Add .load() to _get_dataset_with_source_vars() to improve perform…

…ance

auxiliary_tools/cdat_regression_testing/667-arm_diags/debug/667_debug_perf.py

@@ -0,0 +1,104 @@
+# %%
+import timeit
+
+setup_code = """
+import xarray as xr
+
+AIR_DENS = 1.225  # standard air density 1.225kg/m3
+
+a1 = xr.open_dataarray("qa/667-arms-diags/a1.nc")
+a2 = xr.open_dataarray("qa/667-arms-diags/a2.nc")
+a3 = xr.open_dataarray("qa/667-arms-diags/a3.nc")
+"""
+
+setup_code2 = """
+import xarray as xr
+
+AIR_DENS = 1.225  # standard air density 1.225kg/m3
+
+a1 = xr.open_dataarray("qa/667-arms-diags/a1.nc")
+a2 = xr.open_dataarray("qa/667-arms-diags/a2.nc")
+a3 = xr.open_dataarray("qa/667-arms-diags/a3.nc")
+
+a1.load(scheduler="sync")
+a2.load(scheduler="sync")
+a3.load(scheduler="sync")
+"""
+
+setup_code3 = """
+import xarray as xr
+
+AIR_DENS = 1.225  # standard air density 1.225kg/m3
+
+a1_chunked = xr.open_dataarray("qa/667-arms-diags/a1.nc", chunks={"time": "auto"})
+a2_chunked = xr.open_dataarray("qa/667-arms-diags/a2.nc", chunks={"time": "auto"})
+a3_chunked = xr.open_dataarray("qa/667-arms-diags/a3.nc", chunks={"time": "auto"})
+"""
+
+code_statement1 = """
+with xr.set_options(keep_attrs=True):
+    var = (a1 + a2 + a3) * AIR_DENS / 1e6
+"""
+
+
+code_statement2 = """
+with xr.set_options(keep_attrs=True):
+    var = (a1_chunked + a2_chunked + a3_chunked) * AIR_DENS / 1e6
+"""
+
+code_statement3 = """
+var_data = (a1.values + a2.values + a3.values) * AIR_DENS / 1e6
+var_new = xr.DataArray(
+    var_data,
+    dims=a1.dims,
+    coords=a1.coords,
+    name="a_num",
+    attrs={"units": "/cm3", "long_name": "aerosol number concentration"},
+)
+"""
+
+code_statement4 = """
+var_data2 = (a1.data + a2.data + a3.data) * AIR_DENS / 1e6
+var_new2 = xr.DataArray(
+    name="a_num", data=var_data2, dims=a1.dims, coords=a1.coords, attrs=a1.attrs
+)
+var_new2.attrs.update(
+    {"units": "/cm3", "long_name": "aerosol number concentration"}
+)
+"""
+
+
+def run_timeit(code_statement: str, setup_code: str) -> float:
+    elapsed_time = timeit.repeat(
+        code_statement, setup=setup_code, globals=globals(), repeat=3, number=1
+    )
+
+    return min(elapsed_time)
+
+
+elapsed_time_xarray = run_timeit(code_statement1, setup_code)
+print(f"1. Elapsed time (Xarray non-chunked): {elapsed_time_xarray} seconds")
+
+elapsed_time_xarray_load = run_timeit(code_statement1, setup_code2)
+print(
+    f"2. Elapsed time (Xarray non-chunked with .load()): {elapsed_time_xarray_load} seconds"
+)
+elapsed_time_xarray_chunked = run_timeit(code_statement2, setup_code3)
+print(f"3. Elapsed time (Xarray chunked): {elapsed_time_xarray_chunked} seconds")
+
+elapsed_time_numpy_1 = run_timeit(code_statement3, setup_code)
+print(f"4. Elapsed time (numpy .values): {elapsed_time_numpy_1} seconds")
+
+elapsed_time_numpy_2 = run_timeit(code_statement4, setup_code)
+print(f"5. Elapsed time (numpy .data): {elapsed_time_numpy_2} seconds")
+
+
+"""
+Results
+----------
+1. Elapsed time (Xarray non-chunked): 6.540755605790764 seconds
+2. Elapsed time (Xarray non-chunked with .load()): 0.17097265785560012 seconds
+3. Elapsed time (Xarray chunked): 0.1452920027077198 seconds
+4. Elapsed time (numpy .values): 6.418793010059744 seconds
+5. Elapsed time (numpy .data): 7.334999438840896 seconds
+"""

e3sm_diags/driver/utils/dataset_xr.py

@@ -417,7 +417,8 @@ def get_climo_dataset(self, var: str, season: ClimoFreq) -> xr.Dataset:
         elif self.is_time_series:
             ds = self.get_time_series_dataset(var)
             ds_climo = climo(ds, self.var, season).to_dataset()
-            return ds_climo
+
+        return ds_climo
 
     def _get_climo_dataset(self, season: str) -> xr.Dataset:
         """Get the climatology dataset for the variable and season.
@@ -1054,6 +1055,7 @@ def _get_dataset_with_source_vars(self, vars_to_get: Tuple[str, ...]) -> xr.Data
 
         ds = xr.merge(datasets)
         ds = squeeze_time_dim(ds)
+        ds.load(scheduler="sync")
 
         return ds