with first draft of spreading to multiple index

pymrio/tools/ioutil.py

@@ -1110,7 +1110,7 @@ def convert(
     if isinstance(df_orig, pd.Series):
         df_orig = pd.DataFrame(df_orig)
 
-    # some consitency checks of arguments and restructuring if everything is ok
+    # some consistency checks of arguments and restructuring if everything is ok
     if len(bridge_columns) == 0:
         raise ValueError("No columns with '__' in the mapping DataFrame")
     for col in bridge_columns:
@@ -1149,6 +1149,7 @@ def convert(
     res_collector = []
 
     # loop over each new impact/characterized value
+    # and collect entries, multiply and rename
     for entry in unique_new_index:
         df_cur_map = df_map.loc[[entry]]
         collector = []
@@ -1164,32 +1165,59 @@ def convert(
 
         df_collected = pd.concat(collector, axis=0)
 
+        # renaming part, checks if the old name (bridge.orig) is in the current index
+        # and renames by the new one (bridge.new)
+
+        already_renamed = dict()
         for bridge in bridges:
+            # encountering a bridge with the same orig name but which should
+            # lead to two new index levels
+            if bridge.orig in already_renamed.keys():
+                # duplicate the index level
+                df_collected.reset_index(level=already_renamed[bridge.orig].new, inplace=True)
+                df_collected[bridge.new] = df_cur_map.index.get_level_values(bridge.raw)[0]
+
+                if df_collected.index.name is None:
+                    df_collected.set_index(already_renamed[bridge.orig].new, drop=True, append=False, inplace=True)
+                else:
+                    df_collected.set_index(already_renamed[bridge.orig].new, drop=True, append=True, inplace=True)
+                df_collected.set_index(bridge.new, drop=True, append=True, inplace=True)
+                continue
+
             for idx_old_names in df_collected.index.names:
                 if bridge.orig in idx_old_names:
+                    # rename the index names
                     if isinstance(df_collected.index, pd.MultiIndex):
                         df_collected.index = df_collected.index.set_names( bridge.new, level=idx_old_names)
                     else:
                         df_collected.index = df_collected.index.set_names( bridge.new, level=None)
 
+                    # rename the actual index values
                     df_collected.reset_index(level=bridge.new, inplace=True)
-
                     for row in df_cur_map.reset_index().iterrows():
                         new_row_name = row[1][bridge.raw]
                         old_row_name = row[1][bridge.orig]
                         df_collected.loc[:, bridge.new] = df_collected.loc[
                             :, bridge.new
                         ].str.replace(pat=old_row_name, repl=new_row_name, regex=True)
-                    df_collected.set_index(
-                        # CONT: Make test cases for renaming/chacterization of a df without a multiindex
                         # CONT: Make a test case/method where a matching line gets extended into more index columns
-                        bridge.new, drop=True, append=True, inplace=True
-                    )
+                        # CONT: Ensure that the spread keeps the order as in the original mapping
+
+                    # put the index back
+                    if df_collected.index.name is None:
+                        # The case with a single index where the previous reset index
+                        # left only a numerical index
+                        df_collected.set_index(
+                            bridge.new, drop=True, append=False, inplace=True
+                        )
+                    else:
+                        df_collected.set_index(
+                            bridge.new, drop=True, append=True, inplace=True
+                        )
+                    already_renamed[bridge.orig] = bridge
 
         res_collector.append(
-
             df_collected.groupby(by=df_collected.index.names).agg(agg_func)
-
         )
 
     all_result = pd.concat(res_collector, axis=0)

tests/test_util.py

@@ -354,9 +354,34 @@ def test_util_regex():
     assert len(df_none_match) == 0
     assert len(df_none_match_index) == 0
 
+def test_convert_rename_singleindex():
+    """Testing the renaming of one table with a single index"""
 
-def test_convert_rename():
-    """Testing the renaming of one table"""
+    to_char = pd.DataFrame(
+        data=99.0,
+        index=["em1", "em2", "em3"],
+        columns=["r1", "r2", "r3"]
+    )
+    to_char.index.name = "em_type"
+    to_char.columns.name = "reg"
+
+    rename_bridge_simple = pd.DataFrame(
+        columns=["em_type", "stressor__em_type"],
+        data=[
+            ["em1", "emission1"],
+            ["em2", "emission2"],
+            ["em3", "emission3"],
+        ],
+    )
+
+    renamed = convert(to_char, rename_bridge_simple)
+    assert all(renamed.columns == renamed.columns)
+    assert all(renamed.index == rename_bridge_simple["stressor__em_type"])
+
+
+
+def test_convert_rename_multiindex():
+    """Testing the renaming of one table with a multiindex"""
 
     to_char = pd.DataFrame(
         data=99.0,
@@ -440,6 +465,47 @@ def test_convert_rename():
     pdt.assert_frame_equal(char_res_keep_comp_wo_factor, char_res_keep_comp)
 
 
+def test_convert_rename_spread_index():
+    """Testing the renaming of one table from an index to an multiindex
+
+    This is a specific case for the EXIOBASE to GLAM conversion, 
+    where one stressor level need to be spread to multiple flows/classes
+    """
+
+    to_char = pd.DataFrame(
+        data=99.0,
+        index=["em1", "em2", "em3"],
+        columns=["r1", "r2", "r3"]
+    )
+    to_char.index.name = "stressor"
+    to_char.columns.name = "reg"
+
+    rename_bridge = pd.DataFrame(
+        columns=["stressor", "flow__stressor", "class__stressor", "class2__stressor"],
+        data=[
+            ["em1", "emission1", "to_air", "to_air (unspecified)"],
+            ["em2", "emission2", "to_air", "to_air (specified)"],
+            ["em3", "emission3", "to_water", "to_water (unpecified)"],],
+    )
+
+
+    rename_bridge = pd.DataFrame(
+        columns=["stressor", "class__stressor", "class2__stressor"],
+        data=[
+            ["em1", "to_air", "to_air (unspecified)"],
+            ["em2", "to_air", "to_air (specified)"],
+            ["em3", "to_water", "to_water (unpecified)"],],
+    )
+
+
+    renamed = convert(to_char, rename_bridge)
+
+    assert all(renamed.columns == renamed.columns)
+    assert all(renamed.index == rename_bridge_simple["stressor__em_type"])
+
+
+
+
 def test_convert_characterize():
     """Testing the characterization of one table"""