@amaster.ai/pi-lark 0.1.7 → 0.1.8

package/skills/lark-sheets/scripts/lark_profile_table.py

@@ -0,0 +1,614 @@
+#!/usr/bin/env python3
+# Copyright (c) 2026 Lark Technologies Pte. Ltd.
+# SPDX-License-Identifier: MIT
+"""Profile a candidate table range in a Lark sheet."""
+
+from __future__ import annotations
+
+import argparse
+import re
+from typing import Any
+
+from lark_sheet_range import col_to_index, format_range, index_to_col, parse_range
+from lark_sheet_read_cli import (
+    LarkCliError,
+    add_spreadsheet_args,
+    emit_error,
+    emit_success,
+    envelope_data,
+    run_sheets,
+)
+from lark_detect_subtables import CsvGrid, parse_annotated_csv
+
+ACTION = "profile_table"
+ERROR_VALUES = ("#VALUE!", "#DIV/0!", "#REF!", "#NAME?", "#NULL!", "#NUM!", "#N/A")
+TOTAL_KEYWORDS = ("合计", "总计", "小计", "汇总", "累计")
+TOTAL_EN_RE = re.compile(r"\b(?:grand total|subtotal|total)\b", re.IGNORECASE)
+SIGNATURE_KEYWORDS = ("编制人", "审核人", "审批人", "负责人", "经理", "签名")
+
+
+def _is_number(value: str) -> bool:
+    text = value.strip().replace(",", "")
+    if text.endswith("%"):
+        text = text[:-1]
+    try:
+        float(text)
+        return True
+    except ValueError:
+        return False
+
+
+def _is_date_like(value: str) -> bool:
+    text = value.strip()
+    if re.match(r"^\d{4}[-/]\d{1,2}[-/]\d{1,2}(?:\s+\d{1,2}:\d{2}(:\d{2})?)?$", text):
+        return True
+    if re.match(r"^\d{8}$", text):
+        year = int(text[:4])
+        month = int(text[4:6])
+        day = int(text[6:8])
+        return 1900 <= year <= 2100 and 1 <= month <= 12 and 1 <= day <= 31
+    if re.match(r"^\d{6}$", text):
+        year = int(text[:4])
+        month = int(text[4:6])
+        return 1900 <= year <= 2100 and 1 <= month <= 12
+    return False
+
+
+def _classify_value(value: str) -> str:
+    text = value.strip()
+    if not text:
+        return "empty"
+    if any(err in text for err in ERROR_VALUES):
+        return "error"
+    if re.match(r"^0\d+$", text) or re.match(r"^\d{11,}$", text):
+        return "string_id"
+    if _is_date_like(text):
+        return "date"
+    if _is_number(text):
+        return "number"
+    if text.lower() in {"true", "false", "yes", "no"}:
+        return "bool"
+    return "string"
+
+
+def _row_values(grid: CsvGrid, row_index: int) -> list[str]:
+    return grid.values[row_index]
+
+
+def _header_score(row: list[str]) -> tuple[int, int]:
+    non_empty = [value for value in row if value.strip()]
+    if not non_empty:
+        return 0, 0
+    strings = sum(1 for value in non_empty if _classify_value(value) in {"string", "string_id"})
+    data_like = sum(
+        1 for value in non_empty if _classify_value(value) in {"number", "date", "bool"}
+    )
+    return strings * 2 + len(non_empty) - data_like * 3, len(non_empty)
+
+
+def _header_candidate_score(grid: CsvGrid, index: int) -> int:
+    score, _ = _header_score(grid.values[index])
+    following = grid.values[index + 1 : index + 4]
+    following_non_empty = [value for row in following for value in row if value.strip()]
+    following_data_like = sum(
+        1 for value in following_non_empty if _classify_value(value) in {"number", "date", "bool"}
+    )
+    # A schema row is usually followed by data-shaped values; use this as a
+    # small tie-breaker while preferring earlier candidates with equal evidence.
+    score += min(3, following_data_like)
+    return score - index // 2
+
+
+def detect_header_row(grid: CsvGrid, scan_rows: int = 20) -> int | None:
+    best_row = None
+    best_score = None
+    for idx, row in enumerate(grid.values[:scan_rows]):
+        _, non_empty_count = _header_score(row)
+        if not non_empty_count:
+            continue
+        score = _header_candidate_score(grid, idx)
+        if best_score is None or score > best_score:
+            best_score = score
+            best_row = grid.row_numbers[idx]
+    return best_row
+
+
+def _possible_multi_row_header(grid: CsvGrid, header_row: int | None) -> bool:
+    if header_row is None or header_row not in grid.row_numbers:
+        return False
+    header_index = grid.row_numbers.index(header_row)
+    if header_index + 1 >= len(grid.values):
+        return False
+    header_score, header_non_empty = _header_score(grid.values[header_index])
+    next_score, next_non_empty = _header_score(grid.values[header_index + 1])
+    if header_non_empty < 2 or next_non_empty < 2:
+        return False
+    return next_score >= max(4, (header_score * 3 + 3) // 4)
+
+
+def _header_row_not_first(grid: CsvGrid, header_row: int | None) -> bool:
+    if header_row is None:
+        return False
+    first_non_empty = next(
+        (
+            row_num
+            for row_num, row in zip(grid.row_numbers, grid.values)
+            if any(value.strip() for value in row)
+        ),
+        None,
+    )
+    return first_non_empty is not None and first_non_empty != header_row
+
+
+def detect_special_rows(grid: CsvGrid, header_row: int | None) -> list[dict[str, Any]]:
+    specials = []
+    for idx, row in enumerate(grid.values):
+        row_num = grid.row_numbers[idx]
+        if header_row is not None and row_num <= header_row:
+            continue
+        joined = " ".join(value.strip() for value in row if value.strip())
+        if not joined:
+            specials.append({"row": row_num, "reason": "empty_row", "sample": ""})
+            continue
+        if any(keyword in joined for keyword in TOTAL_KEYWORDS) or TOTAL_EN_RE.search(joined):
+            specials.append({"row": row_num, "reason": "total_row", "sample": joined[:120]})
+        elif any(keyword in joined for keyword in SIGNATURE_KEYWORDS):
+            specials.append({"row": row_num, "reason": "signature_row", "sample": joined[:120]})
+    return specials
+
+
+def _last_data_row(grid: CsvGrid, header_row: int | None, specials: list[dict[str, Any]]) -> int | None:
+    special_rows = {item["row"] for item in specials if item["reason"] != "empty_row"}
+    for idx in range(len(grid.row_numbers) - 1, -1, -1):
+        row_num = grid.row_numbers[idx]
+        if header_row is not None and row_num <= header_row:
+            continue
+        if row_num in special_rows:
+            continue
+        if any(value.strip() for value in grid.values[idx]):
+            return row_num
+    return None
+
+
+def _column_values(grid: CsvGrid, col_index: int, start_row: int, end_row: int | None) -> list[str]:
+    values = []
+    for idx, row_num in enumerate(grid.row_numbers):
+        if row_num < start_row:
+            continue
+        if end_row is not None and row_num > end_row:
+            continue
+        row = grid.values[idx]
+        values.append(row[col_index] if col_index < len(row) else "")
+    return values
+
+
+def _type_guess(type_counts: dict[str, int]) -> str:
+    non_empty_counts = {key: value for key, value in type_counts.items() if key != "empty" and value}
+    if not non_empty_counts:
+        return "empty"
+    if len(non_empty_counts) == 1:
+        return next(iter(non_empty_counts))
+    if "error" in non_empty_counts:
+        return "mixed_with_errors"
+    if "string_id" in non_empty_counts and set(non_empty_counts) <= {"string_id", "string"}:
+        return "string_id"
+    return "mixed"
+
+
+def profile_columns(
+    grid: CsvGrid,
+    *,
+    header_row: int | None,
+    data_start_row: int,
+    data_end_row: int | None,
+) -> list[dict[str, Any]]:
+    header_values = []
+    if header_row in grid.row_numbers:
+        header_values = _row_values(grid, grid.row_numbers.index(header_row))
+
+    profiles = []
+    for idx, col in enumerate(grid.col_letters):
+        name = ""
+        if idx < len(header_values):
+            name = header_values[idx].strip()
+        if not name:
+            name = f"unnamed_{col}"
+
+        values = _column_values(grid, idx, data_start_row, data_end_row)
+        type_counts: dict[str, int] = {}
+        examples = []
+        error_count = 0
+        for value in values:
+            value_type = _classify_value(value)
+            type_counts[value_type] = type_counts.get(value_type, 0) + 1
+            if value_type == "error":
+                error_count += 1
+            if value.strip() and len(examples) < 5:
+                examples.append(value)
+        non_empty = len(values) - type_counts.get("empty", 0)
+        warnings = []
+        if type_counts.get("string_id"):
+            warnings.append("long_numeric_like_id")
+        if error_count:
+            warnings.append("formula_or_value_errors")
+        if non_empty and type_counts.get("empty", 0) / len(values) > 0.5:
+            warnings.append("many_empty_cells")
+        if _type_guess(type_counts) == "mixed":
+            warnings.append("mixed_value_types")
+
+        profiles.append(
+            {
+                "name": name,
+                "col": col,
+                "non_empty": non_empty,
+                "empty": type_counts.get("empty", 0),
+                "type_guess": _type_guess(type_counts),
+                "type_distribution": type_counts,
+                "examples": examples,
+                "warnings": warnings,
+            }
+        )
+    return profiles
+
+
+def _field_map(columns: list[dict[str, Any]]) -> dict[str, list[dict[str, Any]]]:
+    result: dict[str, list[dict[str, Any]]] = {}
+    for column in columns:
+        name = column["name"]
+        result.setdefault(name, []).append(
+            {
+                "col": column["col"],
+                "type_guess": column["type_guess"],
+                "warnings": column["warnings"],
+            }
+        )
+    return result
+
+
+def _risk_warnings(
+    *,
+    columns: list[dict[str, Any]],
+    special_rows: list[dict[str, Any]],
+    header_row: int | None,
+    data_range: str | None,
+    possible_multi_row_header: bool,
+    header_row_not_first: bool,
+    data_range_has_gaps: bool,
+    data_range_has_col_gaps: bool,
+    hidden_rows: list[int],
+    hidden_columns: list[str],
+) -> list[str]:
+    warnings = {warning for column in columns for warning in column["warnings"]}
+    names = [column["name"] for column in columns]
+    if len(names) != len(set(names)):
+        warnings.add("duplicate_headers")
+    if any(name.startswith("unnamed_") for name in names):
+        warnings.add("unnamed_columns")
+    if header_row is None:
+        warnings.add("header_not_detected")
+    if data_range is None:
+        warnings.add("data_range_not_detected")
+    if any(item["reason"] != "empty_row" for item in special_rows):
+        warnings.add("special_rows_present")
+    if any(item["reason"] == "empty_row" for item in special_rows):
+        warnings.add("empty_rows_present")
+    if possible_multi_row_header:
+        warnings.add("possible_multi_row_header")
+    if header_row_not_first:
+        warnings.add("header_row_not_first")
+    if data_range_has_gaps:
+        warnings.add("data_range_has_gaps")
+    if data_range_has_col_gaps:
+        warnings.add("data_range_has_col_gaps")
+    if hidden_rows:
+        warnings.add("hidden_rows_in_range")
+    if hidden_columns:
+        warnings.add("hidden_columns_in_range")
+    return sorted(warnings)
+
+
+def _col_segments(col_letters: list[str]) -> list[list[str]]:
+    """Group column letters into contiguous runs of real column indices.
+
+    With --skip-hidden the returned columns can be non-consecutive (A, C when B
+    is hidden). Reporting a single data_range across that gap would let a caller
+    write the hidden-column-free data back as if it were contiguous, shifting
+    every value right of the gap.
+    """
+    if not col_letters:
+        return []
+    indices = [col_to_index(col) for col in col_letters]
+    segments = []
+    start = previous = indices[0]
+    for current in indices[1:]:
+        if current == previous + 1:
+            previous = current
+            continue
+        segments.append([index_to_col(start), index_to_col(previous)])
+        start = previous = current
+    segments.append([index_to_col(start), index_to_col(previous)])
+    return segments
+
+
+def _row_segments(row_numbers: list[int]) -> list[list[int]]:
+    if not row_numbers:
+        return []
+    segments = []
+    start = end = row_numbers[0]
+    for row in row_numbers[1:]:
+        if row == end + 1:
+            end = row
+            continue
+        segments.append([start, end])
+        start = end = row
+    segments.append([start, end])
+    return segments
+
+
+def _write_hints(
+    grid: CsvGrid,
+    *,
+    header_row: int | None,
+    data_start: int,
+    bounds,
+    hidden_columns: list[str] | None = None,
+    skip_hidden: bool = False,
+) -> dict[str, Any]:
+    last_non_empty_col = None
+    for idx, col in enumerate(grid.col_letters):
+        if any(idx < len(row) and row[idx].strip() for row in grid.values):
+            last_non_empty_col = col_to_index(col)
+    safe_col_num = (last_non_empty_col + 1) if last_non_empty_col else bounds.start_col
+    # Step over hidden columns. Under --skip-hidden they are absent from the
+    # grid entirely, so "one past the last visible column" can land ON a hidden
+    # column that holds data — and when the hidden columns sit at the right edge
+    # there is no gap in the returned letters either, so data_range_has_col_gaps
+    # stays silent too. Appending there would overwrite data nobody can see.
+    hidden_indices = {col_to_index(col) for col in (hidden_columns or [])}
+    skipped_hidden: list[str] = []
+    while safe_col_num in hidden_indices:
+        skipped_hidden.append(index_to_col(safe_col_num))
+        safe_col_num += 1
+    safe_col = index_to_col(safe_col_num)
+    hints = {
+        "last_non_empty_col": index_to_col(last_non_empty_col) if last_non_empty_col else None,
+    }
+    if not (skip_hidden and hidden_columns is None):
+        hints.update(
+            {
+                "safe_append_col": safe_col,
+                "safe_append_header_cell": f"{safe_col}{header_row}" if header_row else None,
+                "safe_append_data_start_cell": f"{safe_col}{data_start}",
+            }
+        )
+    if skipped_hidden:
+        hints["skipped_hidden_cols"] = skipped_hidden
+    return hints
+
+
+def profile_grid(
+    grid: CsvGrid,
+    source_range: str,
+    *,
+    skip_hidden: bool = False,
+    hidden_rows: list[int] | None = None,
+    hidden_columns: list[str] | None = None,
+    all_hidden_columns: list[str] | None = None,
+    header_scan_rows: int = 20,
+) -> dict[str, Any]:
+    max_row = max(grid.row_numbers, default=1)
+    max_col = max((col_to_index(col) for col in grid.col_letters), default=1)
+    bounds = parse_range(source_range, max_row=max_row, max_col=max_col)
+    header_row = detect_header_row(grid, header_scan_rows)
+    data_start = (header_row + 1) if header_row else bounds.start_row
+    specials = detect_special_rows(grid, header_row)
+    data_end = _last_data_row(grid, header_row, specials)
+    data_range = None
+    if data_end is not None and data_end >= data_start:
+        data_range = format_range(data_start, bounds.start_col, data_end, bounds.end_col)
+    columns = profile_columns(
+        grid,
+        header_row=header_row,
+        data_start_row=data_start,
+        data_end_row=data_end,
+    )
+    data_row_numbers = []
+    if data_end is not None and data_end >= data_start:
+        data_row_numbers = [
+            row for row in grid.row_numbers if data_start <= row <= data_end
+        ]
+    data_row_segments = _row_segments(data_row_numbers)
+    data_range_has_gaps = bool(data_row_numbers) and (
+        data_row_numbers[0] != data_start or len(data_row_segments) > 1
+    )
+    data_rows = len(data_row_numbers)
+    data_col_segments = _col_segments(grid.col_letters)
+    data_range_has_col_gaps = len(data_col_segments) > 1
+    hidden_rows = hidden_rows or []
+    hidden_columns = hidden_columns or []
+    possible_multi_row_header = _possible_multi_row_header(grid, header_row)
+    header_row_not_first = _header_row_not_first(grid, header_row)
+
+    return {
+        "summary": {
+            "range": source_range,
+            "header_row": header_row,
+            "data_range": data_range,
+            "data_rows": data_rows,
+            "data_row_segments": data_row_segments,
+            "data_col_segments": data_col_segments,
+            "column_count": len(columns),
+            "special_rows_count": len(specials),
+        },
+        "range": source_range,
+        "header_row": header_row,
+        "data_range": data_range,
+        "data_row_segments": data_row_segments,
+        "data_col_segments": data_col_segments,
+        "columns": columns,
+        "field_map": _field_map(columns),
+        "risk_warnings": _risk_warnings(
+            columns=columns,
+            special_rows=specials,
+            header_row=header_row,
+            data_range=data_range,
+            possible_multi_row_header=possible_multi_row_header,
+            header_row_not_first=header_row_not_first,
+            data_range_has_gaps=data_range_has_gaps,
+            data_range_has_col_gaps=data_range_has_col_gaps,
+            hidden_rows=hidden_rows,
+            hidden_columns=hidden_columns,
+        ),
+        "visibility": {
+            "skip_hidden": skip_hidden,
+            "hidden_rows_in_range": hidden_rows,
+            "hidden_columns_in_range": hidden_columns,
+        },
+        "write_hints": _write_hints(
+            grid,
+            header_row=header_row,
+            data_start=data_start,
+            bounds=bounds,
+            hidden_columns=all_hidden_columns,
+            skip_hidden=skip_hidden,
+        ),
+        "special_rows": specials,
+    }
+
+
+def _hidden_rows_and_columns(grid: CsvGrid, layout: dict[str, Any]) -> tuple[list[int], list[str]]:
+    def indexes(key: str) -> set[int]:
+        values = layout.get(key)
+        if not isinstance(values, list):
+            return set()
+        result = set()
+        for value in values:
+            try:
+                result.add(int(value) + 1)  # +sheet-info uses zero-based indices.
+            except (TypeError, ValueError):
+                continue
+        return result
+
+    hidden_row_indexes = indexes("hidden_rows")
+    hidden_columns = layout.get("hidden_cols") or layout.get("hidden_columns") or []
+    hidden_column_letters = set()
+    for value in hidden_columns if isinstance(hidden_columns, list) else []:
+        if isinstance(value, str) and value.isalpha():
+            hidden_column_letters.add(value.upper())
+        else:
+            try:
+                hidden_column_letters.add(index_to_col(int(value) + 1))
+            except (TypeError, ValueError):
+                continue
+    rows = sorted(row for row in grid.row_numbers if row in hidden_row_indexes)
+    columns = [col for col in grid.col_letters if col.upper() in hidden_column_letters]
+    return rows, columns
+
+
+def _all_hidden_columns(layout: dict[str, Any]) -> list[str]:
+    """Every hidden column on the sheet, not just those inside the grid.
+
+    _hidden_rows_and_columns intersects with the returned grid, which is the
+    right scope for the "there are hidden rows/columns in what you read"
+    warnings. The append hint needs the opposite: the columns that are NOT in
+    the grid precisely because they are hidden.
+    """
+    letters: list[str] = []
+    raw = layout.get("hidden_cols") or layout.get("hidden_columns") or []
+    for value in raw if isinstance(raw, list) else []:
+        if isinstance(value, str) and value.isalpha():
+            letters.append(value.upper())
+            continue
+        try:
+            letters.append(index_to_col(int(value) + 1))
+        except (TypeError, ValueError):
+            continue
+    return letters
+
+
+def profile_table(args) -> tuple[dict[str, Any], list[str]]:
+    warnings = []
+    if args.header_scan_rows < 1:
+        raise ValueError("--header-scan-rows must be at least 1")
+    csv_data = envelope_data(
+        run_sheets(
+            "+csv-get",
+            url=args.url,
+            spreadsheet_token=args.spreadsheet_token,
+            sheet_id=args.sheet_id,
+            sheet_name=args.sheet_name,
+            flags={
+                "range": args.range,
+                "max_chars": args.max_chars,
+                "skip_hidden": True if args.skip_hidden else None,
+            },
+            timeout=args.timeout,
+        )
+    )
+    source_range = str(csv_data.get("actual_range") or args.range)
+    if csv_data.get("has_more"):
+        raise LarkCliError(
+            f"+csv-get truncated the requested range at {source_range}; narrow --range before profiling"
+        )
+    grid = parse_annotated_csv(
+        csv_data.get("annotated_csv", ""),
+        csv_data.get("col_indices"),
+        csv_data.get("row_indices"),
+        source_range,
+    )
+    if grid.row_numbers_inferred:
+        warnings.append("CSV row numbers were inferred from the requested range")
+    hidden_rows: list[int] = []
+    hidden_columns: list[str] = []
+    all_hidden_columns: list[str] | None = None
+    # Fetched in BOTH modes. Under --skip-hidden the hidden rows/columns are
+    # absent from the grid, so hidden_rows/hidden_columns (which are scoped to
+    # what the grid contains) come back empty and no warning fires — but the
+    # write hints still need to know where the hidden columns are, or
+    # safe_append_col can point at one that holds data. all_hidden_columns is
+    # the unscoped list used for exactly that.
+    try:
+        layout = envelope_data(
+            run_sheets(
+                "+sheet-info",
+                url=args.url,
+                spreadsheet_token=args.spreadsheet_token,
+                sheet_id=args.sheet_id,
+                sheet_name=args.sheet_name,
+                flags={"include": "hidden_rows,hidden_cols"},
+                timeout=args.timeout,
+            )
+        )
+        hidden_rows, hidden_columns = _hidden_rows_and_columns(grid, layout)
+        all_hidden_columns = _all_hidden_columns(layout)
+    except LarkCliError as exc:
+        warnings.append(f"hidden row/column detection unavailable: {exc}")
+    return profile_grid(
+        grid,
+        source_range,
+        skip_hidden=args.skip_hidden,
+        hidden_rows=hidden_rows,
+        hidden_columns=hidden_columns,
+        all_hidden_columns=all_hidden_columns,
+        header_scan_rows=args.header_scan_rows,
+    ), warnings
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description=__doc__)
+    add_spreadsheet_args(parser, require_sheet=True, allow_sheet=True)
+    parser.add_argument("--range", required=True)
+    parser.add_argument("--max-chars", type=int, default=25000)
+    parser.add_argument("--header-scan-rows", type=int, default=20)
+    parser.add_argument("--skip-hidden", action="store_true")
+    parser.add_argument("--timeout", type=int, default=60)
+    args = parser.parse_args()
+    try:
+        data, warnings = profile_table(args)
+    except (LarkCliError, ValueError, TypeError) as exc:
+        emit_error(ACTION, str(exc))
+    emit_success(ACTION, data, warnings)
+
+
+if __name__ == "__main__":
+    main()