[SPARK-49693][PYTHON][CONNECT] Refine the string representation of timedelta

### What changes were proposed in this pull request?
Refine the string representation of `timedelta`, by following the ISO format.
Note that the used units in JVM side (`Duration`) and Pandas are different.

### Why are the changes needed?
We should not leak the raw data

### Does this PR introduce _any_ user-facing change?
yes

PySpark Classic:
```
In [1]: from pyspark.sql import functions as sf

In [2]: import datetime

In [3]: sf.lit(datetime.timedelta(1, 1))
Out[3]: Column<'PT24H1S'>
```

PySpark Connect (before):
```
In [1]: from pyspark.sql import functions as sf

In [2]: import datetime

In [3]: sf.lit(datetime.timedelta(1, 1))
Out[3]: Column<'86401000000'>
```

PySpark Connect (after):
```
In [1]: from pyspark.sql import functions as sf

In [2]: import datetime

In [3]: sf.lit(datetime.timedelta(1, 1))
Out[3]: Column<'P1DT0H0M1S'>
```

### How was this patch tested?
added test

### Was this patch authored or co-authored using generative AI tooling?
no

Closes #48159 from zhengruifeng/pc_lit_delta.

Authored-by: Ruifeng Zheng <[email protected]>
Signed-off-by: Ruifeng Zheng <[email protected]>

Author: zhengruifeng

python/pyspark/sql/connect/expressions.py

@@ -489,7 +489,17 @@ def __repr__(self) -> str:
             ts = TimestampNTZType().fromInternal(self._value)
             if ts is not None and isinstance(ts, datetime.datetime):
                 return ts.strftime("%Y-%m-%d %H:%M:%S.%f")
-        # TODO(SPARK-49693): Refine the string representation of timedelta
+        elif isinstance(self._dataType, DayTimeIntervalType):
+            delta = DayTimeIntervalType().fromInternal(self._value)
+            if delta is not None and isinstance(delta, datetime.timedelta):
+                import pandas as pd
+
+                # Note: timedelta itself does not provide isoformat method.
+                # Both Pandas and java.time.Duration provide it, but the format
+                # is sightly different:
+                # java.time.Duration only applies HOURS, MINUTES, SECONDS units,
+                # while Pandas applies all supported units.
+                return pd.Timedelta(delta).isoformat()  # type: ignore[attr-defined]
         return f"{self._value}"
 
 

python/pyspark/sql/tests/test_column.py

@@ -19,12 +19,13 @@
 from enum import Enum
 from itertools import chain
 import datetime
+import unittest
 
 from pyspark.sql import Column, Row
 from pyspark.sql import functions as sf
 from pyspark.sql.types import StructType, StructField, IntegerType, LongType
 from pyspark.errors import AnalysisException, PySparkTypeError, PySparkValueError
-from pyspark.testing.sqlutils import ReusedSQLTestCase
+from pyspark.testing.sqlutils import ReusedSQLTestCase, have_pandas, pandas_requirement_message
 
 
 class ColumnTestsMixin:
@@ -289,6 +290,26 @@ def test_lit_time_representation(self):
         ts = datetime.datetime(2021, 3, 4, 12, 34, 56, 1234)
         self.assertEqual(str(sf.lit(ts)), "Column<'2021-03-04 12:34:56.001234'>")
 
+    @unittest.skipIf(not have_pandas, pandas_requirement_message)
+    def test_lit_delta_representation(self):
+        for delta in [
+            datetime.timedelta(days=1),
+            datetime.timedelta(hours=2),
+            datetime.timedelta(minutes=3),
+            datetime.timedelta(seconds=4),
+            datetime.timedelta(microseconds=5),
+            datetime.timedelta(days=2, hours=21, microseconds=908),
+            datetime.timedelta(days=1, minutes=-3, microseconds=-1001),
+            datetime.timedelta(days=1, hours=2, minutes=3, seconds=4, microseconds=5),
+        ]:
+            import pandas as pd
+
+            # Column<'PT69H0.000908S'> or Column<'P2DT21H0M0.000908S'>
+            s = str(sf.lit(delta))
+
+            # Parse the ISO string representation and compare
+            self.assertTrue(pd.Timedelta(s[8:-2]).to_pytimedelta() == delta)
+
     def test_enum_literals(self):
         class IntEnum(Enum):
             X = 1