From 8a55556582cfe1a926bb74b28532d8e4710596b7 Mon Sep 17 00:00:00 2001
From: "Benjamin T. Schwertfeger" <contact@b-schwertfeger.de>
Date: Fri, 17 May 2024 16:25:32 +0200
Subject: [PATCH] Resolve "The behavior for data sets with different temporal
 resolution are not uniform" (#100)

---
 README.md               |  4 ++++
 cmethods/core.py        | 16 ++++++++--------
 doc/getting_started.rst | 13 +++++++------
 doc/introduction.rst    |  5 +++++
 4 files changed, 24 insertions(+), 14 deletions(-)

diff --git a/README.md b/README.md
index 10c7aff..b7bdb86 100644
--- a/README.md
+++ b/README.md
@@ -108,6 +108,10 @@ Please refer to the official documentation for more information about these
 methods as well as sample scripts:
 https://python-cmethods.readthedocs.io/en/stable/
 
+## Best Practices and important Notes
+
+- The training data should have the same temporal resolution.
+
 - Except for the variance scaling, all methods can be applied on stochastic and
   non-stochastic climate variables. Variance scaling can only be applied on
   non-stochastic climate variables.
diff --git a/cmethods/core.py b/cmethods/core.py
index eb4c2b5..a08c4a9 100644
--- a/cmethods/core.py
+++ b/cmethods/core.py
@@ -60,7 +60,7 @@ def apply_ufunc(
                 'input_core_dims must have three key-value pairs like: {"obs": "time", "simh": "time", "simp": "time"}',
             )
 
-        input_core_dims = kwargs["input_core_dims"]
+        input_core_dims = kwargs.pop("input_core_dims")
     else:
         input_core_dims = {"obs": "time", "simh": "time", "simp": "time"}
 
@@ -68,13 +68,13 @@ def apply_ufunc(
         __METHODS_FUNC__[method],
         obs,
         simh,
-        # Need to spoof a fake time axis since 'time' coord on full dataset is different
-        # than 'time' coord on training dataset.
+        # Need to spoof a fake time axis since 'time' coord on full dataset is
+        # different than 'time' coord on training dataset.
         simp.rename({input_core_dims["simp"]: "__t_simp__"}),
         dask="parallelized",
         vectorize=True,
-        # This will vectorize over the time dimension, so will submit each grid cell
-        # independently
+        # This will vectorize over the time dimension, so will submit each grid
+        # cell independently
         input_core_dims=[
             [input_core_dims["obs"]],
             [input_core_dims["simh"]],
@@ -89,9 +89,9 @@ def apply_ufunc(
     # Rename to proper coordinate name.
     result = result.rename({"__t_simp__": input_core_dims["simp"]})
 
-    # ufunc will put the core dimension to the end (time), so want to preserve original
-    # order where time is commonly first.
-    return result.transpose(*obs.dims)
+    # ufunc will put the core dimension to the end (time), so want to preserve
+    # original order where time is commonly first.
+    return result.transpose(*obs.rename({input_core_dims["obs"]: input_core_dims["simp"]}).dims)
 
 
 def adjust(
diff --git a/doc/getting_started.rst b/doc/getting_started.rst
index 51d1faf..a8b8337 100644
--- a/doc/getting_started.rst
+++ b/doc/getting_started.rst
@@ -149,16 +149,17 @@ dimensions of ``obs`` and ``simp`` have the length, but the time dimension of
     import xarray as xr
 
     obs = xr.open_dataset("examples/input_data/observations.nc")["tas"]
-    simp = xr.open_dataset("examples/input_data/control.nc")["tas"]
     simh = simp.copy(deep=True)[3650:]
+    simp = xr.open_dataset("examples/input_data/control.nc")["tas"]
 
     bc = adjust(
         method="quantile_mapping",
         obs=obs,
         simh=simh.rename({"time": "t_simh"}),
-        simp=simh,
+        simp=simp,
         kind="+",
-        input_core_dims={"obs": "time", "simh": "t_simh", "simp": "time"}
+        input_core_dims={"obs": "time", "simh": "t_simh", "simp": "time"},
+        n_quantiles=100,
     )
 
 In case you are applying a scaling based technique using grouping, you have to
@@ -172,15 +173,15 @@ adjust the group names accordingly to the time dimension names.
     import xarray as xr
 
     obs = xr.open_dataset("examples/input_data/observations.nc")["tas"]
-    simp = xr.open_dataset("examples/input_data/control.nc")["tas"]
     simh = simp.copy(deep=True)[3650:]
+    simp = xr.open_dataset("examples/input_data/control.nc")["tas"]
 
     bc = adjust(
         method="linear_scaling",
         obs=obs,
         simh=simh.rename({"time": "t_simh"}),
-        simp=simh,
+        simp=simp,
         kind="+",
         group={"obs": "time.month", "simh": "t_simh.month", "simp": "time.month"},
-        input_core_dims={"obs": "time", "simh": "t_simh", "simp": "time"}
+        input_core_dims={"obs": "time", "simh": "t_simh", "simp": "time"},
     )
diff --git a/doc/introduction.rst b/doc/introduction.rst
index a207a36..99344e4 100644
--- a/doc/introduction.rst
+++ b/doc/introduction.rst
@@ -74,6 +74,11 @@ Please refer to the official documentation for more information about these
 methods as well as sample scripts:
 https://python-cmethods.readthedocs.io/en/stable/
 
+Best Practices and important Notes
+----------------------------------
+
+- The training data should have the same temporal resolution.
+
 - Except for the variance scaling, all methods can be applied on stochastic and
   non-stochastic climate variables. Variance scaling can only be applied on
   non-stochastic climate variables.