Merge pull request #5188 from VesnaT/excel_missings

ExcelReader: #VALUE! as np.nan

Author: janezd

Orange/data/io.py

@@ -275,6 +275,7 @@ class ExcelReader(_BaseExcelReader):
     """Reader for .xlsx files"""
     EXTENSIONS = ('.xlsx',)
     DESCRIPTION = 'Microsoft Excel spreadsheet'
+    ERRORS = ("#VALUE!", "#DIV/0!", "#REF!", "#NUM!", "#NULL!", "#NAME?")
 
     @property
     def workbook(self) -> openpyxl.Workbook:
@@ -296,7 +297,8 @@ def sheets(self) -> List:
 
     def get_cells(self) -> Iterable:
         def str_(x):
-            return str(x) if x is not None else ""
+            return str(x) if x is not None and x not in ExcelReader.ERRORS \
+                else ""
 
         sheet = self._get_active_sheet()
         min_col = sheet.min_column

Orange/tests/test_xlsx_reader.py

@@ -181,5 +181,12 @@ def test_read(self, reader: Callable[[str], io.FileFormat]):
             table.metas[:, 2], np.array(list("abcdefghijklmnopqrstuvw")))
 
 
+class TestMissingValues(unittest.TestCase):
+    def test_read_errors(self):
+        table = read_file(get_xlsx_reader, "missing")
+        values = table.get_column_view("C")[0]
+        self.assertTrue(np.isnan(values).all())
+
+
 if __name__ == "__main__":
     unittest.main()