tmp_path, wb)
    result = ExcelParser().parse(path)
    assert isinstance(result, ExcelParseResult)
    by_name = {t.name: t for t in result.tables}
    assert by_name["機能一覧"].detected_type == SheetType.FUNCTION
    assert len(by_name["機能一覧"].rows) == 1
    assert by_name["バッチ一覧"].detected_type == SheetType.BATCH


def test_parse_free_text_sheet(tmp_path):
    wb = new_workbook({"メモ": [["新入社員を登録"], [], [], [], []]})
    path = save_workbook(tmp_path, wb)
    result = ExcelParser().parse(path)
    assert len(result.tables) == 1
    assert result.tables[0].extraction_method == "llm_from_free_text"

• Step 2: 运行确认失败

Run: python -m pytest tests/test_excel_parser.py -v Expected: FAIL（导入错误）

• Step 3: 实现 excel_parser.py

src/genesis/parsers/excel_parser.py:

from __future__ import annotations

from dataclasses import dataclass, field
from pathlib import Path

from genesis.data_models import CellComment, ExcelTable
from genesis.parsers.excel_reader import open_workbook, sheet_matrix
from genesis.parsers.sheet_detector import detect_sheet_type
from genesis.parsers.sheet_nature import SheetNature, classify_sheet
from genesis.parsers.merge_fill import forward_fill
from genesis.parsers.table_extractor import extract_table
from genesis.parsers.formatting_detector import collect_comments
from genesis.parsers.free_text_extractor import build_free_text_table, extract_text_blocks


@dataclass
class ExcelParseResult:
    file_name: str
    tables: list[ExcelTable] = field(default_factory=list)
    comments: list[CellComment] = field(default_factory=list)
    skipped: list[str] = field(default_factory=list)


class ExcelParser:
    """要件定义 Excel 解析入口。"""

    def parse(self, path: str | Path) -> ExcelParseResult:
        wb = open_workbook(path)
        file_name = Path(path).name
        result = ExcelParseResult(file_name=file_name)
        for ws in wb.worksheets:
            matrix = sheet_matrix(ws)
            if not matrix:
                result.skipped.append(ws.title)
                continue
            detected_type = detect_sheet_type(ws.title, matrix)
            nature = classify_sheet(matrix)
            if nature == SheetNature.FREE_TEXT:
                blocks = extract_text_blocks(matrix)
                result.tables.append(
                    build_free_text_table(ws.title, blocks, file_name, detected_type)
                )
            else:
                merged = [
                    (r.min_row, r.min_col, r.max_row, r.max_col)
                    for r in ws.merged_cells.ranges
                ]
                filled = forward_fill(matrix, merged) if merged else matrix
                result.tables.append(extract_table(ws.title, filled, file_name, detected_type))
            result.comments.extend(collect_comments(ws, file_name))
        return result

• Step 4: 运行确认通过

Run: python -m pytest tests/test_excel_parser.py -v Expected: PASS（2 passed）

• Step 5: Commit

git add src/genesis/parsers/excel_parser.py tests/test_excel_parser.py
git commit -m "feat: ExcelParser 编排器（类型/性质/提取/汇总）"

---

Task 8: 真实样本集成测试

Files:

• Create: tests/test_real_samples.py

说明: 使用 samples/ 下 3 个脱敏要件定义样本做端到端验证。样本缺失时对应用例 pytest.skip。

• Step 1: 写集成测试

tests/test_real_samples.py:

from pathlib import Path

import pytest

from genesis.parsers.excel_parser import ExcelParser

SAMPLES = Path(__file__).resolve().parents[1] / "samples"


def _x(name: str) -> Path:
    return SAMPLES / name


def test_new_dev_sample_has_tables():
    p = _x("要件定義_新規開発.xlsx")
    if not p.exists():
        pytest.skip("样本缺失")
    result = ExcelParser().parse(p)
    by_name = {t.name: t for t in result.tables}
    assert "機能一覧" in by_name
    assert by_name["機能一覧"].rows
    assert any(t.name == "DB定義" for t in result.tables)


def test_additional_modification_has_tables():
    p = _x("要件定義_追加改修.xlsx")
    if not p.exists():
        pytest.skip("样本缺失")
    result = ExcelParser().parse(p)
    assert result.tables
    assert any(t.rows for t in result.tables)


def test_free_text_sample_detected():
    p = _x("要件定義_自由記述.xlsx")
    if not p.exists():
        pytest.skip("样本缺失")
    result = ExcelParser().parse(p)
    assert any(t.extraction_method == "llm_from_free_text" for t in result.tables)

• Step 2: 运行集成测试

Run: python -m pytest tests/test_real_samples.py -v Expected: PASS（样本存在时）或 至少 2 passed 1 skipped

• Step 3: 全量回归

Run: python -m pytest -v Expected: PASS（全体收集成功）

• Step 4: Commit

git add tests/test_real_samples.py
git commit -m "test: 真实样本集成测试（3 类型 xlsx）"

---

自审结论（计划完稿时执行）

• Spec 覆盖：implementation-plan §2.1→Task1、§2.2→Task2、§2.3→Task3、§2.4+§2.5→Task4、§2.7→Task5、§2.6→Task6、§2.9→Task7+Task8；design §3.5.1→T2、§3.5.2→T3、§3.5.3→T4、§3.5.4/§3.5.5→T5；sample-spec §3.1-3.4→T8；验收标准（表格型/自由记述型/混合型解析、合并填充、source_uri）逐项覆盖。
• 占位符扫描：无 TBD/TODO；任务指引 Task1 的 Step 编号已显式标注（Step 3 即实现、Step 4 运行），不含空白步骤。
• 类型一致性：CellValue / ExcelTable / SheetType / CellFormatting / CellComment 与 data_models.py 完全一致；build_source_uri(file, sheet, ref) 全局同名；extract_table.column 用 column_letter 生成。
• 依赖边界：全程不依赖 InferenceEngine；FreeText 表 extraction_method="llm_from_free_text" 显式标记后续替换，符合模块边界与 YAGNI。