Improvements to scripts/tools/memory/diffsyms.py (#6236)

#### Summary of Changes

- Fixed a bug causing symbols to be skipped.
- Sort by size change rather than (mangled) symbol name.
- Report absent symbols as size 0 rather than NaN.
- Remove trailing blanks from text report format.
- Enable selection options.

scripts/tools/memory/diffsyms.py

@@ -29,6 +29,7 @@
 CONFIG: ConfigDescription = {
     **memdf.util.config.CONFIG,
     **memdf.collect.CONFIG,
+    **memdf.select.CONFIG,
     **memdf.report.REPORT_CONFIG,
     **memdf.report.OUTPUT_CONFIG,
 }
@@ -47,35 +48,43 @@ def main(argv):
         a_dfs = memdf.collect.collect_files(config, files=[inputs[0]])
         b_dfs = memdf.collect.collect_files(config, files=[inputs[1]])
 
-        a_syms = a_dfs[SymbolDF.name].sort_values(by='symbol')
-        b_syms = b_dfs[SymbolDF.name].sort_values(by='symbol')
+        a_syms = a_dfs[SymbolDF.name].sort_values(by='symbol',
+                                                  ignore_index=True)
+        b_syms = b_dfs[SymbolDF.name].sort_values(by='symbol',
+                                                  ignore_index=True)
 
         # TBD: Differences other than size, configurably.
         differences = []
         ai = a_syms.itertuples()
         bi = b_syms.itertuples()
-        while True:
-            if (a := next(ai, None)) is None:
-                break
-            if (b := next(bi, None)) is None:
-                differences.append((a.symbol, a.size, None))
-                break
+        a = next(ai, None)
+        b = next(bi, None)
+        while a and b:
             if a.symbol < b.symbol:
-                differences.append((a.symbol, a.size, None))
+                differences.append((-a.size, a.size, 0, a.symbol))
                 a = next(ai, None)
                 continue
             if a.symbol > b.symbol:
-                differences.append((b.symbol, None, b.size))
+                differences.append((b.size, 0, b.size, b.symbol))
                 b = next(bi, None)
                 continue
             if a.size != b.size:
-                differences.append((a.symbol, a.size, b.size))
+                differences.append((b.size - a.size, a.size, b.size, a.symbol))
+            a = next(ai, None)
+            b = next(bi, None)
         for a in ai:
-            differences.append((a.symbol, a.Index, None))
+            differences.append((-a.size, a.size, 0, a.symbol))
         for b in bi:
-            differences.append((b.symbol, None, b.Index))
-
-        df = pd.DataFrame(differences, columns=['symbol', 'a', 'b'])
+            differences.append((b.size, 0, b.size, b.symbol))
+
+        df = pd.DataFrame(differences,
+                          columns=['change', 'a-size', 'b-size', 'symbol'])
+        if config['report.demangle']:
+            # Demangle early to sort by demangled name.
+            df['symbol'] = df['symbol'].apply(memdf.report.demangle)
+            config['report.demangle'] = False
+        df.sort_values(by=['change', 'symbol'], ascending=[False, True],
+                       inplace=True)
         memdf.report.write_dfs(config, {'Differences': df})
 
     except Exception as exception:

scripts/tools/memory/memdf/report.py

@@ -175,18 +175,30 @@ def write_table(config: Config, df: DF, output: IO) -> None:
         df = df.copy()
         if 'symbol' in df.columns and config['report.demangle']:
             df['symbol'] = df['symbol'].apply(demangle)
+        last_column_is_left_justified = False
+        formatters = []
         for column in df.columns:
             if column.endswith('address'):
                 # Hex format address.
                 width = (int(df[column].max()).bit_length() + 3) // 4
-                df[column] = df[column].apply(lambda s: '{0:0{width}X}'.format(
-                    s, width=width))
+                formatters.append(lambda x:
+                                  '{0:0{width}X}'.format(x, width=width))
             elif pd.api.types.is_string_dtype(df.dtypes[column]):
-                # Left justify strings.
                 df[column] = df[column].astype(str)
+                # Left justify strings.
                 width = max(len(column), df[column].str.len().max())
-                df[column] = df[column].apply(lambda s: s.ljust(width))
-        print(df.to_string(index=False), file=output)
+                formatters.append(lambda x: x.ljust(width))
+                if column == df.columns[-1]:
+                    last_column_is_left_justified = True
+            else:
+                formatters.append(str)
+        s = df.to_string(index=False, formatters=formatters, justify='left')
+        if last_column_is_left_justified:
+            # Strip trailing spaces.
+            for line in s.split('\n'):
+                print(line.rstrip())
+        else:
+            print(s, file=output)
     else:
         # No rows. `df.to_string()` doesn't look like a text table in this case.
         print(' '.join(df.columns))