Preserve legacy Xarray data_vars settings (#325)

Author: tomvothecoder

e3sm_to_cmip/__init__.py

@@ -24,3 +24,14 @@
 # Path to the directory where MPAS handlers are defined.
 # TODO: MPAS handlers should eventually be refactored to `handlers.yaml` too.
 MPAS_HANDLER_DIR_PATH = f"{ROOT_HANDLERS_DIR}/mpas_vars"
+
+# Settings to preserve legacy Xarray behavior when opening multiple files with
+# xarray.open_mfdataset(). These settings are automatically set when using
+# xcdat>=0.10.1.
+# See https://xarray.pydata.org/en/stable/user-guide/io.html#combining-multiple-files
+# and https://xarray.pydata.org/en/stable/whats-new.html#id14
+LEGACY_XARRAY_MERGE_SETTINGS = {
+    # "override", "exact" are the new defaults as of Xarray v2025.08.0
+    "compat": "no_conflicts",
+    "join": "outer",
+}

e3sm_to_cmip/cmor_handlers/handler.py

@@ -9,6 +9,7 @@
 import xcdat as xc
 import yaml
 
+from e3sm_to_cmip import LEGACY_XARRAY_MERGE_SETTINGS
 from e3sm_to_cmip._logger import _setup_child_logger
 from e3sm_to_cmip.cmor_handlers import FILL_VALUE, _formulas
 from e3sm_to_cmip.util import _get_table_for_non_monthly_freq
@@ -390,12 +391,7 @@ def _get_mfdataset(
             combine="nested",
             data_vars="minimal",
             coords="minimal",
-            # NOTE: Preserve legacy Xarray behavior by setting compat="no_conflicts"
-            # and join="outer" ("override" and "exact" are the new defaults as of
-            # Xarray v2025.08.0). These settings are automatically set when
-            # using xcdat>=0.10.1.
-            compat="no_conflicts",
-            join="outer",
+            **LEGACY_XARRAY_MERGE_SETTINGS,
         )
 
         # If the output CMIP variable has an alternative time dimension name (e.g.,

e3sm_to_cmip/cmor_handlers/mpas_vars/pbo.py

@@ -4,7 +4,7 @@
 
 import xarray
 
-from e3sm_to_cmip import mpas, util
+from e3sm_to_cmip import LEGACY_XARRAY_MERGE_SETTINGS, mpas, util
 from e3sm_to_cmip._logger import _setup_child_logger
 from e3sm_to_cmip.util import print_message
 
@@ -88,11 +88,11 @@ def handle(infiles, tables, user_input_path, cmor_log_dir, **kwargs):
 
     ds = mpas.remap(ds, "mpasocean", mappingFileName)
 
-    # NOTE: Preserve legacy Xarray behavior by setting compat="no_conflicts"
-    # and join="outer" ("override" and "exact" are the new defaults as of
-    # Xarray v2025.08.0)
     with xarray.open_mfdataset(
-        pslFileNames, concat_dim="time", compat="no_conflicts", join="outer"
+        pslFileNames,
+        concat_dim="time",
+        data_vars="all",
+        **LEGACY_XARRAY_MERGE_SETTINGS,  # type: ignore
     ) as dsIn:
         ds[VAR_NAME] = ds[VAR_NAME] + dsIn.PSL.values
 

e3sm_to_cmip/cmor_handlers/mpas_vars/pso.py

@@ -4,7 +4,7 @@
 
 import xarray
 
-from e3sm_to_cmip import mpas, util
+from e3sm_to_cmip import LEGACY_XARRAY_MERGE_SETTINGS, mpas, util
 from e3sm_to_cmip._logger import _setup_child_logger
 from e3sm_to_cmip.util import print_message
 
@@ -85,11 +85,11 @@ def handle(infiles, tables, user_input_path, cmor_log_dir, **kwargs):
 
     ds = mpas.remap(ds, "mpasocean", mappingFileName)
 
-    # NOTE: Preserve legacy Xarray behavior by setting compat="no_conflicts"
-    # and join="outer" ("override" and "exact" are the new defaults as of
-    # Xarray v2025.08.0)
     with xarray.open_mfdataset(
-        pslFileNames, concat_dim="time", compat="no_conflicts", join="exact"
+        pslFileNames,
+        concat_dim="time",
+        data_vars="all",
+        **LEGACY_XARRAY_MERGE_SETTINGS,  # type: ignore
     ) as dsIn:
         ds[VAR_NAME] = ds[VAR_NAME] + dsIn.PSL.values
 

e3sm_to_cmip/mpas.py

@@ -20,6 +20,7 @@
 import xarray
 from dask.diagnostics import ProgressBar
 
+from e3sm_to_cmip import LEGACY_XARRAY_MERGE_SETTINGS
 from e3sm_to_cmip._logger import _setup_child_logger
 
 logger = _setup_child_logger(__name__)
@@ -77,13 +78,9 @@ def remap_seaice_sgs(inFileName, outFileName, mappingFileName, renorm_threshold=
     # With  data_vars='minimal', only data variables in which the dimension already appears are included.
     ds_out_all = xarray.open_mfdataset(
         f"{outFilePath}/temp_out*nc",
-        data_vars="minimal",
         decode_times=False,
-        # NOTE: Preserve legacy Xarray behavior by setting compat="no_conflicts"
-        # and join="outer" ("override" and "exact" are the new defaults as of
-        # Xarray v2025.08.0)
-        compat="no_conflicts",
-        join="outer",
+        data_vars="minimal",
+        **LEGACY_XARRAY_MERGE_SETTINGS,  # type: ignore
     )
     ds_out_all = ds_out_all.drop("timeMonthly_avg_iceAreaCell")
     ds_out_all.to_netcdf(outFileName)
@@ -333,11 +330,8 @@ def open_mfdataset(
         concat_dim="Time",
         mask_and_scale=False,
         chunks=chunks,
-        # NOTE: Preserve legacy Xarray behavior by setting compat="no_conflicts"
-        # and join="outer" ("override" and "exact" are the new defaults as of
-        # Xarray v2025.08.0).
-        compat="no_conflicts",
-        join="outer",
+        data_vars="all",
+        **LEGACY_XARRAY_MERGE_SETTINGS,  # type: ignore
     )
 
     if variableList is not None: