Merge pull request #4 from pfnet-research/masanori/support-latest-python

support python 3.12, 3.13

Author: masanorihirano

.github/workflows/ci-python.yaml

@@ -24,7 +24,15 @@ jobs:
       max-parallel: 5
       matrix:
         platform: [ubuntu-latest, macos-latest, windows-latest]
-        python-version: [3.9, "3.10", "3.11"] # TODO (masanori): add 3.12, 3.13 when it is available
+        python-version: [3.9, "3.10", "3.11", "3.12", "3.13"]
+        pandas-and-numpy-version:
+          - '"pandas>=1.0.0,<2.0.0" "numpy>=1.0.0,<2.0.0"'
+          - '"pandas>=2.0.0" "numpy>=2.0.0"'
+        exclude:
+          - python-version: "3.12"
+            pandas-and-numpy-version: '"pandas>=1.0.0,<2.0.0" "numpy>=1.0.0,<2.0.0"'
+          - python-version: "3.13"
+            pandas-and-numpy-version: '"pandas>=1.0.0,<2.0.0" "numpy>=1.0.0,<2.0.0"'
 
     # Steps represent a sequence of tasks that will be executed as part of the job
     steps:
@@ -41,6 +49,7 @@ jobs:
       - name: Install dependencies
         run: |
           uv sync --dev
+          uv run pip install -U ${{ matrix.pandas-and-numpy-version }}
       - name: Test
         run: |
           uv run black --check --diff --quiet .

README.md

@@ -1,5 +1,5 @@
 # qfeval-data
-[![python](https://img.shields.io/badge/python-%3E=3.9,%3C3.12-blue.svg)](https://pypi.org/project/qfeval_data/)
+[![python](https://img.shields.io/badge/python-%3E=3.9-blue.svg)](https://pypi.org/project/qfeval_data/)
 [![pypi](https://img.shields.io/pypi/v/qfeval_data.svg)](https://pypi.org/project/qfeval_data/)
 [![CI](https://github.com/pfnet-research/qfeval-data/actions/workflows/ci-python.yaml/badge.svg)](https://github.com/pfnet-research/qfeval-data/actions/workflows/ci-python.yaml)
 [![codecov](https://codecov.io/gh/pfnet-research/qfeval-data/graph/badge.svg?token=5A02B1JV7V)](https://codecov.io/gh/pfnet-research/qfeval-data)

pyproject.toml

@@ -4,18 +4,17 @@ version = "0.1.0"
 description = "Data structures for quantitative finance"
 readme = "README.md"
 # Don't forget to fix the badge in README.md after changing requires-python
-# TODO (masanori): Support python 3.12, 3.13 (pandas 2.0.0+). also update CI
-requires-python = ">=3.9,<3.12"
+requires-python = ">=3.9"
 dependencies = [
-    "pandas>=1.5.0,<2.0.0",
-    "numpy>=1.24.0,<2.0.0",
+    "pandas",
+    "numpy",
     "torch",
     "qfeval-functions",
 ]
 
 [project.optional-dependencies]
 plot = [
-  "matplotlib>=3.8.3"
+  "matplotlib"
 ]
 
 [dependency-groups]

qfeval_data/data.py

@@ -1497,6 +1497,25 @@ def __set_xticks(self, ax: matplotlib.axes.Axes) -> None:
                 # first candlestick of a month would start with its previous
                 # month.
                 unit, _ = np.datetime_data(delta)
+                unit = typing.cast(
+                    typing.Literal[
+                        "Y",
+                        "M",
+                        "W",
+                        "D",
+                        "h",
+                        "m",
+                        "s",
+                        "ms",
+                        "us",
+                        "μs",
+                        "ns",
+                        "ps",
+                        "fs",
+                        "as",
+                    ],
+                    unit,
+                )
                 label_ticks = util.ceil_time(
                     timestamps[tick_indices], np.timedelta64(1, unit)
                 )
@@ -1521,8 +1540,14 @@ def __xtick_deltas(
             "m": [(30, 10), (20, 5), (10, 5), (5, 1), (2, 1), (1, 1)],
             "s": [(30, 10), (20, 5), (10, 5), (5, 1), (2, 1), (1, 1)],
         }
-        last_deltas = (np.timedelta64(100, "Y"), np.timedelta64(100, "Y"))
+        last_deltas: typing.Tuple[
+            np.timedelta64,
+            np.timedelta64,
+        ] = (np.timedelta64(100, "Y"), np.timedelta64(100, "Y"))
         for unit, sizes in deltas.items():
+            unit = typing.cast(
+                typing.Literal["Y", "M", "D", "h", "m", "s"], unit
+            )
             for major, minor in sizes:
                 major_delta = np.timedelta64(major, unit)
                 minor_delta = np.timedelta64(minor, unit)
@@ -1642,8 +1667,8 @@ def downsample(
         dest_unit = np.datetime_data(delta)[0]
         timestamps = util.floor_time(self.timestamps, delta, origin, offset)
         timestamps = timestamps.astype(f"datetime64[{dest_unit}]")
-        timestamps, group_ids = np.unique(timestamps, return_inverse=True)
-        group_ids = torch.tensor(group_ids, device=self.device)
+        timestamps, _group_ids = np.unique(timestamps, return_inverse=True)
+        group_ids = torch.tensor(_group_ids, device=self.device)
 
         tensors = {}
         for k, v in self.__tensors.items():

qfeval_data/util.py

@@ -169,7 +169,27 @@ def floor_time(
 
 def time_origin(d: np.timedelta64) -> np.datetime64:
     unit, _ = np.datetime_data(d)
+    unit = typing.cast(
+        typing.Literal[
+            "Y",
+            "M",
+            "W",
+            "D",
+            "h",
+            "m",
+            "s",
+            "ms",
+            "us",
+            "μs",
+            "ns",
+            "ps",
+            "fs",
+            "as",
+        ],
+        unit,
+    )
     if unit in ("M", "Y"):
+        unit = typing.cast(typing.Literal["M", "Y"], unit)
         # qfeval uses 1000-01-01 as the datetime origin for monthly/yearly
         # ticks so as to show multiples of 10/100/1000 years for a long term
         # x-axis.

tests/test_data.py

@@ -495,7 +495,7 @@ def test_to_matrix(self) -> None:
         df = pd.read_csv(io.StringIO(csv))
         data = Data.from_dataframe(df)
         actual_df = data.weight.to_matrix()
-        assert actual_df.to_csv(line_terminator="\n") == expected
+        assert actual_df.to_csv(lineterminator="\n") == expected
 
     def test_missing_to_matrix(self) -> None:
         csv = """timestamp,symbol,return,weight
@@ -507,7 +507,7 @@ def test_missing_to_matrix(self) -> None:
         df = pd.read_csv(io.StringIO(csv))
         data = Data.from_dataframe(df)
         actual_df = data.weight.to_matrix()
-        assert actual_df.to_csv(line_terminator="\n") == expected
+        assert actual_df.to_csv(lineterminator="\n") == expected
 
     def test_rename(self) -> None:
         data = Data.from_dataframe(self.create_simple_dataframe())
@@ -834,7 +834,7 @@ def test_downsample(self) -> None:
                 np.timedelta64(2, "D"), origin=np.datetime64("2018-01-01")
             )
             .to_dataframe()
-            .to_csv(index=False, line_terminator="\n")
+            .to_csv(index=False, lineterminator="\n")
             == "timestamp,symbol,open,high,low,close\n"
             + "2018-01-01,AAPL,154.0,162.0,154.0,160.0\n"
             + "2018-01-03,AAPL,159.0,161.0,154.0,159.0\n"
@@ -849,7 +849,7 @@ def test_downsample(self) -> None:
         assert (
             data.weekly(origin=np.datetime64("2018-01-01"))
             .to_dataframe()
-            .to_csv(index=False, line_terminator="\n")
+            .to_csv(index=False, lineterminator="\n")
             == "timestamp,symbol,open,high,low,close\n"
             + "2018-01-01,AAPL,154.0,163.0,154.0,163.0\n"
             + "2018-01-08,AAPL,161.0,177.0,160.0,162.0\n"