sheet-edit 多级表头审计修复(上轮讨论定稿的「立即做」批次):① locate.detect_multi_level 识别「分组行+重复子列名行」相邻形态,inspect 每 sheet 返回 multi_level_header——AI 第一轮即可判型,不再烧两轮试错;build_col_map 重复列名报错、_resolve_change 命中分组行时报错,文案一律直指 change.cells 出口(审计指正:这类表任何 header_row 都无解,此前建议修正 suggest 行号其实救不了,改口径为「检测并拒得明白」)。② apply 新增 cells 地址模式({sheet, cells:{D4:{old,new}}},与 match+set 互斥,允许可选 $)——多级表头从「不支持」降级为「坐标多确认一步」,old 校验/合并区防护/全簿逐格自检/LO 重算与列名模式完全同源,真正封死「terminal 裸改交付物」的危险兜底动机。③ header_row 支持 change 级覆盖(多 sheet 表头位置不一的隐患提前拆掉)。④ SKILL.md:工作流程/返回字段/cells 小节/硬约束加「禁止绕开技能手改」/边界与常见错误同步;清理 MergedCell 死 import。事故同款结构(大标题+合并分组+重复子列)16 项容器实测全过,含 old 校验拒绝无产物、非法地址、per-change 混单、插删链与 CSV 防回归
This commit is contained in:
@@ -27,7 +27,7 @@ description: 结构化修改表格文件。内容定位(列名 + 值),最
|
||||
|
||||
## 工作流程
|
||||
|
||||
1. `action=inspect` 查看表结构与公式(注意采用返回里的 `suggested_header_row`)
|
||||
1. `action=inspect` 查看表结构与公式(注意采用返回里的 `suggested_header_row`;若 `multi_level_header` 非 null,列名模式不可用,直接走 `cells` 地址模式)
|
||||
2. 按需求选择 `action=apply` / `insert_rows` / `delete_rows` / `insert_cols` / `delete_cols`
|
||||
|
||||
## inspect
|
||||
@@ -56,6 +56,7 @@ description: 结构化修改表格文件。内容定位(列名 + 值),最
|
||||
"cols": 7,
|
||||
"header_row": 1,
|
||||
"suggested_header_row": 1,
|
||||
"multi_level_header": null,
|
||||
"headers": ["序号", "项目", "单位", "数量", "单价", "金额", "备注"],
|
||||
"preview": [
|
||||
[1, "管道", "米", 100, 18, {"formula": "=D3*E3"}, ""],
|
||||
@@ -74,6 +75,7 @@ description: 结构化修改表格文件。内容定位(列名 + 值),最
|
||||
要点:
|
||||
- `headers` 是表头行;合并单元格只有首个有值,其余为 `null`
|
||||
- `suggested_header_row` 是自动检测的表头行(中文表格常见「第 1 行合并标题 + 第 2 行真表头」);与你的 `header_row` 不一致时直接采用它继续后续调用,不用自己推断
|
||||
- `multi_level_header` 非 null(如 `{"group_row": 2, "column_row": 3}`)= 多级表头(分组行 + 重复子列名行):列名定位必死,改用 `cells` 地址模式,别反复换 `header_row` 重试
|
||||
- 公式以 `{"formula": "=..."}` 表示
|
||||
- `preview` 可能被 `limit` 截断;`formulas` 最多列 500 条(超出时 `formulas_truncated: true`)
|
||||
- `merged` 是合并区域列表(如 `["A1:G1"]`)
|
||||
@@ -97,10 +99,28 @@ description: 结构化修改表格文件。内容定位(列名 + 值),最
|
||||
|
||||
参数:
|
||||
- `file`(必填)
|
||||
- `changes`(必填):改动清单
|
||||
- `header_row`(可选,默认 1)
|
||||
- `changes`(必填):改动清单;每项可用 `match+set`(列名定位)或 `cells`(地址定位),二选一
|
||||
- `header_row`(可选,默认 1);各 change 可单独带 `header_row` 字段覆盖(多 sheet 表头位置不同时用)
|
||||
- `output`(可选):输出文件名(放原文件同目录;平台预处理成会话目录下的绝对路径也接受),默认原文件名加「(已改)」
|
||||
|
||||
### cells 地址定位(多级表头 / 分组汇总表的正经出口)
|
||||
|
||||
```json
|
||||
{
|
||||
"action": "apply",
|
||||
"file": "预算表.xlsx",
|
||||
"changes": [
|
||||
{"sheet": "总表", "cells": {
|
||||
"D4": {"old": 157, "new": 190},
|
||||
"F4": {"old": 120, "new": 150}
|
||||
}}
|
||||
]
|
||||
}
|
||||
```
|
||||
|
||||
- 地址从 inspect 的 `preview`(含列坐标提示)/`formulas` 里数出来;允许可选 `$`(`$D$4` 同 `D4`)
|
||||
- old 校验、合并区防护、全簿逐格自检、LO 重算与列名模式**完全同源**,不是降级通道
|
||||
|
||||
### 定位
|
||||
|
||||
`match` 用列名 + 值定位目标行,可多列组合:
|
||||
@@ -204,6 +224,7 @@ description: 结构化修改表格文件。内容定位(列名 + 值),最
|
||||
- 改动后自检:目标改动生效、其他内容未变(含公式原文)、文件可打开
|
||||
- 任一项失败 → 产物自动删除、不交付
|
||||
- 遇不支持的引用形态(数组公式、外部簿引用、结构化引用)直接报错,绝不半改交付
|
||||
- 超出能力的表(多级表头等)用 `cells` 地址模式处理;**禁止绕开本技能用 terminal/脚本手改交付物**——手改没有 old 校验、逐格自检、跨表引用检查与重算门禁,错了没人兜底
|
||||
|
||||
## 边界
|
||||
|
||||
@@ -214,7 +235,7 @@ description: 结构化修改表格文件。内容定位(列名 + 值),最
|
||||
重算失败时自动回退交付「未重算版」并在结果 `recalc` 字段标注 `failed:...`(未重算版仍带「打开时全量重算」标记,
|
||||
Excel/WPS 打开结果一样正确);结果里看到 recalc=failed 时简短告知用户一句即可,不要重试。
|
||||
- CSV 单元格一律按文本保留(「007」「1.10」不会变成 7/1.1);数字与数字文本在定位比较时互认。
|
||||
- 多级表头、复杂合并:可能定位失败,需手工处理。
|
||||
- 多级表头(分组行 + 重复子列名,如「新建顶管|数量·单价」):`match/set` 列名定位不支持(子列名不唯一,任何 `header_row` 都无解);`inspect` 会标 `multi_level_header`,请用 `cells` 地址模式 + 从 preview 确认坐标。更深层级(≥3 行表头)同样走 `cells`。
|
||||
|
||||
## 常见错误
|
||||
|
||||
@@ -225,3 +246,4 @@ description: 结构化修改表格文件。内容定位(列名 + 值),最
|
||||
5. **sheet 不存在**:报错会列出可用 sheet 名
|
||||
6. **写入合并区域非锚点格**:报错提示改用区域左上角锚点
|
||||
7. **报「不支持数组公式/外部簿引用」**:该表需人工处理,不要反复重试
|
||||
8. **报「重复列名 / 检测到多级表头」**:换 `header_row` 重试没有用(子列名天然不唯一),改用 `cells` 地址模式;也不要因此绕开技能去手改文件
|
||||
|
||||
@@ -32,6 +32,35 @@ def detect_header_row(ws, scan=10, col_cap=200) -> int:
|
||||
return best_row
|
||||
|
||||
|
||||
def detect_multi_level(ws, scan=10, col_cap=200):
|
||||
"""检测“分组行 + 重复子列名行”相邻出现的多级表头形态,返回 (分组行, 列名行) 或 None。
|
||||
|
||||
判定:相邻两行均全为字符串;上行值唯一且 ≥2 个(分组标题),下行有重复且 ≥3 个、格数多于上行
|
||||
(真列名在多组下重复)。只读不写,纯诊断用。"""
|
||||
max_c = min(ws.max_column, col_cap)
|
||||
prev = None # (row, 全字符串, 值列表)
|
||||
for r in range(1, min(scan, ws.max_row) + 1):
|
||||
vals = []
|
||||
ok = True
|
||||
for c in range(1, max_c + 1):
|
||||
v = ws.cell(r, c).value
|
||||
if v is None or v == '':
|
||||
continue
|
||||
if isinstance(v, str) and not v.startswith('='):
|
||||
vals.append(v.strip())
|
||||
else:
|
||||
ok = False
|
||||
break
|
||||
if prev is not None and ok and prev[1] and prev[2]:
|
||||
pu, cu = prev[2], vals
|
||||
if (len(pu) >= 2 and len(set(pu)) == len(pu)
|
||||
and len(cu) >= 3 and len(set(cu)) != len(cu)
|
||||
and len(cu) > len(pu)):
|
||||
return (prev[0], r)
|
||||
prev = (r, ok, vals)
|
||||
return None
|
||||
|
||||
|
||||
def value_eq(a, b) -> bool:
|
||||
if a is None and b is None:
|
||||
return True
|
||||
@@ -57,6 +86,13 @@ def build_col_map(ws, header_row: int) -> dict:
|
||||
if not name:
|
||||
continue
|
||||
if name in col_map:
|
||||
ml = detect_multi_level(ws)
|
||||
if ml:
|
||||
raise UserError(
|
||||
f"表头行 {header_row} 存在重复列名('{name}'),且检测到多级表头"
|
||||
f"(第 {ml[0]} 行分组、第 {ml[1]} 行子列名):列名模式无法唯一定位,"
|
||||
"请从 inspect 的 preview/formulas 确认目标坐标后改用 change.cells 地址模式"
|
||||
)
|
||||
raise UserError(f"表头行 {header_row} 存在重复列名: '{name}'")
|
||||
col_map[name] = c
|
||||
if not col_map:
|
||||
|
||||
@@ -1,12 +1,12 @@
|
||||
"""sheet-edit 的各个 op 实现。"""
|
||||
import datetime
|
||||
import re
|
||||
import shutil
|
||||
import tempfile
|
||||
from pathlib import Path
|
||||
|
||||
import openpyxl
|
||||
from openpyxl.utils import get_column_letter
|
||||
from openpyxl.cell.cell import MergedCell
|
||||
from openpyxl.utils import column_index_from_string, get_column_letter
|
||||
from openpyxl.worksheet.cell_range import CellRange, MultiCellRange
|
||||
from openpyxl.worksheet.formula import ArrayFormula
|
||||
|
||||
@@ -74,6 +74,7 @@ def _sheet_info(ws, name, header_row, limit):
|
||||
'cols': ws.max_column,
|
||||
'header_row': header_row,
|
||||
'suggested_header_row': locate.detect_header_row(ws),
|
||||
'multi_level_header': None,
|
||||
'headers': [
|
||||
_json_safe(ws.cell(header_row, c).value)
|
||||
for c in range(1, ws.max_column + 1)
|
||||
@@ -84,6 +85,9 @@ def _sheet_info(ws, name, header_row, limit):
|
||||
'merged': [str(r) for r in ws.merged_cells.ranges],
|
||||
'truncated': False,
|
||||
}
|
||||
ml = locate.detect_multi_level(ws)
|
||||
if ml:
|
||||
info['multi_level_header'] = {'group_row': ml[0], 'column_row': ml[1]}
|
||||
for row in ws.iter_rows():
|
||||
for cell in row:
|
||||
v = cell.value
|
||||
@@ -173,44 +177,79 @@ def _resolve_change(wb, ch, header_row):
|
||||
raise UserError(
|
||||
f"sheet 不存在: {sheet}(可用: {', '.join(wb.sheetnames)})"
|
||||
)
|
||||
match = ch.get('match') or {}
|
||||
set_spec = ch.get('set') or {}
|
||||
if not isinstance(set_spec, dict) or not set_spec:
|
||||
raise UserError("change.set 必须是非空对象")
|
||||
|
||||
hr = _int(ch, 'header_row', header_row) # change 级可覆盖全局 header_row
|
||||
ws = wb[sheet]
|
||||
col_map = locate.build_col_map(ws, header_row)
|
||||
row = locate.find_row(ws, col_map, match, header_row)
|
||||
|
||||
set_spec = ch.get('set')
|
||||
cells_spec = ch.get('cells')
|
||||
if set_spec is not None and cells_spec is not None:
|
||||
raise UserError("change 只能二选一:match+set(列名定位)或 cells(地址定位)")
|
||||
if cells_spec is not None:
|
||||
return _resolve_cells(ws, sheet, cells_spec)
|
||||
|
||||
match = ch.get('match') or {}
|
||||
if not isinstance(set_spec, dict) or not set_spec:
|
||||
raise UserError("change.set 必须是非空对象(多级表头等无法按列名定位的表改用 change.cells)")
|
||||
|
||||
ml = locate.detect_multi_level(ws)
|
||||
if ml and hr == ml[0]:
|
||||
raise UserError(
|
||||
f"检测到多级表头(第 {ml[0]} 行分组、第 {ml[1]} 行子列名):列名模式无法唯一定位,"
|
||||
"请从 inspect 的 preview/formulas 确认目标坐标后改用 change.cells 地址模式"
|
||||
)
|
||||
col_map = locate.build_col_map(ws, hr)
|
||||
row = locate.find_row(ws, col_map, match, hr)
|
||||
|
||||
items = []
|
||||
for col_name, spec in set_spec.items():
|
||||
if not isinstance(spec, dict) or 'old' not in spec or 'new' not in spec:
|
||||
raise UserError(
|
||||
f"set['{col_name}'] 必须是 {{\"old\": ..., \"new\": ...}} 格式"
|
||||
)
|
||||
col = locate.find_col(col_map, col_name)
|
||||
actual = ws.cell(row, col).value
|
||||
if isinstance(actual, ArrayFormula):
|
||||
raise UserError(
|
||||
f"{sheet}!{get_column_letter(col)}{row} 是数组公式,"
|
||||
"本工具不支持修改(请在 Excel/WPS 手工处理)"
|
||||
)
|
||||
new = spec['new']
|
||||
if isinstance(new, (dict, list)):
|
||||
raise UserError(f"set['{col_name}'].new 只能是数字/字符串/布尔/null,收到 {new!r}")
|
||||
items.append({
|
||||
'sheet': sheet,
|
||||
'row': row,
|
||||
'col': col,
|
||||
'col_name': col_name,
|
||||
'match': dict(match),
|
||||
'expected_old': spec['old'],
|
||||
'actual_old': actual,
|
||||
'new': new,
|
||||
})
|
||||
items.append(_mk_item(ws, sheet, row, col, col_name, spec, dict(match)))
|
||||
return items
|
||||
|
||||
|
||||
_ADDR_RE = re.compile(r'^\$?([A-Za-z]{1,3})\$?(\d{1,7})$')
|
||||
|
||||
|
||||
def _resolve_cells(ws, sheet, cells_spec):
|
||||
"""地址模式:cells = {"C5": {"old": ..., "new": ...}, ...}。多级表头等列名不可用场景的正经出口,
|
||||
old 校验/合并区防护/全簿自检/重算与列名模式完全同源。"""
|
||||
if not isinstance(cells_spec, dict) or not cells_spec:
|
||||
raise UserError("change.cells 必须是非空对象:地址 → {\"old\": ..., \"new\": ...}")
|
||||
items = []
|
||||
for addr, spec in cells_spec.items():
|
||||
m = _ADDR_RE.match(str(addr).strip())
|
||||
if not m:
|
||||
raise UserError(f"非法单元格地址: {addr!r}(形如 C5,允许可选 $ 绝对符)")
|
||||
col = column_index_from_string(m.group(1).upper())
|
||||
row = int(m.group(2))
|
||||
items.append(_mk_item(ws, sheet, row, col, str(addr).strip().upper(), spec, {}))
|
||||
return items
|
||||
|
||||
|
||||
def _mk_item(ws, sheet, row, col, label, spec, match_info):
|
||||
if not isinstance(spec, dict) or 'old' not in spec or 'new' not in spec:
|
||||
raise UserError(f"'{label}' 的设值必须是 {{\"old\": ..., \"new\": ...}} 格式")
|
||||
actual = ws.cell(row, col).value
|
||||
if isinstance(actual, ArrayFormula):
|
||||
raise UserError(
|
||||
f"{sheet}!{get_column_letter(col)}{row} 是数组公式,"
|
||||
"本工具不支持修改(请在 Excel/WPS 手工处理)"
|
||||
)
|
||||
new = spec['new']
|
||||
if isinstance(new, (dict, list)):
|
||||
raise UserError(f"'{label}'.new 只能是数字/字符串/布尔/null,收到 {new!r}")
|
||||
return {
|
||||
'sheet': sheet,
|
||||
'row': row,
|
||||
'col': col,
|
||||
'col_name': label,
|
||||
'match': match_info,
|
||||
'expected_old': spec['old'],
|
||||
'actual_old': actual,
|
||||
'new': new,
|
||||
}
|
||||
|
||||
|
||||
def _verify_apply(out_path, before, plan):
|
||||
with tempfile.TemporaryDirectory() as td:
|
||||
xlsx = convert.to_workbook(out_path, Path(td))
|
||||
|
||||
Reference in New Issue
Block a user