SNOW-1356685: Fix precision loss bug in DataFrame/Series with dtype=np.uint64

CHANGELOG.md

@@ -15,6 +15,7 @@
 - Fixed a bug where python stored procedure with table return type fails when run in a task.
 - Fixed a bug where df.dropna fails due to `RecursionError: maximum recursion depth exceeded` when the DataFrame has more than 500 columns.
 - Fixed a bug where `AsyncJob.result("no_result")` doesn't wait for the query to finish execution.
+- Fixed a bug regarding precision loss when converting to Snowpark pandas `DataFrame` or `Series` with `dtype=np.uint64`.
 
 ### Snowpark Local Testing Updates
 
@@ -50,6 +51,7 @@
 - Fixed a bug that causes output of GroupBy.aggregate's columns to be ordered incorrectly.
 - Fixed a bug where `DataFrame.describe` on a frame with duplicate columns of differing dtypes could cause an error or incorrect results.
 - Fixed a bug in `DataFrame.rolling` and `Series.rolling` so `window=0` now throws `NotImplementedError` instead of `ValueError`
+- Fixed a bug in `DataFrame` and `Series` with `dtype=np.uint64` resulting in precision errors
 
 #### Improvements
 

src/snowflake/snowpark/_internal/server_connection.py

@@ -734,7 +734,7 @@ def _fix_pandas_df_fixed_type(
                         pd_df[pandas_col_name] = pd_df[pandas_col_name].astype("int64")
                     except OverflowError:
                         pd_df[pandas_col_name] = pandas.to_numeric(
-                            pd_df[pandas_col_name], downcast="integer"
+                            pd_df[pandas_col_name]
                         )
                 else:
                     pd_df[pandas_col_name] = pandas.to_numeric(

tests/integ/modin/series/test_unique.py

@@ -1,6 +1,7 @@
 #
 # Copyright (c) 2012-2024 Snowflake Computing Inc. All rights reserved.
 #
+from decimal import Decimal
 from typing import Any
 
 import modin.pandas as pd
@@ -26,13 +27,20 @@ def _make_nan_interleaved_float_series():
     [12.0, 11.999999, 11.999999],
     ["A", "A", "C", "C", "A"],
     [None, "A", None, "B"],
+    _make_nan_interleaved_float_series(),
+    native_pd.Series([1, 2, 2**63, 2**63], dtype=np.uint64),
     pytest.param(
-        native_pd.Series([1, 2, 2**63, 2**63], dtype=np.uint64),
+        native_pd.Series([1, 2, -(2**63) - 1, -(2**64)]),
         marks=pytest.mark.xfail(
-            reason="SNOW-1356685: Dtype with unsigned int results in precision error"
+            reason="Represent overflow using float instead of integer",
+        ),
+    ),
+    pytest.param(
+        native_pd.Series([Decimal(1.5), Decimal(2**64 - 1)], dtype=object),
+        marks=pytest.mark.xfail(
+            reason="Represent Decimal using float instead of integer as pandas does not recognize it",
         ),
     ),
-    _make_nan_interleaved_float_series(),
 ]