Fix generic defaults enrichment for single-country runs

When running single-country simulations, the *Default tables
(pGenDataInputDefault, pAvailabilityDefault) could become empty after
zone filtering, causing the generic enrichment to be skipped entirely.

Now handles the empty case by creating defaults from scratch using
generic data, similar to how pCapexTrajectoriesDefault already worked.
For pAvailabilityDefault, seasons are taken from pHours instead of
hardcoded values.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Author: lucas-vivier

epm/input_treatment.py

@@ -2146,50 +2146,83 @@ def _enrich_defaults_with_generic(db: gt.Container):
         # GAMS-loaded generic data is already in long format: (tech, f, pGenDataInputHeader, value)
         if "gendata" in generic and "pGenDataInputDefault" in db:
             gendata_default = db["pGenDataInputDefault"].records
-            if gendata_default is not None and not gendata_default.empty:
+            generic_gendata = generic["gendata"]  # Already has (tech, f, pGenDataInputHeader, value)
+
+            if gendata_default is None or gendata_default.empty:
+                # pGenDataInputDefault is empty - create it entirely from generic data
+                to_add = required.merge(generic_gendata, on=["tech", "f"], how="inner")
+                if not to_add.empty:
+                    db.data["pGenDataInputDefault"].setRecords(to_add)
+                    _write_back(db, "pGenDataInputDefault")
+                    n_combos = to_add[["z", "tech", "f"]].drop_duplicates().shape[0]
+                    gams.printLog(f"[input_treatment][generic] Created pGenDataInputDefault with {n_combos} (zone, tech, fuel) combination(s) from generic.")
+            else:
+                # pGenDataInputDefault has data - only add missing combinations
                 gendata_keys = gendata_default[["z", "tech", "f"]].drop_duplicates()
                 merged = required.merge(gendata_keys, on=["z", "tech", "f"], how="left", indicator=True)
                 missing = merged[merged["_merge"] == "left_only"][["z", "tech", "f"]]
-                
+
                 if not missing.empty:
-                    generic_gendata = generic["gendata"]  # Already has (tech, f, pGenDataInputHeader, value)
                     # Join missing combinations with generic values
                     to_add = missing.merge(generic_gendata, on=["tech", "f"], how="inner")
                     if not to_add.empty:
                         updated = pd.concat([gendata_default, to_add], ignore_index=True)
                         db.data["pGenDataInputDefault"].setRecords(updated)
                         _write_back(db, "pGenDataInputDefault")
                         n_combos = to_add[["z", "tech", "f"]].drop_duplicates().shape[0]
-                        gams.printLog(f"[input_treatment][generic] Enriched pGenDataInputDefault with {n_combos} tech-fuel combination(s) from generic.")
+                        gams.printLog(f"[input_treatment][generic] Enriched pGenDataInputDefault with {n_combos} (zone, tech, fuel) combination(s) from generic.")
 
         # Enrich pAvailabilityDefault
         # GAMS-loaded generic data is (tech, f, value) - single annual value
-        if "availability" in generic and "pAvailabilityDefault" in db:
+        # Get seasons from pHours
+        seasons_from_hours = []
+        if "pHours" in db:
+            phours_records = db["pHours"].records
+            if phours_records is not None and not phours_records.empty:
+                q_col = "q" if "q" in phours_records.columns else phours_records.columns[0]
+                seasons_from_hours = phours_records[q_col].unique().tolist()
+
+        if "availability" in generic and "pAvailabilityDefault" in db and seasons_from_hours:
             avail_default = db["pAvailabilityDefault"].records
-            if avail_default is not None and not avail_default.empty:
+            generic_avail = generic["availability"]  # (tech, f, value)
+
+            if avail_default is None or avail_default.empty:
+                # pAvailabilityDefault is empty - create it entirely from generic data
+                to_add = required.merge(generic_avail, on=["tech", "f"], how="inner")
+                if not to_add.empty:
+                    new_records = []
+                    for _, row in to_add.iterrows():
+                        for q in seasons_from_hours:
+                            new_records.append({"z": row["z"], "tech": row["tech"], "f": row["f"],
+                                              "q": q, "value": row["value"]})
+                    if new_records:
+                        new_df = pd.DataFrame(new_records)
+                        db.data["pAvailabilityDefault"].setRecords(new_df)
+                        _write_back(db, "pAvailabilityDefault")
+                        n_combos = to_add[["z", "tech", "f"]].drop_duplicates().shape[0]
+                        gams.printLog(f"[input_treatment][generic] Created pAvailabilityDefault with {n_combos} (zone, tech, fuel) combination(s) from generic.")
+            else:
+                # pAvailabilityDefault has data - only add missing combinations
                 avail_keys = avail_default[["z", "tech", "f"]].drop_duplicates()
                 merged = required.merge(avail_keys, on=["z", "tech", "f"], how="left", indicator=True)
                 missing = merged[merged["_merge"] == "left_only"][["z", "tech", "f"]]
-                
+
                 if not missing.empty:
-                    generic_avail = generic["availability"]  # (tech, f, value)
-                    # Get seasons from existing default (q column in long format)
-                    seasons = avail_default["q"].unique().tolist() if "q" in avail_default.columns else ["Q1"]
-                    
                     # Join missing combinations with generic values, expand to all seasons
                     to_add = missing.merge(generic_avail, on=["tech", "f"], how="inner")
                     if not to_add.empty:
                         new_records = []
                         for _, row in to_add.iterrows():
-                            for q in seasons:
-                                new_records.append({"z": row["z"], "tech": row["tech"], "f": row["f"], 
+                            for q in seasons_from_hours:
+                                new_records.append({"z": row["z"], "tech": row["tech"], "f": row["f"],
                                                   "q": q, "value": row["value"]})
                         if new_records:
                             new_df = pd.DataFrame(new_records)
                             updated = pd.concat([avail_default, new_df], ignore_index=True)
                             db.data["pAvailabilityDefault"].setRecords(updated)
                             _write_back(db, "pAvailabilityDefault")
-                            gams.printLog(f"[input_treatment][generic] Enriched pAvailabilityDefault with {len(to_add)} tech-fuel combination(s) from generic.")
+                            n_combos = to_add[["z", "tech", "f"]].drop_duplicates().shape[0]
+                            gams.printLog(f"[input_treatment][generic] Enriched pAvailabilityDefault with {n_combos} (zone, tech, fuel) combination(s) from generic.")
 
         # Enrich pCapexTrajectoriesDefault
         # GAMS-loaded generic data is already in long format: (tech, f, y, value)