feat: multi programming language select

app.py

@@ -81,6 +81,38 @@ initialize_data_directories()
 # Load data for leaderboard
 LEADERBOARD_DF = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS, BENCHMARK_COLS)
 
+# Extract unique languages for filtering
+def get_unique_languages(df):
+    """Extract all unique individual languages from the Language column"""
+    if df.empty or auto_eval_column_attrs.language.name not in df.columns:
+        return []
+    
+    all_languages = set()
+    for value in df[auto_eval_column_attrs.language.name].unique():
+        if isinstance(value, str):
+            if "/" in value:
+                languages = [lang.strip() for lang in value.split("/")]
+                all_languages.update(languages)
+            else:
+                all_languages.add(value.strip())
+    
+    return sorted(list(all_languages))
+
+# Create a mapping for language filtering
+UNIQUE_LANGUAGES = get_unique_languages(LEADERBOARD_DF)
+
+# Create a special column for individual language filtering
+if not LEADERBOARD_DF.empty:
+    # Create a column that contains all individual languages as a list
+    LEADERBOARD_DF["_languages_list"] = LEADERBOARD_DF[auto_eval_column_attrs.language.name].apply(
+        lambda x: [lang.strip() for lang in str(x).split("/")] if pd.notna(x) else []
+    )
+    
+    # Create a text version of Active Maintenance for checkboxgroup filtering
+    LEADERBOARD_DF["_maintenance_filter"] = LEADERBOARD_DF[auto_eval_column_attrs.availability.name].apply(
+        lambda x: "Active" if x else "Inactive"
+    )
+
 # Load queue data
 (
     finished_eval_queue_df,
@@ -96,30 +128,46 @@ def init_leaderboard(dataframe):
         empty_df = pd.DataFrame(columns=pd.Index(all_columns))
         print("Warning: Leaderboard DataFrame is empty. Using empty dataframe.")
         dataframe = empty_df
-    filter_columns = [
-            ColumnFilter(auto_eval_column_attrs.library_type.name, type="checkboxgroup", label="Library types"),
-            ColumnFilter(auto_eval_column_attrs.language.name, type="checkboxgroup", label="Programming Language"),
-            ColumnFilter(
-                auto_eval_column_attrs.stars.name,
-                type="slider",
-                min=0,
-                max=50000,
-                label="GitHub Stars",
-            ),
-            ColumnFilter(
-                auto_eval_column_attrs.availability.name, type="boolean", label="Show only active libraries" # type: ignore
-            ),
-        ]
+        
+    # Create filter columns list with proper typing
+    filter_columns = []
+    
+    # 1. Library types
+    filter_columns.append(ColumnFilter(auto_eval_column_attrs.library_type.name, type="checkboxgroup", label="Library types"))
+    
+    # 2. Programming Language (checkboxgroup - OR filtering)
+    filter_columns.append(ColumnFilter(auto_eval_column_attrs.language.name, type="checkboxgroup", label="Programming Language"))
+    
+    # 3. GitHub Stars
+    filter_columns.append(ColumnFilter(
+        auto_eval_column_attrs.stars.name,
+        type="slider",
+        min=0,
+        max=50000,
+        label="GitHub Stars",
+    ))
+    
+    # 4. Maintenance Status (checkboxgroup - separate from languages)
+    filter_columns.append(ColumnFilter("_maintenance_filter", type="checkboxgroup", label="Maintenance Status"))
+    
+    # Hide columns
+    hidden_columns = [getattr(auto_eval_column_attrs, field).name for field in AutoEvalColumn.model_fields if getattr(auto_eval_column_attrs, field).hidden]
+    hidden_columns.extend(["_languages_list", "_maintenance_filter", "_original_language"])  # Hide helper columns
+    
+    # Update datatypes
+    datatypes = [getattr(auto_eval_column_attrs, field).type for field in AutoEvalColumn.model_fields]
+    datatypes.extend(["str", "str", "str"])  # For helper columns
+    
     return Leaderboard(
         value=dataframe,
-        datatype=[getattr(auto_eval_column_attrs, field).type for field in AutoEvalColumn.model_fields],
+        datatype=datatypes,
         select_columns=SelectColumns(
             default_selection=[getattr(auto_eval_column_attrs, field).name for field in AutoEvalColumn.model_fields if getattr(auto_eval_column_attrs, field).displayed_by_default],
             cant_deselect=[getattr(auto_eval_column_attrs, field).name for field in AutoEvalColumn.model_fields if getattr(auto_eval_column_attrs, field).never_hidden],
             label="Select Columns to Display:",
         ),
         search_columns=[auto_eval_column_attrs.library.name, auto_eval_column_attrs.license_name.name],
-        hide_columns=[getattr(auto_eval_column_attrs, field).name for field in AutoEvalColumn.model_fields if getattr(auto_eval_column_attrs, field).hidden],
+        hide_columns=hidden_columns,
         filter_columns=filter_columns, # type: ignore
         bool_checkboxgroup_label="Filter libraries",
         interactive=False,
@@ -197,8 +245,8 @@ with demo:
                     language = gr.Dropdown(
                         choices=[i.value.name for i in Language if i != Language.Other],
                         label="Programming Language",
-                        multiselect=False,
-                        value="Python",
+                        multiselect=True,
+                        value=["Python"],
                         interactive=True,
                     )
                     framework = gr.Textbox(label="Framework/Ecosystem (e.g., PyTorch, React)")

src/leaderboard/read_evals.py

@@ -19,6 +19,7 @@ class AssessmentResult(BaseModel):
     results: dict # Risk scores
     framework: str = ""
     language: Language = Language.Other
+    language_str: str = ""  # Original language string to support multiple languages
     library_type: LibraryType = LibraryType.Unknown
     license: str = "?"
     stars: int = 0
@@ -58,7 +59,14 @@ class AssessmentResult(BaseModel):
         # Library metadata
         framework = assessment.get("framework", "")
         language_str = assessment.get("language", "Other")
-        language = next((lang for lang in Language if lang.value.name == language_str), Language.Other)
+        
+        # Handle multiple languages separated by /
+        if "/" in language_str:
+            language_parts = [lang.strip() for lang in language_str.split("/")]
+            # Store the full string but parse the first language for enum
+            language = next((lang for lang in Language if lang.value.name == language_parts[0]), Language.Other)
+        else:
+            language = next((lang for lang in Language if lang.value.name == language_str), Language.Other)
         
         # Availability and verification
         last_update = assessment.get("last_updated", "")
@@ -80,6 +88,7 @@ class AssessmentResult(BaseModel):
             results=risk_scores,
             framework=framework,
             language=language,
+            language_str=language_str,
             license=assessment.get("license", "?"),
             availability=assessment.get("active_maintenance", True),
             verified=assessment.get("independently_verified", False),
@@ -115,7 +124,7 @@ class AssessmentResult(BaseModel):
             "assessment_id": self.assessment_id,  # not a column, just a save name
             auto_eval_column_attrs.library_type.name: self.library_type.value.name,
             auto_eval_column_attrs.library_type_symbol.name: self.library_type.value.symbol,
-            auto_eval_column_attrs.language.name: self.language.value.name, 
+            auto_eval_column_attrs.language.name: self.language_str if self.language_str else self.language.value.name, 
             auto_eval_column_attrs.framework.name: self.framework,
             auto_eval_column_attrs.library.name: make_clickable_library(self.library_name),
             auto_eval_column_attrs.version.name: self.version,

src/populate.py

@@ -6,6 +6,33 @@ from src.display.utils import auto_eval_column_attrs
 from src.leaderboard.read_evals import get_raw_assessment_results
 
 
+def expand_multi_language_entries(df):
+    """Expand multi-language entries (like 'Python/C++') into separate rows for OR filtering"""
+    if df.empty or auto_eval_column_attrs.language.name not in df.columns:
+        return df
+        
+    expanded_rows = []
+    
+    for idx, row in df.iterrows():
+        lang_value = row[auto_eval_column_attrs.language.name]
+        
+        # If language contains /, create separate rows for each language
+        if isinstance(lang_value, str) and "/" in lang_value:
+            languages = [lang.strip() for lang in lang_value.split("/")]
+            for lang in languages:
+                new_row = row.copy()
+                new_row[auto_eval_column_attrs.language.name] = lang
+                new_row["_original_language"] = lang_value  # Keep original for display
+                expanded_rows.append(new_row)
+        else:
+            # Keep single language rows as is
+            row_copy = row.copy()
+            row_copy["_original_language"] = lang_value
+            expanded_rows.append(row_copy)
+    
+    return pd.DataFrame(expanded_rows).reset_index(drop=True)
+
+
 def get_leaderboard_df(eval_results_path, eval_requests_path, cols, benchmark_cols):
     """Read all the runs in the folder and return a dataframe
     
@@ -26,6 +53,9 @@ def get_leaderboard_df(eval_results_path, eval_requests_path, cols, benchmark_co
             # Create dataframe from assessment results
             all_df = pd.DataFrame.from_records([r.to_dict() for r in assessment_results])
             
+            # Expand multi-language entries for OR filtering
+            all_df = expand_multi_language_entries(all_df)
+            
             # Ensure we have all the needed display columns
             all_columns = set(all_df.columns)
             for col in benchmark_cols:

src/submission/submit.py

@@ -57,7 +57,7 @@ def add_new_eval(
         "library": library_name,
         "version": library_version,
         "repository_url": repository_url,
-        "language": language,
+        "language": "/".join(language) if isinstance(language, list) else language,
         "framework": framework,
         "library_type": library_type.value.name,
         "license": license_name,