Backport PR #31513: REGR: Fixed truncation with na_rep (#31542)

Co-authored-by: Tom Augspurger <[email protected]>

Author: meeseeksmachine

doc/source/whatsnew/v1.0.1.rst

@@ -87,6 +87,7 @@ MultiIndex
 I/O
 ^^^
 
+- Fixed regression in :meth:`~DataFrame.to_csv` where specifying an ``na_rep`` might truncate the values written (:issue:`31447`)
 -
 -
 

pandas/core/internals/blocks.py

@@ -680,7 +680,10 @@ def to_native_types(self, slicer=None, na_rep="nan", quoting=None, **kwargs):
         itemsize = writers.word_len(na_rep)
 
         if not self.is_object and not quoting and itemsize:
-            values = values.astype(f"<U{itemsize}")
+            values = values.astype(str)
+            if values.dtype.itemsize / np.dtype("U1").itemsize < itemsize:
+                # enlarge for the na_rep
+                values = values.astype(f"<U{itemsize}")
         else:
             values = np.array(values, dtype="object")
 

pandas/tests/io/formats/test_to_csv.py

@@ -583,3 +583,17 @@ def test_to_csv_timedelta_precision(self):
         ]
         expected = tm.convert_rows_list_to_csv_str(expected_rows)
         assert result == expected
+
+    def test_na_rep_truncated(self):
+        # https://github.com/pandas-dev/pandas/issues/31447
+        result = pd.Series(range(8, 12)).to_csv(na_rep="-")
+        expected = tm.convert_rows_list_to_csv_str([",0", "0,8", "1,9", "2,10", "3,11"])
+        assert result == expected
+
+        result = pd.Series([True, False]).to_csv(na_rep="nan")
+        expected = tm.convert_rows_list_to_csv_str([",0", "0,True", "1,False"])
+        assert result == expected
+
+        result = pd.Series([1.1, 2.2]).to_csv(na_rep=".")
+        expected = tm.convert_rows_list_to_csv_str([",0", "0,1.1", "1,2.2"])
+        assert result == expected