fix: fix the bug in unselecting anonymous

src/utils.py

@@ -113,10 +113,10 @@ def select_columns(
         selected_cols.append(c)
     # We use COLS to maintain sorting
     filtered_df = df[FIXED_COLS + selected_cols]
-    filtered_df[COL_NAME_AVG] = filtered_df[selected_cols].apply(calculate_mean, axis=1).round(decimals=2)
-    filtered_df.sort_values(by=[COL_NAME_AVG], ascending=False, inplace=True)
-    filtered_df.reset_index(inplace=True, drop=True)
     if reset_ranking:
+        filtered_df[COL_NAME_AVG] = filtered_df[selected_cols].apply(calculate_mean, axis=1).round(decimals=2)
+        filtered_df.sort_values(by=[COL_NAME_AVG], ascending=False, inplace=True)
+        filtered_df.reset_index(inplace=True, drop=True)
         filtered_df[COL_NAME_RANK] = filtered_df[COL_NAME_AVG].rank(ascending=False, method="min")
 
     return filtered_df

tests/test_utils.py

@@ -1,7 +1,8 @@
 import pandas as pd
 import pytest
 
-from src.utils import filter_models, search_table, filter_queries, select_columns, update_table_long_doc, get_iso_format_timestamp, get_default_cols
+from src.utils import filter_models, search_table, filter_queries, select_columns, update_table_long_doc, get_iso_format_timestamp, get_default_cols, update_table
+from src.display.utils import COL_NAME_IS_ANONYMOUS, COL_NAME_REVISION, COL_NAME_TIMESTAMP, COL_NAME_RERANKING_MODEL, COL_NAME_RETRIEVAL_MODEL, COL_NAME_RANK, COL_NAME_AVG
 
 
 @pytest.fixture
@@ -92,4 +93,23 @@ def test_get_default_cols():
     cols, types = get_default_cols("qa")
     for c, t in zip(cols, types):
         print(f"type({c}): {t}")
-    assert len(frozenset(cols)) == len(cols)
+    assert len(frozenset(cols)) == len(cols)
+
+
+def test_update_table():
+    df = pd.DataFrame(
+        {
+            COL_NAME_IS_ANONYMOUS: [False, False, False],
+            COL_NAME_REVISION: ["a1", "a2", "a3"],
+            COL_NAME_TIMESTAMP: ["2024-05-12T12:24:02Z"] * 3,
+            COL_NAME_RERANKING_MODEL: ["NoReranker"] * 3,
+            COL_NAME_RETRIEVAL_MODEL: ["Foo"] * 3,
+            COL_NAME_RANK: [1, 2, 3],
+            COL_NAME_AVG: [0.1, 0.2, 0.3],  # unsorted values
+            "wiki_en": [0.1, 0.2, 0.3]
+        }
+    )
+    results = update_table(df, "wiki", "en", ["NoReranker"], "", show_anonymous=False, reset_ranking=False, show_revision_and_timestamp=False)
+    # keep the RANK as the same regardless of the unsorted averages
+    assert results[COL_NAME_RANK].to_list() == [1, 2, 3]
+