SNOW-1875145: Support querying json element in col

CHANGELOG.md

@@ -47,6 +47,7 @@
 #### Experimental Features
 
 - Added `Catalog` class to manage snowflake objects. It can be accessed via `Session.catalog`.
+- Added support for querying json element of a VARIANT column in `functions.col` and `functions.column` with an optional keyword argument `json_element`.
 - Allow user input schema when reading JSON file on stage.
 - Added support for specifying a schema string (including implicit struct syntax) when calling `DataFrame.create_dataframe`.
   - `snowflake.core` is a dependency required for this feature.

src/snowflake/snowpark/column.py

@@ -252,11 +252,27 @@ def __init__(
         self,
         expr1: Union[str, Expression],
         expr2: Optional[str] = None,
+        json_element: bool = False,
         _ast: Optional[proto.Expr] = None,
         _emit_ast: bool = True,
     ) -> None:
         self._ast = _ast
 
+        def derive_json_element_expr(
+            expr: str, df_alias: Optional[str] = None
+        ) -> UnresolvedAttribute:
+            parts = expr.split(".")
+            if len(parts) == 1:
+                return UnresolvedAttribute(quote_name(parts[0]), df_alias=df_alias)
+            else:
+                # According to https://docs.snowflake.com/en/user-guide/querying-semistructured#dot-notation,
+                # the json value on the path should be case-sensitive
+                return UnresolvedAttribute(
+                    f"{quote_name(parts[0])}:{'.'.join(quote_name(part, keep_case=True) for part in parts[1:])}",
+                    is_sql_text=True,
+                    df_alias=df_alias,
+                )
+
         if expr2 is not None:
             if not (isinstance(expr1, str) and isinstance(expr2, str)):
                 raise ValueError(
@@ -265,6 +281,8 @@ def __init__(
 
             if expr2 == "*":
                 self._expression = Star([], df_alias=expr1)
+            elif json_element:
+                self._expression = derive_json_element_expr(expr2, expr1)
             else:
                 self._expression = UnresolvedAttribute(
                     quote_name(expr2), df_alias=expr1
@@ -279,6 +297,8 @@ def __init__(
         elif isinstance(expr1, str):
             if expr1 == "*":
                 self._expression = Star([])
+            elif json_element:
+                self._expression = derive_json_element_expr(expr1)
             else:
                 self._expression = UnresolvedAttribute(quote_name(expr1))
 
@@ -1483,9 +1503,15 @@ class CaseExpr(Column):
     """
 
     def __init__(
-        self, expr: CaseWhen, _ast: Optional[proto.Expr] = None, _emit_ast: bool = True
+        self,
+        expr: CaseWhen,
+        json_element: bool = False,
+        _ast: Optional[proto.Expr] = None,
+        _emit_ast: bool = True,
     ) -> None:
-        super().__init__(expr, _ast=_ast, _emit_ast=_emit_ast)
+        super().__init__(
+            expr, json_element=json_element, _ast=_ast, _emit_ast=_emit_ast
+        )
         self._branches = expr.branches
 
     @publicapi

src/snowflake/snowpark/functions.py

@@ -262,20 +262,31 @@ def _check_column_parameters(name1: str, name2: Optional[str]) -> None:
 
 @overload
 @publicapi
-def col(col_name: str, _emit_ast: bool = True) -> Column:
+def col(col_name: str, json_element: bool = False, _emit_ast: bool = True) -> Column:
     """Returns the :class:`~snowflake.snowpark.Column` with the specified name.
 
+    Args:
+        col_name: The name of the column.
+        json_element: Whether the column is a JSON element. If a column is a VARIANT column in Snowflake,
+            you can dot notation `.` to query the nested json element, e.g., "name.firstname" and "name.lastname".
+
     Example::
         >>> df = session.sql("select 1 as a")
         >>> df.select(col("a")).collect()
         [Row(A=1)]
+
+        >>> df = session.sql("select parse_json('{\"firstname\": \"John\", \"lastname\": \"Doe\"}') as name")
+        >>> df.select(col("name.firstname", json_element=True)).collect()
+        [Row(FIRSTNAME='John')]
     """
     ...  # pragma: no cover
 
 
 @overload
 @publicapi
-def col(df_alias: str, col_name: str, _emit_ast: bool = True) -> Column:
+def col(
+    df_alias: str, col_name: str, json_element: bool = False, _emit_ast: bool = True
+) -> Column:
     """Returns the :class:`~snowflake.snowpark.Column` with the specified dataframe alias and column name.
 
     Example::
@@ -287,7 +298,12 @@ def col(df_alias: str, col_name: str, _emit_ast: bool = True) -> Column:
 
 
 @publicapi
-def col(name1: str, name2: Optional[str] = None, _emit_ast: bool = True) -> Column:
+def col(
+    name1: str,
+    name2: Optional[str] = None,
+    json_element: bool = False,
+    _emit_ast: bool = True,
+) -> Column:
 
     _check_column_parameters(name1, name2)
 
@@ -296,27 +312,38 @@ def col(name1: str, name2: Optional[str] = None, _emit_ast: bool = True) -> Colu
         ast = create_ast_for_column(name1, name2, "col")
 
     if name2 is None:
-        return Column(name1, _ast=ast)
+        return Column(name1, json_element=json_element, _ast=ast)
     else:
-        return Column(name1, name2, _ast=ast)
+        return Column(name1, name2, json_element=json_element, _ast=ast)
 
 
 @overload
 @publicapi
-def column(col_name: str, _emit_ast: bool = True) -> Column:
+def column(col_name: str, json_element: bool = False, _emit_ast: bool = True) -> Column:
     """Returns a :class:`~snowflake.snowpark.Column` with the specified name. Alias for col.
 
+    Args:
+         col_name: The name of the column.
+         json_element: Whether the column is a JSON element. If a column is a VARIANT column in Snowflake,
+             you can dot notation `.` to query the nested json element, e.g., "name.firstname" and "name.lastname".
+
     Example::
-        >>> df = session.sql("select 1 as a")
-        >>> df.select(column("a")).collect()
-        [Row(A=1)]
+         >>> df = session.sql("select 1 as a")
+         >>> df.select(column("a")).collect()
+         [Row(A=1)]
+
+         >>> df = session.sql("select parse_json('{\"firstname\": \"John\", \"lastname\": \"Doe\"}') as name")
+         >>> df.select(column("name.firstname", json_element=True)).collect()
+         [Row(FIRSTNAME='John')]
     """
     ...  # pragma: no cover
 
 
 @overload
 @publicapi
-def column(df_alias: str, col_name: str, _emit_ast: bool = True) -> Column:
+def column(
+    df_alias: str, col_name: str, json_element: bool = False, _emit_ast: bool = True
+) -> Column:
     """Returns a :class:`~snowflake.snowpark.Column` with the specified name and dataframe alias name. Alias for col.
 
     Example::
@@ -328,15 +355,20 @@ def column(df_alias: str, col_name: str, _emit_ast: bool = True) -> Column:
 
 
 @publicapi
-def column(name1: str, name2: Optional[str] = None, _emit_ast: bool = True) -> Column:
+def column(
+    name1: str,
+    name2: Optional[str] = None,
+    json_element: bool = False,
+    _emit_ast: bool = True,
+) -> Column:
     _check_column_parameters(name1, name2)
 
     ast = create_ast_for_column(name1, name2, "column") if _emit_ast else None
 
     if name2 is None:
-        return Column(name1, _ast=ast)
+        return Column(name1, json_element=json_element, _ast=ast)
     else:
-        return Column(name1, name2, _ast=ast)
+        return Column(name1, name2, json_element=json_element, _ast=ast)
 
 
 @publicapi

tests/integ/test_function.py

@@ -185,6 +185,53 @@
 )
 
 
+@pytest.mark.skipif(
+    "config.getoption('local_testing_mode', default=False)",
+    reason="querying json element is not supported in local testing",
+)
+def test_col_json_element(session):
+    # 2-level deep
+    df = session.sql(
+        'select parse_json(\'{"firstname": "John", "lastname": "Doe"}\') as name'
+    )
+    Utils.check_answer(
+        df.select(
+            col("name.firstname", json_element=True),
+            col("name.lastname", json_element=True),
+        ),
+        [Row('"John"', '"Doe"')],
+    )
+    Utils.check_answer(
+        df.select(
+            col('name."firstname"', json_element=True),
+            col('NAME."lastname"', json_element=True),
+        ),
+        [Row('"John"', '"Doe"')],
+    )
+    Utils.check_answer(df.select(col("name.FIRSTNAME", json_element=True)), [Row(None)])
+
+    # 3-level deep
+    with pytest.raises(SnowparkSQLException, match="invalid identifier"):
+        df.select(col("name:firstname", json_element=True)).collect()
+
+    with pytest.raises(SnowparkSQLException, match="invalid identifier"):
+        df.select(col("name.firstname")).collect()
+
+    df = session.sql('select parse_json(\'{"l1": {"l2": "xyz"}}\') as value')
+    Utils.check_answer(df.select(col("value.l1.l2", json_element=True)), Row('"xyz"'))
+    Utils.check_answer(
+        df.select(col('value."l1"."l2"', json_element=True)), Row('"xyz"')
+    )
+    Utils.check_answer(df.select(col("value.L1.l2", json_element=True)), Row(None))
+    Utils.check_answer(df.select(col("value.l1.L2", json_element=True)), Row(None))
+
+    with pytest.raises(SnowparkSQLException, match="invalid identifier"):
+        df.select(col("value:l1.l2", json_element=True)).collect()
+
+    with pytest.raises(SnowparkSQLException, match="invalid identifier"):
+        df.select(col("value.l1.l2")).collect()
+
+
 def test_order(session):
     null_data1 = TestData.null_data1(session)
     assert null_data1.sort(asc(null_data1["A"])).collect() == [