Add docstrings and test for strings.convert_durations APIs for pylibcudf

python/cudf/cudf/_lib/string_casting.pyx

@@ -599,12 +599,11 @@ def timedelta2int(Column input_col, dtype, format):
 
     """
     dtype = dtype_to_pylibcudf_type(dtype)
-    cdef string c_timestamp_format = format.encode('UTF-8')
     return Column.from_pylibcudf(
         plc.strings.convert.convert_durations.to_durations(
             input_col.to_pylibcudf(mode="read"),
             dtype,
-            c_timestamp_format
+            format
         )
     )
 
@@ -623,12 +622,10 @@ def int2timedelta(Column input_col, str format):
     A Column with Timedelta represented in string format
 
     """
-
-    cdef string c_duration_format = format.encode('UTF-8')
     return Column.from_pylibcudf(
         plc.strings.convert.convert_durations.from_durations(
             input_col.to_pylibcudf(mode="read"),
-            c_duration_format
+            format
         )
     )
 

python/pylibcudf/pylibcudf/libcudf/strings/convert/convert_durations.pxd

@@ -10,7 +10,7 @@ from pylibcudf.libcudf.types cimport data_type
 cdef extern from "cudf/strings/convert/convert_durations.hpp" namespace \
         "cudf::strings" nogil:
     cdef unique_ptr[column] to_durations(
-        const column_view & strings_col,
+        const column_view & input,
         data_type duration_type,
         const string & format) except +
 

python/pylibcudf/pylibcudf/strings/convert/convert_durations.pxd

@@ -8,10 +8,10 @@ from pylibcudf.types cimport DataType
 cpdef Column to_durations(
     Column input,
     DataType duration_type,
-    const string& format
+    str format
 )
 
 cpdef Column from_durations(
-    Column input,
-    const string& format
+    Column durations,
+    str format=*
 )

python/pylibcudf/pylibcudf/strings/convert/convert_durations.pyx

@@ -15,27 +15,80 @@ from pylibcudf.types import DataType
 cpdef Column to_durations(
     Column input,
     DataType duration_type,
-    const string& format
+    str format
 ):
+    """
+    Returns a new duration column converting a strings column into
+    durations using the provided format pattern.
+
+    For details, see cpp:func:`cudf::strings::to_durations`
+
+    Parameters
+    ----------
+    input : Column
+        Strings instance for this operation.
+
+    duration_type : DataType
+        The duration type used for creating the output column.
+
+    format : str
+        String specifying the duration format in strings.
+
+    Returns
+    -------
+    Column
+        New duration column.
+    """
     cdef unique_ptr[column] c_result
+    cdef string c_format = format.encode()
+
     with nogil:
-        c_result = cpp_convert_durations.to_durations(
-            input.view(),
-            duration_type.c_obj,
-            format
+        c_result = move(
+            cpp_convert_durations.to_durations(
+                input.view(),
+                duration_type.c_obj,
+                c_format
+            )
         )
 
     return Column.from_libcudf(move(c_result))
 
 cpdef Column from_durations(
-    Column input,
-    const string& format
+    Column durations,
+    str format=None
 ):
+    """
+    Returns a new strings column converting a duration column into
+    strings using the provided format pattern.
+
+    For details, see cpp:func:`cudf::strings::from_durations`
+
+    Parameters
+    ----------
+    durations : Column
+        Duration values to convert.
+
+    format : str
+        The string specifying output format.
+        Default format is ""%D days %H:%M:%S".
+
+    Returns
+    -------
+    Column
+        New strings column with formatted durations.
+    """
     cdef unique_ptr[column] c_result
+
+    if format is None:
+        format = "%D days %H:%M:%S"
+    cdef string c_format = format.encode()
+
     with nogil:
-        c_result = cpp_convert_durations.from_durations(
-            input.view(),
-            format
+        c_result = move(
+            cpp_convert_durations.from_durations(
+                durations.view(),
+                c_format
+            )
         )
 
     return Column.from_libcudf(move(c_result))

python/pylibcudf/pylibcudf/tests/test_string_convert.py

@@ -1,7 +1,5 @@
 # Copyright (c) 2024, NVIDIA CORPORATION.
 
-from datetime import datetime
-
 import pyarrow as pa
 import pylibcudf as plc
 import pytest
@@ -21,39 +19,16 @@ def timestamp_type(request):
     return request.param
 
 
-@pytest.fixture(
-    scope="module",
-    params=[
-        pa.duration("ns"),
-        pa.duration("us"),
-        pa.duration("ms"),
-        pa.duration("s"),
-    ],
-)
-def duration_type(request):
-    return request.param
-
-
 @pytest.fixture(scope="module")
 def pa_timestamp_col():
     return pa.array(["2011-01-01", "2011-01-02", "2011-01-03"])
 
 
-@pytest.fixture(scope="module")
-def pa_duration_col():
-    return pa.array(["05:20:25"])
-
-
 @pytest.fixture(scope="module")
 def plc_timestamp_col(pa_timestamp_col):
     return plc.interop.from_arrow(pa_timestamp_col)
 
 
-@pytest.fixture(scope="module")
-def plc_duration_col(pa_duration_col):
-    return plc.interop.from_arrow(pa_duration_col)
-
-
 @pytest.mark.parametrize("format", ["%Y-%m-%d"])
 def test_to_datetime(
     pa_timestamp_col, plc_timestamp_col, timestamp_type, format
@@ -65,21 +40,3 @@ def test_to_datetime(
         format.encode(),
     )
     assert_column_eq(expect, got)
-
-
-@pytest.mark.parametrize("format", ["%H:%M:%S"])
-def test_to_duration(pa_duration_col, plc_duration_col, duration_type, format):
-    def to_timedelta(duration_str):
-        date = datetime.strptime(duration_str, format)
-        return date - datetime(1900, 1, 1)  # "%H:%M:%S" zero date
-
-    expect = pa.array([to_timedelta(d.as_py()) for d in pa_duration_col]).cast(
-        duration_type
-    )
-
-    got = plc.strings.convert.convert_durations.to_durations(
-        plc_duration_col,
-        plc.interop.from_arrow(duration_type),
-        format.encode(),
-    )
-    assert_column_eq(expect, got)

python/pylibcudf/pylibcudf/tests/test_string_convert_durations.py

@@ -0,0 +1,61 @@
+# Copyright (c) 2024, NVIDIA CORPORATION.
+
+from datetime import datetime, timedelta
+
+import pyarrow as pa
+import pylibcudf as plc
+import pytest
+from utils import assert_column_eq
+
+
+@pytest.fixture(
+    params=[
+        pa.duration("ns"),
+        pa.duration("us"),
+        pa.duration("ms"),
+        pa.duration("s"),
+    ],
+)
+def duration_type(request):
+    return request.param
+
+
+@pytest.fixture
+def pa_duration_col():
+    return pa.array(["05:20:25"])
+
+
+@pytest.fixture
+def plc_duration_col(pa_duration_col):
+    return plc.interop.from_arrow(pa_duration_col)
+
+
+def test_to_duration(pa_duration_col, plc_duration_col, duration_type):
+    format = "%H:%M:%S"
+
+    def to_timedelta(duration_str):
+        date = datetime.strptime(duration_str, format)
+        return date - datetime(1900, 1, 1)  # "%H:%M:%S" zero date
+
+    expect = pa.array([to_timedelta(d.as_py()) for d in pa_duration_col]).cast(
+        duration_type
+    )
+
+    got = plc.strings.convert.convert_durations.to_durations(
+        plc_duration_col,
+        plc.interop.from_arrow(duration_type),
+        format,
+    )
+    assert_column_eq(expect, got)
+
+
+@pytest.mark.parametrize("format", [None, "%D days %H:%M:%S"])
+def test_from_durations(format):
+    pa_array = pa.array(
+        [timedelta(days=1, hours=1, minutes=1, seconds=1), None]
+    )
+    result = plc.strings.convert.convert_durations.from_durations(
+        plc.interop.from_arrow(pa_array), format
+    )
+    expected = pa.array(["1 days 01:01:01", None])
+    assert_column_eq(result, expected)