Merge pull request #316 from openscm/xarray-update

Fix xarray pin

Author: znicholls

.github/workflows/ci.yaml

@@ -64,8 +64,8 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        python-version: [ "3.9" ]
-        xarray-version: [ "0.16.2", "0.17.0", "0.18.2" ]
+        python-version: [ "3.11" ]
+        xarray-version: [ "0.18.2", "2025.01", "2025.03" ]
 
     steps:
       - name: Checkout repository

.pre-commit-config.yaml

@@ -45,6 +45,6 @@ repos:
     hooks:
       - id: blackdoc
   - repo: https://github.com/python-poetry/poetry
-    rev: '1.4.2'
+    rev: '2.1.2'
     hooks:
       - id: poetry-check

Makefile

@@ -54,7 +54,7 @@ licence-check:  ## Check that licences of the dependencies are suitable
 
 .PHONY: virtual-environment
 virtual-environment:  ## update virtual environment, create a new one if it doesn't already exist
-	poetry lock --no-update
+	poetry lock
 	# Put virtual environments in the project
 	poetry config virtualenvs.in-project true
 	poetry install --all-extras

changelog/316.fix.md

@@ -0,0 +1 @@
+Fixed up incompatibility with latest xarray releases

docs/source/notebooks/ops.py

@@ -217,7 +217,7 @@
 # [Pint](https://pint.readthedocs.io) and some other work, this operation
 # is also unit aware.
 #
-# There are two methods of integration available, `cumtrapz` and `cumsum`.
+# There are two methods of integration available, `cumulative_trapezoid` and `cumsum`.
 # The method that should be used depends on the data you are integrating,
 # specifically whether the data are piecewise linear or piecewise constant.
 #
@@ -226,18 +226,18 @@
 #
 # Other output such as effective radiative forcing, concentrations or decadal
 # timeseries of emissions, represent a  point estimate or an average over a period.
-# These timeseries are therefore piecewise linear and should use the `cumtrapz`
+# These timeseries are therefore piecewise linear and should use the `cumulative_trapezoid`
 # method.
 
 # %%
 with warnings.catch_warnings():
     # Ignore warning about nans in the historical timeseries
     warnings.filterwarnings("ignore", module="scmdata.ops")
 
-    # Radiative Forcings are piecewise-linear so `cumtrapz` should be used
+    # Radiative Forcings are piecewise-linear so `cumulative_trapezoid` should be used
     erf_integral = (
         db_forcing.filter(variable="Effective Radiative Forcing")
-        .cumtrapz()
+        .cumulative_trapezoid()
         .convert_unit("TJ / m^2")
     )
 
@@ -416,13 +416,7 @@
 )
 erf_total_for_shift_rel_to_ref_period_shifted[
     "label"
-] = "rel. to {} - {} (median of {} - {} mean adjusted to {})".format(
-    ref_period[0],
-    ref_period[-1],
-    evaluation_period[0],
-    evaluation_period[-1],
-    target,
-)
+] = f"rel. to {ref_period[0]} - {ref_period[-1]} (median of {evaluation_period[0]} - {evaluation_period[-1]} mean adjusted to {target})"  # noqa: E501
 
 # %%
 pdf = run_append(

pyproject.toml

@@ -27,7 +27,8 @@ classifiers = [
 ]
 
 [tool.poetry.dependencies]
-python = "^3.9"
+# TODO: get rid of this stupid upper level bound
+python = ">=3.9,<4"
 cftime = ">=1.5"
 numpy = "*"
 openscm-units = "*"
@@ -38,10 +39,13 @@ pint-pandas = "*"
 python-dateutil = "*"
 tqdm = "*"
 six = "*"
-xarray = "*"
+xarray = [
+    { version = "*", python = ">=3.10" },
+    { version = "<=2025.03", python = ">=3.9,<3.10" },
+]
 nc-time-axis = { version = ">=1.2.0", optional = true }
 typing-extensions = "*"
-matplotlib = { version = "^3.7.1", optional = true }
+matplotlib = { version = ">=3.7.1", optional = true }
 seaborn = { version = ">=0.12.0", optional = true }
 netCDF4 = { version = "*", optional = true }
 openpyxl = { version = "*", optional = true }
@@ -88,6 +92,9 @@ sphinx-autodocgen = "^1.3"
 jupytext = "^1.14.5"
 sphinx-copybutton = "^0.5.2"
 
+[tool.poetry.requires-plugins]
+poetry-plugin-export = ">=1.8"
+
 [build-system]
 requires = ["poetry-core"]
 build-backend = "poetry.core.masonry.api"
@@ -260,6 +267,7 @@ authorized_licenses = [
     "mit",
     "mit license",
     "Mozilla Public License 2.0 (MPL 2.0)",
+    "psf-2.0",
     "python software foundation",
     "python software foundation license",
     "zpl 2.1",

src/scmdata/groupby.py

@@ -1,6 +1,7 @@
 """
 Functionality for grouping and filtering ScmRun objects
 """
+
 from __future__ import annotations
 
 import warnings
@@ -9,13 +10,20 @@
 
 import numpy as np
 import pandas as pd
+import xarray as xr
 from numpy.typing import NDArray
-from xarray.core import ops
 from xarray.core.common import ImplementsArrayReduce
 
 from scmdata._typing import MetadataValue
 from scmdata.run import BaseScmRun, GenericRun
 
+# Hack central, a better way to do this would be nice
+xr_version_split = xr.__version__.split(".")
+if int(xr_version_split[0]) < 2025 or xr_version_split[1] == "01":  # noqa: PLR2004
+    from xarray.core import ops
+else:
+    from xarray.computation import ops
+
 if TYPE_CHECKING:
     from pandas.core.groupby.generic import DataFrameGroupBy
     from typing_extensions import Concatenate, ParamSpec
@@ -246,7 +254,7 @@ def _combine(
 
     def reduce(
         self,
-        func: Callable[Concatenate[NDArray[np.float_], P], NDArray[np.float_]],
+        func: Callable[Concatenate[NDArray[np.float64], P], NDArray[np.float64]],
         dim: str | Iterable[str] | None = None,
         axis: str | Iterable[int] | None = None,
         *args: P.args,

src/scmdata/ops.py

@@ -5,6 +5,7 @@
 `Pint's Pandas interface <https://pint.readthedocs.io/en/0.13/pint-pandas.html>`_
 to handle unit conversions automatically
 """
+
 import warnings
 
 import numpy as np
@@ -113,10 +114,9 @@ def _perform_op(base, other, op, use_pint_units=True):
                 elif op == "subtract":
                     out.append(base[col] - other[col])
 
-            except KeyError:
-                raise KeyError(  # noqa: TRY200
-                    f"No equivalent in `other` for {list(zip(col_names, col))}"
-                )
+            except KeyError as exc:
+                msg = f"No equivalent in `other` for {list(zip(col_names, col))}"
+                raise KeyError(msg) from exc
 
         elif op == "multiply":
             out.append(base[col] * other[col])
@@ -227,17 +227,17 @@ def subtract(self, other, op_cols, **kwargs):
     >>> fos_minus_afolu.head()
     time                                                                  2010-01-01  2020-01-01
     model     region   scenario  unit       variable
-    idealised World|NH idealised gigatC / a Emissions|CO2|Fossil - AFOLU      -0.001       3.995
-              World|SH idealised gigatC / a Emissions|CO2|Fossil - AFOLU       1.997       5.993
+    idealised World|NH idealised gigatC / yr Emissions|CO2|Fossil - AFOLU      -0.001       3.995
+              World|SH idealised gigatC / yr Emissions|CO2|Fossil - AFOLU       1.997       5.993
 
     >>> nh = start.filter(region="World|NH")
     >>> sh = start.filter(region="World|SH")
     >>> nh_minus_sh = nh.subtract(sh, op_cols={"region": "World|NH - SH"})
     >>> nh_minus_sh.head()
     time                                                               2010-01-01  2020-01-01
     model     region        scenario  unit       variable
-    idealised World|NH - SH idealised gigatC / a Emissions|CO2|Fossil        -2.0        -2.0
-                                      megatC / a Emissions|CO2|AFOLU         -2.0        -2.0
+    idealised World|NH - SH idealised gigatC / yr Emissions|CO2|Fossil        -2.0        -2.0
+                                      megatC / yr Emissions|CO2|AFOLU         -2.0        -2.0
     """
     out = _perform_op(
         prep_for_op(self, op_cols, self.meta.columns, **kwargs),
@@ -336,17 +336,17 @@ def add(self, other, op_cols, **kwargs):
     >>> fos_plus_afolu.head()
     time                                                                  2010-01-01  2020-01-01
     model     region   scenario  unit       variable
-    idealised World|NH idealised gigatC / a Emissions|CO2|Fossil + AFOLU       0.001       4.005
-              World|SH idealised gigatC / a Emissions|CO2|Fossil + AFOLU       2.003       6.007
+    idealised World|NH idealised gigatC / yr Emissions|CO2|Fossil + AFOLU       0.001       4.005
+              World|SH idealised gigatC / yr Emissions|CO2|Fossil + AFOLU       2.003       6.007
 
     >>> nh = start.filter(region="World|NH")
     >>> sh = start.filter(region="World|SH")
     >>> nh_plus_sh = nh.add(sh, op_cols={"region": "World|NH + SH"})
     >>> nh_plus_sh.head()
     time                                                               2010-01-01  2020-01-01
     model     region        scenario  unit       variable
-    idealised World|NH + SH idealised gigatC / a Emissions|CO2|Fossil         2.0        10.0
-                                      megatC / a Emissions|CO2|AFOLU          4.0        12.0
+    idealised World|NH + SH idealised gigatC / yr Emissions|CO2|Fossil         2.0        10.0
+                                      megatC / yr Emissions|CO2|AFOLU          4.0        12.0
     """
     out = _perform_op(
         prep_for_op(self, op_cols, self.meta.columns, **kwargs),
@@ -544,14 +544,14 @@ def divide(self, other, op_cols, **kwargs):
               World|SH idealised MtC / yr Emissions|CO2|AFOLU         3.0         7.0
 
     >>> fos_divide_afolu = fos.divide(
-    ...     afolu, op_cols={"variable": "Emissions|CO2|Fossil / AFOLU"}
+    ...     afolu, op_cols={"variable": "Emissions|CO2|Fossil / yrFOLU"}
     ... )
     >>> # The rows align and the units are handled automatically
     >>> fos_divide_afolu.convert_unit("dimensionless").head()
     time                                                                     2010-01-01  2020-01-01
     model     region   scenario  unit          variable
-    idealised World|NH idealised dimensionless Emissions|CO2|Fossil / AFOLU    0.000000  800.000000
-              World|SH idealised dimensionless Emissions|CO2|Fossil / AFOLU  666.666667  857.142857
+    idealised World|NH idealised dimensionless Emissions|CO2|Fossil / yrFOLU    0.000000  800.000000
+              World|SH idealised dimensionless Emissions|CO2|Fossil / yrFOLU  666.666667  857.142857
 
     >>> nh = start.filter(region="World|NH")
     >>> sh = start.filter(region="World|SH")
@@ -597,7 +597,7 @@ def cumsum(self, out_var=None, check_annual=True):
     arbitrary day/month of the year has not been implemented, if that would be
     useful raise an issue on GitHub.
 
-    If the timeseries are piecewise-linear, :meth:`cumtrapz` should be used instead.
+    If the timeseries are piecewise-linear, :meth:`cumulative_trapezoid` should be used instead.
 
     Parameters
     ----------
@@ -618,7 +618,7 @@ def cumsum(self, out_var=None, check_annual=True):
 
     See Also
     --------
-    :func:`cumtrapz`
+    :func:`cumulative_trapezoid`
 
     Raises
     ------
@@ -677,7 +677,7 @@ def cumsum(self, out_var=None, check_annual=True):
     return out
 
 
-def cumtrapz(self, out_var=None):
+def cumulative_trapezoid(self, out_var=None):
     """
     Integrate with respect to time using the trapezoid rule
 
@@ -743,7 +743,9 @@ def cumtrapz(self, out_var=None):
     # some thinking about unit handling
     _initial = 0.0
     out = pd.DataFrame(
-        scipy.integrate.cumtrapz(y=ts, x=times_in_s, axis=1, initial=_initial)
+        scipy.integrate.cumulative_trapezoid(
+            y=ts, x=times_in_s, axis=1, initial=_initial
+        )
     )
     out.index = ts.index
     out.columns = ts.columns
@@ -790,7 +792,7 @@ def integrate(self, out_var=None):
     See Also
     --------
     :meth:`cumsum`
-    :meth:`cumtrapz`
+    :meth:`cumulative_trapezoid`
 
     Raises
     ------
@@ -805,14 +807,14 @@ def integrate(self, out_var=None):
         data will also contain nans.
     DeprecationWarning
         This function has been deprecated in preference to :meth:`cumsum` and
-        :meth:`cumtrapz`.
+        :meth:`cumulative_trapezoid`.
     """
     warnings.warn(
-        "integrate has been deprecated in preference of cumsum and cumtrapz",
+        "integrate has been deprecated in preference of cumsum and cumulative_trapezoid",
         DeprecationWarning,
     )
 
-    return cumtrapz(self, out_var)
+    return cumulative_trapezoid(self, out_var)
 
 
 def delta_per_delta_time(self, out_var=None):
@@ -1144,7 +1146,7 @@ def inject_ops_methods(cls):
         ("divide", divide),
         ("integrate", integrate),
         ("cumsum", cumsum),
-        ("cumtrapz", cumtrapz),
+        ("cumulative_trapezoid", cumulative_trapezoid),
         ("delta_per_delta_time", delta_per_delta_time),
         ("linear_regression", linear_regression),
         ("linear_regression_gradient", linear_regression_gradient),