Fix: src/tablib/formats/_xlsx.py's dset_sheet() does cell.value = col...

src/tablib/formats/_xlsx.py

@@ -14,6 +14,7 @@
 import re
 from io import BytesIO
 
+from openpyxl.cell.cell import ILLEGAL_CHARACTERS_RE
 from openpyxl.reader.excel import ExcelReader, load_workbook
 from openpyxl.styles import Alignment, Font
 from openpyxl.utils import get_column_letter
@@ -23,11 +24,30 @@
 
 INVALID_TITLE_REGEX = re.compile(r'[\\*?:/\[\]]')
 
+# Lone surrogates (e.g. produced by ``surrogateescape`` error handling)
+# cannot be encoded to valid XML/UTF-8 either, so they need stripping
+# alongside the ASCII control characters openpyxl itself rejects.
+SURROGATES_RE = re.compile(r'[\ud800-\udfff]')
+
 
 def safe_xlsx_sheet_title(s, replace="-"):
     return re.sub(INVALID_TITLE_REGEX, replace, s)[:31]
 
 
+def _sanitize_value(value):
+    """Strip characters that openpyxl/XML cannot represent in a cell.
+
+    openpyxl raises ``IllegalCharacterError`` (not a ``ValueError``) for
+    strings containing certain ASCII control characters, and silently
+    writes an unreadable file for lone surrogate characters. Sanitize
+    both cases up front so the value can be written safely.
+    """
+    if isinstance(value, str):
+        value = ILLEGAL_CHARACTERS_RE.sub('', value)
+        value = SURROGATES_RE.sub('', value)
+    return value
+
+
 class XLSXFormat:
     title = 'xlsx'
     extensions = ('xlsx',)
@@ -182,7 +202,7 @@ def dset_sheet(cls, dataset, ws, freeze_panes=True, escape=False):
                         cell.alignment = wrap_text
 
                 try:
-                    cell.value = col
+                    cell.value = _sanitize_value(col)
                 except ValueError:
                     cell.value = str(col)
 

tests/test_tablib.py

@@ -1644,6 +1644,16 @@ def test_xlsx_raise_ValueError_on_cell_write_during_export(self):
         wb = load_workbook(filename=BytesIO(_xlsx))
         self.assertEqual('[1]', wb.active['A1'].value)
 
+    def test_xlsx_export_strips_illegal_characters(self):
+        """Cell values containing XML-illegal ASCII control characters
+        must be sanitized rather than raising IllegalCharacterError.
+        See: https://github.com/jazzband/tablib/issues/370
+        """
+        data.append((f'a{chr(31)}b',))
+        _xlsx = data.export('xlsx')
+        wb = load_workbook(filename=BytesIO(_xlsx))
+        self.assertEqual('ab', wb.active['A1'].value)
+
     def test_xlsx_column_width_adaptive(self):
         """ Test that column width adapts to value length"""
         width_before, width_after = self._helper_export_column_width("adaptive")