Don't copy dataframes or use inplace=True (#305)

Pandas is moving away from the `inplace=True` argument in its next
version, and recommends against using it as most methods make copies of
the underlying data anyway, and using inplace is more likely to lead to
bugs:
pandas-dev/pandas#16529
pandas-dev/pandas#51466

We also have a lot of code that looks like:
```python
df = table.dataframe.copy()
df.foobar(..., inplace=True)
...
table = replace(table, dataframe=df)
```
which is better written as the following, as most pandas methods
`foobar` return a new dataframe with the results of the operation:
```python
df = table.dataframe.foobar(...)
...
table = replace(table, dataframe=df)
```

It is not necessary to copy dataframes unless in very special cases, for
example, you are adding a column to the dataframe:
```python
df = model.topology.copy()  # This copy is needed, otherwise the next line adds a column to `model.topology`
df['new_col'] = df['old_col'] * 2
```

Author: siddharth-krishna

xl2times/__main__.py

@@ -313,8 +313,7 @@ def keep_last_by_file_order(df):
             df = df.sort_values(by="file_order", kind="stable")
             df = df.drop(columns=["source_filename", "file_order"])
         df = df.drop_duplicates(keep="last")
-        df.reset_index(drop=True, inplace=True)
-        return df
+        return df.reset_index(drop=True)
 
     result = {}
     used_tables = set()
@@ -326,7 +325,7 @@ def keep_last_by_file_order(df):
             )
         else:
             used_tables.add(mapping.xl_name)
-            df = input[mapping.xl_name].copy()
+            df = input[mapping.xl_name]
             # Filter rows according to filter_rows mapping:
             for filter_col, filter_val in mapping.filter_rows.items():
                 if filter_col not in df.columns:
@@ -352,7 +351,7 @@ def keep_last_by_file_order(df):
                 # Keep only the required columns
                 cols_to_keep = set(mapping.times_cols).union({"source_filename"})
                 cols_to_drop = [x for x in df.columns if x not in cols_to_keep]
-                df.drop(columns=cols_to_drop, inplace=True)
+                df = df.drop(columns=cols_to_drop)
                 # Drop duplicates, keeping last seen rows as per file order
                 df = keep_last_by_file_order(df)
                 # Drop rows with missing values

xl2times/datatypes.py

@@ -295,8 +295,8 @@ def veda_cgs(self) -> dict[tuple[str, str, str], str]:
         cols = ["region", "process", "commodity", "csets"]
         # Exclude auxillary flows
         index = self.topology["io"].isin({"IN", "OUT"})
-        veda_cgs = self.topology[cols + ["io"]][index].copy()
-        veda_cgs.drop_duplicates(subset=cols, keep="last", inplace=True)
+        veda_cgs = self.topology[cols + ["io"]][index]
+        veda_cgs = veda_cgs.drop_duplicates(subset=cols, keep="last")
         veda_cgs["veda_cg"] = veda_cgs["csets"] + veda_cgs["io"].str[:1]
         veda_cgs = veda_cgs.set_index(["region", "process", "commodity"])[
             "veda_cg"

xl2times/excel.py

@@ -170,7 +170,7 @@ def extract_table(
             # Make all columns names strings as some are integers e.g. years
             table_df.columns = [str(x) for x in df.iloc[header_row, start_col:end_col]]
 
-    table_df.reset_index(drop=True, inplace=True)
+    table_df = table_df.reset_index(drop=True)
 
     # Don't use applymap because it can convert ints to floats
     # https://pandas.pydata.org/pandas-docs/stable/user_guide/gotchas.html#gotchas-intna