Address some pandas FuturesWarnings (#653)

Polish for a demo. See Also: https://stackoverflow.com/a/61185987 --------- Co-authored-by: Sergiy Matusevych <sergiym@microsoft.com>
microsoft · Jan 27, 2024 · 3a36797 · 3a36797
1 parent d073fc1
commit 3a36797
Show file tree

Hide file tree

Showing 3 changed files with 6 additions and 4 deletions.
diff --git a/mlos_bench/mlos_bench/environments/local/local_env.py b/mlos_bench/mlos_bench/environments/local/local_env.py
@@ -193,7 +193,7 @@ def run(self) -> Tuple[Status, Optional[Dict[str, TunableValue]]]:
             _LOG.info("Local results have (metric,value) header and %d rows: assume long format", len(data))
             data = pandas.DataFrame([data.value.to_list()], columns=data.metric.to_list())
             # Try to convert string metrics to numbers.
-            data = data.apply(pandas.to_numeric, errors="ignore")   # type: ignore[assignment]  # (false positive)
+            data = data.apply(pandas.to_numeric, errors='coerce').fillna(data)    # type: ignore[assignment]  # (false positive)
         elif len(data) == 1:
             _LOG.info("Local results have 1 row: assume wide format")
         else:

diff --git a/mlos_bench/mlos_bench/optimizers/mlos_core_optimizer.py b/mlos_bench/mlos_bench/optimizers/mlos_core_optimizer.py
@@ -138,7 +138,7 @@ def _to_df(self, configs: Sequence[Dict[str, TunableValue]]) -> pd.DataFrame:
             if tunable.name in missing_cols:
                 df_configs[tunable.name] = tunable.default
             else:
-                df_configs[tunable.name].fillna(tunable.default, inplace=True)
+                df_configs.fillna({tunable.name: tunable.default}, inplace=True)
             # External data can have incorrect types (e.g., all strings).
             df_configs[tunable.name] = df_configs[tunable.name].astype(tunable.dtype)
             # Add columns for tunables with special values.

diff --git a/mlos_bench/mlos_bench/storage/sql/common.py b/mlos_bench/mlos_bench/storage/sql/common.py
@@ -138,7 +138,8 @@ def get_results_df(
             columns=['trial_id', 'tunable_config_id', 'param', 'value']
         ).pivot(
             index=["trial_id", "tunable_config_id"], columns="param", values="value",
-        ).apply(pandas.to_numeric, errors='ignore')
+        )
+        configs_df = configs_df.apply(pandas.to_numeric, errors='coerce').fillna(configs_df)    # type: ignore[assignment]  # (fp)
 
         # Get each trial's results in wide format.
         results_stmt = schema.trial_result.select().with_only_columns(
@@ -164,7 +165,8 @@ def get_results_df(
             columns=['trial_id', 'metric', 'value']
         ).pivot(
             index="trial_id", columns="metric", values="value",
-        ).apply(pandas.to_numeric, errors='ignore')
+        )
+        results_df = results_df.apply(pandas.to_numeric, errors='coerce').fillna(results_df)    # type: ignore[assignment]  # (fp)
 
         # Concat the trials, configs, and results.
         return trials_df.merge(configs_df, on=["trial_id", "tunable_config_id"], how="left") \