BUG: ignore errors for invalid dates in to_datetime with coerce (pandas-dev#25512)

parsing.try_parse_year_month_day() in _attempt_YYYYMMDD() throws not only ValueError but also OverFlowError for incorrect dates. So handling of this error was added.

Author: nathalier

pandas/core/tools/datetimes.py

@@ -775,21 +775,21 @@ def calc_with_mask(carg, mask):
     # try intlike / strings that are ints
     try:
         return calc(arg.astype(np.int64))
-    except ValueError:
+    except (ValueError, OverflowError):
         pass
 
     # a float with actual np.nan
     try:
         carg = arg.astype(np.float64)
         return calc_with_mask(carg, notna(carg))
-    except ValueError:
+    except (ValueError, OverflowError):
         pass
 
     # string with NaN-like
     try:
         mask = ~algorithms.isin(arg, list(tslib.nat_strings))
         return calc_with_mask(arg, mask)
-    except ValueError:
+    except (ValueError, OverflowError):
         pass
 
     return None

pandas/tests/indexes/datetimes/test_tools.py

@@ -83,6 +83,21 @@ def test_to_datetime_format_YYYYMMDD(self, cache):
         result = to_datetime(s, format='%Y%m%d', cache=cache)
         assert_series_equal(result, expected)
 
+        # GH 25512
+        # strings with invalid date values, errors=coerce
+        s = Series(['19801222', '20010012', '10019999', np.nan])
+        result = pd.to_datetime(s, format='%Y%m%d', errors='coerce',
+                                cache=cache)
+        expected = Series([Timestamp('19801222'), np.nan, np.nan, np.nan])
+        tm.assert_series_equal(result, expected)
+
+        # integers with invalid date values, errors=coerce
+        s = Series([20010012, 20190813, 20019999, np.nan])
+        result = pd.to_datetime(s, format='%Y%m%d', errors='coerce',
+                                cache=cache)
+        expected = Series([np.nan, Timestamp('20190813'), np.nan, np.nan])
+        tm.assert_series_equal(result, expected)
+
         # coercion
         # GH 7930
         s = Series([20121231, 20141231, 99991231])