Pandas 3.0

.github/workflows/pytest.yml

@@ -16,7 +16,7 @@ jobs:
       fail-fast: false
       matrix:
         os: ['ubuntu-latest']
-        python-version: ['3.10', '3.11', '3.12', '3.13', '3.14']
+        python-version: ['3.11', '3.12', '3.13', '3.14']
     steps:
       - uses: actions/checkout@v6
       - name: Install uv

.github/workflows/pytest_upstream_nightly.yml

@@ -12,7 +12,7 @@ jobs:
       fail-fast: false
       matrix:
         os: ['ubuntu-latest']
-        python-version: ['3.10', '3.11', '3.12', '3.13', '3.14']
+        python-version: ['3.11', '3.12', '3.13', '3.14']
     steps:
       - uses: actions/checkout@v6
       - name: Install uv

chainladder/__init__.py

@@ -4,7 +4,7 @@ class Options:
     ARRAY_BACKEND = "numpy"
     AUTO_SPARSE = True
     ARRAY_PRIORITY = ["dask", "sparse", "cupy", "numpy"]
-    ULT_VAL = "2261-12-31 23:59:59.999999999"
+    ULT_VAL = "2261-12-31 23:59:59.999999"
 
     @classmethod
     def get_option(cls, option=None):
@@ -18,7 +18,7 @@ def reset_option(self):
         self.set_option('ARRAY_BACKEND', "numpy")
         self.set_option('AUTO_SPARSE', True)
         self.set_option('ARRAY_PRIORITY', ["dask", "sparse", "cupy", "numpy"])
-        self.set_option('ULT_VAL', "2261-12-31 23:59:59.999999999")
+        self.set_option('ULT_VAL', "2261-12-31 23:59:59.999999")
 
     def describe_option(self):
         pass

chainladder/core/base.py

@@ -540,12 +540,12 @@ def valuation(self):
         ddim_arr = ddims - ddims[0]
         origin = np.minimum(self.odims, np.datetime64(self.valuation_date))
         val_array = origin.astype("datetime64[M]") + np.timedelta64(ddims[0], "M")
-        val_array = val_array.astype("datetime64[ns]") - np.timedelta64(1, "ns")
+        val_array = val_array.astype("datetime64[us]") - np.timedelta64(1, "us")
         val_array = val_array[:, None]
         s = slice(None, -1) if ddims[-1] == 9999 else slice(None, None)
         val_array = (
             val_array.astype("datetime64[M]") + ddim_arr[s][None, :] + 1
-        ).astype("datetime64[ns]") - np.timedelta64(1, "ns")
+        ).astype("datetime64[us]") - np.timedelta64(1, "us")
         if ddims[-1] == 9999:
             ult = np.repeat(np.datetime64(options.ULT_VAL), val_array.shape[0])[:, None]
             val_array = np.concatenate(

chainladder/core/pandas.py

@@ -505,7 +505,7 @@ def agg_func(
         # If axis is development, set the ddims to be the valuation date.
         if axis == 3 and obj.values.shape[axis] == 1 and len(obj.ddims) > 1:
             obj.ddims = pd.DatetimeIndex(
-                [self.valuation_date], dtype="datetime64[ns]", freq=None
+                [self.valuation_date], dtype="datetime64[us]", freq=None
             )
         obj._set_slicers()
         if auto_sparse:

chainladder/core/tests/test_triangle.py

@@ -3,7 +3,8 @@
 import numpy as np
 import pytest
 import io
-from datetime import datetime
+
+from io import StringIO
 
 try:
     from IPython.core.display import HTML
@@ -13,7 +14,7 @@
 
 def test_repr(raa):
     np.testing.assert_array_equal(
-        pd.read_html(raa._repr_html_())[0].set_index("Unnamed: 0").values,
+        pd.read_html(StringIO(raa._repr_html_()))[0].set_index("Unnamed: 0").values,
         raa.to_frame(origin_as_datetime=False).values,
     )
 
@@ -317,7 +318,7 @@ def test_groupby_axis1(clrd, prism):
     clrd = clrd.sum("origin").sum("development")
     groups = [i.find("Loss") >= 0 for i in clrd.columns]
     assert np.all(
-        clrd.to_frame(origin_as_datetime=False).groupby(groups, axis=1).sum()
+        clrd.to_frame(origin_as_datetime=False).T.groupby(groups).sum().T
         == clrd.groupby(groups, axis=1).sum().to_frame(origin_as_datetime=False)
     )
     assert np.all(
@@ -908,7 +909,7 @@ def test_single_valuation_date_preserves_exact_date():
     )
 
     # Valuation date should be end of October 2025, not converted to a fiscal year
-    assert triangle.valuation_date == pd.Timestamp('2025-10-31 23:59:59.999999999')
+    assert triangle.valuation_date == pd.Timestamp('2025-10-31 23:59:59.999999')
     assert triangle.development_grain == 'M'
     assert int(triangle.valuation_date.strftime('%Y%m')) == 202510
 def test_OXDX_triangle():

chainladder/core/triangle.py

@@ -577,7 +577,7 @@ def _split_ult(
         if (
             development
             and len(development) == 1
-            and data[development[0]].dtype == "<M8[ns]"
+            and pd.api.types.is_datetime64_dtype(data[development[0]])
         ):
             u = data[data[development[0]] == options.ULT_VAL].copy()
             if len(u) > 0 and len(u) != len(data):

chainladder/development/base.py

@@ -9,6 +9,7 @@
 from chainladder.utils.utility_functions import num_to_nan
 from chainladder.core.io import EstimatorIO
 from chainladder.core.common import Common
+from pandas.api.types import is_string_dtype
 
 
 class DevelopmentBase(BaseEstimator, TransformerMixin, EstimatorIO, Common):
@@ -421,7 +422,7 @@ def _drop_func(self, factor):
         # accommodate ldf triangle as factor, where the dimensions are '12-24'
         dev_list = (
             factor.development.str.split("-", expand=True)[0]
-            if factor.development.dtype == object
+            if is_string_dtype(factor.development)
             else factor.development.astype("string")
         )
         # create ndarray of drop_list for further operation in numpy

chainladder/utils/tests/test_utilities.py

@@ -7,7 +7,6 @@
 
 from pathlib import Path
 
-
 def test_non_vertical_line():
     true_olf = (
         1.20
@@ -161,7 +160,6 @@ def test_read_csv_single(raa):
     # Test the read_csv function for a single dimensional input.
 
     # Read in the csv file.
-    from pathlib import Path
     raa_csv_path = Path(__file__).parent.parent / "data" / "raa.csv"
 
     assert raa == cl.read_csv(

chainladder/utils/utility_functions.py

@@ -504,7 +504,7 @@ def read_json(json_str, array_backend=None):
                 setattr(getattr(tri, k), "development_grain", tri.development_grain)
         if "dfs" in json_dict.keys():
             for k, v in json_dict["dfs"].items():
-                df = pd.read_json(v)
+                df = pd.read_json(StringIO(v))
                 if len(df.columns) == 1:
                     df = df.iloc[:, 0]
                 setattr(tri, k, df)
@@ -708,7 +708,7 @@ def concat(
     if ignore_index and axis == 0:
         out.key_labels = ["Index"]
     out.valuation_date = pd.Series([obj.valuation_date for obj in objs]).max()
-    if out.ddims.dtype == "datetime64[ns]" and type(out.ddims) == np.ndarray:
+    if out.ddims.dtype == "datetime64[us]" and type(out.ddims) == np.ndarray:
         out.ddims = pd.DatetimeIndex(out.ddims)
     out._set_slicers()
     if sort:

pyproject.toml

@@ -14,23 +14,22 @@ maintainers = [
 description = "Chainladder Package - P&C Loss Reserving package"
 readme = "README.rst"
 license = {text = "MPL-2.0"}
-requires-python = ">=3.10"
+requires-python = ">=3.11"
 classifiers = [
     "Development Status :: 4 - Beta",
     "Intended Audience :: Developers",
     "Intended Audience :: Science/Research",
     "License :: OSI Approved :: Mozilla Public License 2.0 (MPL 2.0)",
     "Operating System :: OS Independent",
     "Programming Language :: Python :: 3",
-    "Programming Language :: Python :: 3.10",
     "Programming Language :: Python :: 3.11",
     "Programming Language :: Python :: 3.12",
     "Programming Language :: Python :: 3.13",
     "Programming Language :: Python :: 3.14",
 ]
 keywords = ["actuarial", "reserving", "insurance", "chainladder", "IBNR"]
 dependencies = [
-    "pandas >=2.0, <3.0",
+    "pandas >=3",
     "scikit-learn>1.4.2",
     "numba>0.54",
     "sparse>=0.9",