fix(skills): inspect a legacy .xls by its values, not by its storage

astrbot/builtin_stars/astrbot/skills/spreadsheets/scripts/inspect_workbook.py

@@ -7,6 +7,7 @@
 import csv
 import json
 import zipfile
+from datetime import datetime, time
 from pathlib import Path
 
 import chardet
@@ -56,6 +57,35 @@ def _inspect_delimited(path: Path, sample_rows: int, sample_cols: int) -> dict:
     }
 
 
+def _xls_cell_value(cell: xlrd.sheet.Cell, datemode: int) -> object:
+    """Read one xlrd cell as its value, the way the XLSX path already reads one.
+
+    xlrd hands back the raw storage: a date is the serial number Excel keeps it
+    as, a boolean is 1 or 0, and every number is a double, so a whole number
+    arrives as 12.0. Left alone, the same workbook is described one way as .xls
+    and another as .xlsx, and the date cannot be recovered from the sample.
+    """
+    if cell.ctype == xlrd.XL_CELL_DATE:
+        try:
+            year, month, day, hour, minute, second = xlrd.xldate_as_tuple(
+                cell.value, datemode
+            )
+        except (ValueError, xlrd.XLDateError):
+            return cell.value
+        if (year, month, day) == (0, 0, 0):
+            # A time-only cell has no date part; openpyxl reads one as a time.
+            return time(hour, minute, second)
+        return datetime(year, month, day, hour, minute, second)
+    if cell.ctype == xlrd.XL_CELL_BOOLEAN:
+        return bool(cell.value)
+    if cell.ctype == xlrd.XL_CELL_NUMBER and float(cell.value).is_integer():
+        return int(cell.value)
+    if cell.ctype == xlrd.XL_CELL_ERROR:
+        # openpyxl reports the text Excel shows, e.g. #DIV/0!
+        return xlrd.error_text_from_code.get(cell.value, "")
+    return cell.value
+
+
 def _inspect_xls(path: Path, sample_rows: int, sample_cols: int) -> dict:
     """Inspect a legacy XLS workbook.
 
@@ -71,13 +101,17 @@ def _inspect_xls(path: Path, sample_rows: int, sample_cols: int) -> dict:
     sheets = []
     for name in workbook.sheet_names():
         sheet = workbook.sheet_by_name(name)
+        columns = min(sheet.ncols, sample_cols)
         sheets.append(
             {
                 "name": name,
                 "rows": sheet.nrows,
                 "columns": sheet.ncols,
                 "sample": [
-                    sheet.row_values(row, end_colx=min(sheet.ncols, sample_cols))
+                    [
+                        _xls_cell_value(cell, workbook.datemode)
+                        for cell in sheet.row_slice(row, 0, columns)
+                    ]
                     for row in range(min(sheet.nrows, sample_rows))
                 ],
             }

pyproject.toml

@@ -80,6 +80,7 @@ dev = [
   "pytest>=8.4.1",
   "pytest-asyncio>=1.1.0",
   "pytest-cov>=6.2.1",
+  "xlwt>=1.3.0",  # writes the legacy .xls fixtures the spreadsheet skill tests read
   "ruff==0.15.22",
 ]
 

tests/test_builtin_office_skills.py

@@ -74,6 +74,98 @@ def test_spreadsheet_skill_converts_inspects_and_validates_csv(tmp_path: Path) -
     assert json.loads(validated.stdout)["valid"] is True
 
 
+def test_spreadsheet_skill_inspects_legacy_xls_values_not_their_storage(
+    tmp_path: Path,
+) -> None:
+    """xlrd hands back the raw storage, so the two formats described one sheet
+    two different ways: a date as its serial number, a boolean as 1, and a whole
+    number as a float.
+    """
+    import pytest
+
+    xlwt = pytest.importorskip("xlwt")
+
+    import datetime
+
+    date_style = xlwt.XFStyle()
+    date_style.num_format_str = "YYYY-MM-DD"
+
+    book = xlwt.Workbook()
+    sheet = book.add_sheet("Data")
+    for column, heading in enumerate(["When", "Active", "Units", "Rate"]):
+        sheet.write(0, column, heading)
+    sheet.write(1, 0, datetime.date(2024, 1, 1), date_style)
+    sheet.write(1, 1, True)
+    sheet.write(1, 2, 12)
+    sheet.write(1, 3, 1.5)
+    legacy = tmp_path / "legacy.xls"
+    book.save(legacy)
+
+    modern = tmp_path / "modern.xlsx"
+    workbook = Workbook()
+    worksheet = workbook.active
+    worksheet.title = "Data"
+    worksheet.append(["When", "Active", "Units", "Rate"])
+    worksheet.append([datetime.date(2024, 1, 1), True, 12, 1.5])
+    workbook.save(modern)
+    workbook.close()
+
+    inspected_xls = _run_script(SPREADSHEET_SCRIPTS / "inspect_workbook.py", legacy)
+    inspected_xlsx = _run_script(SPREADSHEET_SCRIPTS / "inspect_workbook.py", modern)
+
+    assert inspected_xls.returncode == 0, inspected_xls.stderr
+    assert inspected_xlsx.returncode == 0, inspected_xlsx.stderr
+
+    xls_sample = json.loads(inspected_xls.stdout)["sheets"][0]["sample"]
+    xlsx_sample = json.loads(inspected_xlsx.stdout)["sheets"][0]["sample"]
+
+    # [45292.0, 1, 12.0, 1.5] before this.
+    assert xls_sample[1] == ["2024-01-01 00:00:00", True, 12, 1.5]
+    assert xls_sample == xlsx_sample
+
+
+def test_spreadsheet_skill_inspects_a_legacy_xls_time_only_cell(
+    tmp_path: Path,
+) -> None:
+    """A time carries no date, so xlrd reports year, month and day as zero."""
+    import pytest
+
+    xlwt = pytest.importorskip("xlwt")
+
+    import datetime
+
+    time_style = xlwt.XFStyle()
+    time_style.num_format_str = "HH:MM:SS"
+
+    book = xlwt.Workbook()
+    sheet = book.add_sheet("Data")
+    sheet.write(0, 0, "Starts")
+    sheet.write(1, 0, datetime.time(12, 0, 0), time_style)
+    legacy = tmp_path / "legacy.xls"
+    book.save(legacy)
+
+    modern = tmp_path / "modern.xlsx"
+    workbook = Workbook()
+    worksheet = workbook.active
+    worksheet.title = "Data"
+    worksheet.append(["Starts"])
+    worksheet.append([datetime.time(12, 0, 0)])
+    workbook.save(modern)
+    workbook.close()
+
+    inspected_xls = _run_script(SPREADSHEET_SCRIPTS / "inspect_workbook.py", legacy)
+    inspected_xlsx = _run_script(SPREADSHEET_SCRIPTS / "inspect_workbook.py", modern)
+
+    assert inspected_xls.returncode == 0, inspected_xls.stderr
+    assert inspected_xlsx.returncode == 0, inspected_xlsx.stderr
+
+    xls_sample = json.loads(inspected_xls.stdout)["sheets"][0]["sample"]
+    xlsx_sample = json.loads(inspected_xlsx.stdout)["sheets"][0]["sample"]
+
+    assert xls_sample[1] == ["12:00:00"]
+    assert xls_sample == xlsx_sample
+
+
 def test_spreadsheet_skill_rejects_broken_formula_reference(tmp_path: Path) -> None:
     path = tmp_path / "broken.xlsx"
     workbook = Workbook()