refactor concat_hist_future (#241)

* refactor concat_hist_future

* Apply suggestions from code review

mesmer/create_emulations/create_emus_gt.py

@@ -6,8 +6,8 @@
 Functions to create global trend emulations with MESMER.
 """
 
-import numpy as np
 
+from mesmer.create_emulations.utils import _concatenate_hist_future
 from mesmer.io.save_mesmer_bundle import save_mesmer_data
 
 
@@ -67,33 +67,22 @@ def create_emus_gt(params_gt, preds_gt, cfg, concat_h_f=False, save_emus=True):
     pred_names = list(preds_gt.keys())
     scenarios_emus = list(preds_gt[pred_names[0]].keys())
 
-    scens_out_f = list(map(lambda x: x.replace("h-", ""), scenarios_emus))
-    # does nothing in case 'h-' not actually included
+    if "h-" in scenarios_emus[0]:
+        scenarios_emus = ["hist"] + [scen.replace("h-", "") for scen in scenarios_emus]
 
-    if concat_h_f:
-        scens_out = scenarios_emus
-    else:
-        if "h-" in scenarios_emus[0]:
-            scens_out = ["hist"] + scens_out_f
-        else:
-            scens_out = scens_out_f
-
-    # initialize global trend emulations dictionary with scenarios as keys
     emus_gt = {}
 
     # apply the chosen method
     if "LOWESS" in params_gt["method"]:
-        if concat_h_f:
-            for scen_out, scen_out_f in zip(scens_out, scens_out_f):
-                emus_gt[scen_out] = np.concatenate(
-                    [params_gt["hist"], params_gt[scen_out_f]]
-                )
-        else:
-            for scen_out in scens_out:
-                emus_gt[scen_out] = params_gt[scen_out]
+        for scen in scenarios_emus:
+            emus_gt[scen] = params_gt[scen]
     else:
         raise ValueError("The chosen method is currently not implemented.")
 
+    if concat_h_f:
+        emus_gt = _concatenate_hist_future(emus_gt)
+        scenarios_emus = list(emus_gt.keys())
+
     # save the global trend emulation if requested
     if save_emus:
         save_mesmer_data(
@@ -107,7 +96,7 @@ def create_emus_gt(params_gt, preds_gt, cfg, concat_h_f=False, save_emus=True):
                 *params_gt["preds"],
                 params_gt["targ"],
                 params_gt["esm"],
-                *scens_out,
+                *scenarios_emus,
             ],
         )
 

mesmer/create_emulations/create_emus_lt.py

@@ -7,10 +7,12 @@
 """
 
 
-import numpy as np
-
 import mesmer.stats
-from mesmer.create_emulations.utils import _gather_params, _gather_preds
+from mesmer.create_emulations.utils import (
+    _concatenate_hist_future,
+    _gather_params,
+    _gather_preds,
+)
 from mesmer.io.save_mesmer_bundle import save_mesmer_data
 
 
@@ -77,18 +79,8 @@ def create_emus_lt(params_lt, preds_lt, cfg, concat_h_f=False, save_emus=True):
     pred_names = list(preds_lt.keys())
     scenarios_emus = list(preds_lt[pred_names[0]].keys())
 
-    if concat_h_f:
-        if scenarios_emus[0] == "hist":
-            scens_out_f = scenarios_emus[1:]
-            scens_out = ["h-" + s for s in scens_out_f]
-        else:
-            raise ValueError("This combination is not supported.")
-    else:
-        if "h-" in scenarios_emus[0]:
-            scens_out_f = list(map(lambda x: x.replace("h-", ""), scenarios_emus))
-            scens_out = ["hist"] + scens_out_f
-        else:
-            scens_out = scens_out_f = scenarios_emus
+    if "h-" in scenarios_emus[0]:
+        scenarios_emus = ["hist"] + [scen.replace("h-", "") for scen in scenarios_emus]
 
     # check if correct predictors
     if pred_names != params_lt["preds"]:
@@ -112,18 +104,13 @@ def create_emus_lt(params_lt, preds_lt, cfg, concat_h_f=False, save_emus=True):
 
     # create emulations
     emus_lt = {}
+
+    for scen in scenarios_emus:
+        emus_lt[scen] = create_emus_method_lt(params_lt, preds_lt, scen)
+
     if concat_h_f:
-        lt_hist = create_emus_method_lt(params_lt, preds_lt, "hist")
-        for scen_out, scen_out_f in zip(scens_out, scens_out_f):
-            lt_scen_f = create_emus_method_lt(params_lt, preds_lt, scen_out_f)
-            emus_lt[scen_out] = {}
-            for targ in params_lt["targs"]:
-                emus_lt[scen_out][targ] = np.concatenate(
-                    [lt_hist[targ], lt_scen_f[targ]]
-                )
-    else:
-        for scen_out in scens_out:
-            emus_lt[scen_out] = create_emus_method_lt(params_lt, preds_lt, scen_out)
+        emus_lt = _concatenate_hist_future(emus_lt)
+        scenarios_emus = list(emus_lt.keys())
 
     # save the local trends emulation if requested
     if save_emus:
@@ -138,7 +125,7 @@ def create_emus_lt(params_lt, preds_lt, cfg, concat_h_f=False, save_emus=True):
                 *params_lt["preds"],
                 *params_lt["targs"],
                 params_lt["esm"],
-                *scens_out,
+                *scenarios_emus,
             ],
         )
 

mesmer/create_emulations/utils.py

@@ -1,6 +1,54 @@
+import numpy as np
 import xarray as xr
 
 
+def _concatenate_hist_future(data):
+    """concatenate historical and future data
+
+    Parameters
+    ----------
+    data : dict
+        Possibly nested dictionary containing arrays to concatenate. The keys of data
+        must correspond to the scenarios to use. The values can either be numpy arrays
+        or dicts of numpy arrays.
+
+    Returns
+    -------
+    concatenated : dict
+        Possibly nested dictionary with concatenated arrays.
+    """
+
+    scens_in = list(data.keys())
+
+    if "hist" not in scens_in:
+        raise ValueError("data does not contain 'hist' scenario")
+
+    scens_in.remove("hist")
+    scens_out = [f"h-{scen}" for scen in scens_in]
+
+    concatenated = {}
+
+    hist = data.pop("hist")
+
+    # data is a nested dict
+    if isinstance(hist, dict):
+
+        for scen_out, scen_in in zip(scens_out, scens_in):
+            concatenated[scen_out] = {}
+            for targ in data[scen_in].keys():
+                concatenated[scen_out][targ] = np.concatenate(
+                    [hist[targ], data[scen_in][targ]]
+                )
+
+    # data is not a nested dict
+    else:
+
+        for scen_out, scen_in in zip(scens_out, scens_in):
+            concatenated[scen_out] = np.concatenate([hist, data[scen_in]])
+
+    return concatenated
+
+
 def _gather_preds(preds_dict, predictor_names, scen, dims):
     """gather predictors for linear regression from legacy data structures