BUG: read_json not reading in large ints properly

doc/source/whatsnew/v1.3.0.rst

@@ -795,6 +795,7 @@ I/O
 - Bug in :func:`read_sas` raising ``ValueError`` when ``datetimes`` were null (:issue:`39725`)
 - Bug in :func:`read_excel` dropping empty values from single-column spreadsheets (:issue:`39808`)
 - Bug in :meth:`DataFrame.to_string` misplacing the truncation column when ``index=False`` (:issue:`40907`)
+- Bug in :func:`read_json` reading large integers incorrectly if dtype is not specified (:issue:`20608`)
 - Bug in :func:`read_orc` always raising ``AttributeError`` (:issue:`40918`)
 
 Period

pandas/io/json/_json.py

@@ -947,12 +947,17 @@ def _try_convert_data(self, name, data, use_dtypes=True, convert_dates=True):
                 return new_data, True
 
         if data.dtype == "object":
-
-            # try float
-            try:
+            if len(data) == 0:
                 data = data.astype("float64")
-            except (TypeError, ValueError):
-                pass
+            else:
+                try:
+                    data = data.astype("int64")
+                except (TypeError, ValueError, OverflowError):
+                    try:
+                        # Maybe out of integer range, try float
+                        data = data.astype("float64")
+                    except (TypeError, ValueError, OverflowError):
+                        pass
 
         if data.dtype.kind == "f":
 
@@ -965,7 +970,7 @@ def _try_convert_data(self, name, data, use_dtypes=True, convert_dates=True):
                     pass
 
         # don't coerce 0-len data
-        if len(data) and (data.dtype == "float" or data.dtype == "object"):
+        if len(data) and data.dtype == "float":
 
             # coerce ints if we can
             try:

pandas/tests/io/json/test_pandas.py

@@ -1305,6 +1305,13 @@ def test_read_json_large_numbers2(self):
         expected = DataFrame(1.404366e21, index=["articleId"], columns=[0])
         tm.assert_frame_equal(result, expected)
 
+    def test_large_ints_from_json_strings(self, orient):
+        # GH 20608
+        expected = DataFrame([9999999999999999, 10000000000000001])
+        df_temp = expected.copy().astype(str)
+        result = read_json(df_temp.to_json(orient=orient), orient=orient)
+        tm.assert_frame_equal(result, expected)
+
     def test_to_jsonl(self):
         # GH9180
         df = DataFrame([[1, 2], [1, 2]], columns=["a", "b"])