duckrun 0.2.16.dev2__py3-none-any.whl → 0.2.19.dev5__py3-none-any.whl

duckrun/semantic_model.py

@@ -129,29 +129,136 @@ def check_dataset_exists(dataset_name, workspace_id, client):
         return False
 
 
-def refresh_dataset(dataset_name, workspace_id, client, dataset_id=None):
-    """Refresh a dataset and monitor progress using Power BI API"""
+def refresh_dataset(dataset_name, workspace_id, client, dataset_id=None, refresh="full"):
+    """Refresh a dataset and monitor progress using Power BI API
+    
+    For DirectLake models, performs refresh based on refresh parameter:
+    - refresh="full": Two-step refresh (clearValues + full reframe)
+    - refresh="ignore": Skip refresh entirely
+    
+    If a refresh is already in progress, waits for it to complete before starting a new one.
+    """
+    
+    # Skip refresh entirely if refresh is "ignore"
+    if refresh == "ignore":
+        print("   Ignoring refresh - skipping refresh")
+        return
     
     # If dataset_id not provided, look it up by name
     if not dataset_id:
         dataset_id = get_dataset_id(dataset_name, workspace_id, client)
     
-    payload = {
-        "type": "full",
+    # Use Power BI API for refresh (not Fabric API)
+    powerbi_url = f"https://api.powerbi.com/v1.0/myorg/datasets/{dataset_id}/refreshes"
+    headers = client._get_headers()
+    
+    # Check for in-progress refreshes
+    print("   Checking for in-progress refreshes...")
+    try:
+        status_response = requests.get(f"{powerbi_url}?$top=1", headers=headers)
+        if status_response.status_code == 200:
+            refreshes = status_response.json().get('value', [])
+            if refreshes:
+                latest_refresh = refreshes[0]
+                status = latest_refresh.get('status')
+                if status in ['InProgress', 'Unknown']:
+                    refresh_id = latest_refresh.get('requestId')
+                    print(f"   ⚠️  Found in-progress refresh (ID: {refresh_id})")
+                    print(f"   Waiting for current refresh to complete...")
+                    
+                    # Wait for the in-progress refresh to complete
+                    max_wait_attempts = 60
+                    for attempt in range(max_wait_attempts):
+                        time.sleep(5)
+                        check_response = requests.get(f"{powerbi_url}/{refresh_id}", headers=headers)
+                        if check_response.status_code == 200:
+                            current_status = check_response.json().get('status')
+                            
+                            if current_status == 'Completed':
+                                print(f"   ✓ Previous refresh completed")
+                                break
+                            elif current_status == 'Failed':
+                                print(f"   ⚠️  Previous refresh failed, continuing with new refresh")
+                                break
+                            elif current_status == 'Cancelled':
+                                print(f"   ⚠️  Previous refresh was cancelled, continuing with new refresh")
+                                break
+                            
+                            if attempt % 6 == 0:
+                                print(f"   Still waiting... (status: {current_status})")
+                    else:
+                        print(f"   ⚠️  Timeout waiting for previous refresh, will attempt new refresh anyway")
+    except Exception as e:
+        print(f"   ⚠️  Could not check refresh status: {e}")
+        print(f"   Continuing with refresh attempt...")
+    
+    # Step 1: clearValues - Purge data from memory
+    print("   Step 1: Clearing values from memory...")
+    clearvalues_payload = {
+        "type": "clearValues",
         "commitMode": "transactional",
         "maxParallelism": 10,
         "retryCount": 2,
         "objects": []
     }
     
-    # Use Power BI API for refresh (not Fabric API)
-    powerbi_url = f"https://api.powerbi.com/v1.0/myorg/datasets/{dataset_id}/refreshes"
-    headers = client._get_headers()
+    response = requests.post(powerbi_url, headers=headers, json=clearvalues_payload)
     
-    response = requests.post(powerbi_url, headers=headers, json=payload)
+    if response.status_code in [200, 202]:
+        # For 202, monitor the clearValues operation
+        if response.status_code == 202:
+            location = response.headers.get('Location')
+            if location:
+                clear_refresh_id = location.split('/')[-1]
+                print("   ✓ Clear values initiated, monitoring progress...")
+                
+                max_attempts = 60
+                for attempt in range(max_attempts):
+                    time.sleep(2)
+                    
+                    status_url = f"https://api.powerbi.com/v1.0/myorg/datasets/{dataset_id}/refreshes/{clear_refresh_id}"
+                    status_response = requests.get(status_url, headers=headers)
+                    status_response.raise_for_status()
+                    status = status_response.json().get('status')
+                    
+                    if status == 'Completed':
+                        print(f"   ✓ Clear values completed")
+                        break
+                    elif status == 'Failed':
+                        error = status_response.json().get('serviceExceptionJson', '')
+                        raise Exception(f"Clear values failed: {error}")
+                    elif status == 'Cancelled':
+                        raise Exception("Clear values was cancelled")
+                    
+                    if attempt % 10 == 0 and attempt > 0:
+                        print(f"   Clear values status: {status}...")
+                else:
+                    raise Exception(f"Clear values timed out")
+        else:
+            print("   ✓ Clear values completed")
+    else:
+        # Provide detailed error message
+        try:
+            error_details = response.json()
+            error_message = error_details.get('error', {}).get('message', response.text)
+            raise Exception(f"Clear values failed with status {response.status_code}: {error_message}")
+        except (json.JSONDecodeError, ValueError):
+            response.raise_for_status()
+    
+    # Step 2: full refresh - Reframe data from Delta tables
+    print("   Step 2: Full refresh to reframe data...")
+    full_payload = {
+        "type": "full",
+        "commitMode": "transactional",
+        "maxParallelism": 10,
+        "retryCount": 2,
+        "objects": []
+    }
+    
+    response = requests.post(powerbi_url, headers=headers, json=full_payload)
     
     if response.status_code in [200, 202]:
-        print(f"✓ Refresh initiated")
+        print(f"   ✓ Refresh initiated")
         
         # For 202, get the refresh_id from the Location header
         if response.status_code == 202:
@@ -183,7 +290,13 @@ def refresh_dataset(dataset_name, workspace_id, client, dataset_id=None):
                 
                 raise Exception(f"Refresh timed out")
     else:
-        response.raise_for_status()
+        # Provide detailed error message
+        try:
+            error_details = response.json()
+            error_message = error_details.get('error', {}).get('message', response.text)
+            raise Exception(f"Refresh request failed with status {response.status_code}: {error_message}")
+        except (json.JSONDecodeError, ValueError):
+            response.raise_for_status()
 
 
 def download_bim_from_github(url_or_path):
@@ -431,7 +544,7 @@ def create_dataset_from_bim(dataset_name, bim_content, workspace_id, client):
 
 
 def deploy_semantic_model(workspace_name_or_id, lakehouse_name_or_id, schema_name, dataset_name, 
-                         bim_url_or_path, wait_seconds=5):
+                         bim_url_or_path, wait_seconds=5, refresh="full"):
     """
     Deploy a semantic model using DirectLake mode.
     
@@ -442,6 +555,9 @@ def deploy_semantic_model(workspace_name_or_id, lakehouse_name_or_id, schema_nam
         dataset_name: Name for the semantic model
         bim_url_or_path: URL to the BIM file or local file path (e.g., 'model.bim' or 'https://...')
         wait_seconds: Seconds to wait before refresh (default: 5)
+        refresh: Refresh strategy (default: "full")
+            - "full": Clear values and process full refresh
+            - "ignore": Skip refresh entirely
     
     Returns:
         1 for success, 0 for failure
@@ -454,6 +570,9 @@ def deploy_semantic_model(workspace_name_or_id, lakehouse_name_or_id, schema_nam
         # Using a local file
         dr.deploy("./my_model.bim")
         dr.deploy("C:/path/to/model.bim")
+        
+        # Deploy without refresh
+        dr.deploy("./my_model.bim", refresh="ignore")
     """
     print("=" * 70)
     print("Semantic Model Deployment (DirectLake)")
@@ -471,14 +590,14 @@ def deploy_semantic_model(workspace_name_or_id, lakehouse_name_or_id, schema_nam
         dataset_exists = check_dataset_exists(dataset_name, workspace_id, client)
         
         if dataset_exists:
-            print(f"\n✓ Dataset exists - refreshing...")
+            print(f"✓ Dataset '{dataset_name}' already exists - skipping deployment")
             
             if wait_seconds > 0:
                 print(f"   Waiting {wait_seconds} seconds...")
                 time.sleep(wait_seconds)
             
-            print("\n[Step 6/6] Refreshing semantic model...")
-            refresh_dataset(dataset_name, workspace_id, client)
+            print("\n[Step 3/3] Refreshing existing semantic model...")
+            refresh_dataset(dataset_name, workspace_id, client, refresh=refresh)
             
             print("\n" + "=" * 70)
             print("🎉 Refresh Completed!")
@@ -510,7 +629,7 @@ def deploy_semantic_model(workspace_name_or_id, lakehouse_name_or_id, schema_nam
         
         # Step 6: Refresh using the dataset ID returned from creation
         print("\n[Step 6/6] Refreshing semantic model...")
-        refresh_dataset(dataset_name, workspace_id, client, dataset_id=dataset_id)
+        refresh_dataset(dataset_name, workspace_id, client, dataset_id=dataset_id, refresh=refresh)
         
         print("\n" + "=" * 70)
         print("🎉 Deployment Completed!")
@@ -537,7 +656,7 @@ def deploy_semantic_model(workspace_name_or_id, lakehouse_name_or_id, schema_nam
         return 0
 
 
-def copy_model(ws_source, model_name, destination, new_model_name=None, wait_seconds=5):
+def copy_model(ws_source, model_name, destination, new_model_name=None, wait_seconds=5, refresh="full"):
     """
     Copy a semantic model from one workspace to another.
     
@@ -550,6 +669,9 @@ def copy_model(ws_source, model_name, destination, new_model_name=None, wait_sec
         destination: Destination in format "workspace/lakehouse.lakehouse/schema"
         new_model_name: Name for the new semantic model (default: same as source)
         wait_seconds: Seconds to wait before refresh (default: 5)
+        refresh: Refresh strategy (default: "full")
+            - "full": Clear values and process full refresh
+            - "ignore": Skip refresh entirely
     
     Returns:
         1 for success, 0 for failure
@@ -562,6 +684,9 @@ def copy_model(ws_source, model_name, destination, new_model_name=None, wait_sec
         copy_model("Source WS", "Production Model", "Target WS/Data Lake.lakehouse/analytics", 
                    new_model_name="Production Model - Copy")
         
+        # Copy without refresh
+        copy_model("Source WS", "Model", "Target WS/LH.lakehouse/dbo", refresh="ignore")
+        
         # Using the connect pattern
         import duckrun
         duckrun.semantic_model.copy_model("Source", "Model", "Target/LH.lakehouse/dbo")
@@ -688,7 +813,8 @@ def copy_model(ws_source, model_name, destination, new_model_name=None, wait_sec
             schema_name=schema,
             dataset_name=new_model_name,
             bim_url_or_path=temp_bim_path,
-            wait_seconds=wait_seconds
+            wait_seconds=wait_seconds,
+            refresh=refresh
         )
         
         # Clean up temp file

duckrun/stats.py

@@ -60,32 +60,89 @@ def _get_existing_tables_in_schema(duckrun_instance, schema_name: str) -> list:
         return []
 
 
-def get_stats(duckrun_instance, source: str):
+def _match_tables_by_pattern(duckrun_instance, pattern: str) -> dict:
+    """Match tables across all schemas using a wildcard pattern.
+    Pattern can be:
+    - '*.summary' - matches 'summary' table in all schemas
+    - '*summary' - matches any table ending with 'summary'
+    - 'schema.*' - matches all tables in 'schema'
+    Returns a dict mapping schema names to lists of matching table names."""
+    import fnmatch
+    
+    try:
+        # Query all schemas and tables in one go
+        query = """
+            SELECT table_schema, table_name 
+            FROM information_schema.tables 
+            WHERE table_schema NOT LIKE 'pg_%' 
+            AND table_schema != 'information_schema'
+            AND table_name NOT LIKE 'tbl_%'
+        """
+        result = duckrun_instance.con.execute(query).fetchall()
+        
+        matched = {}
+        
+        # Check if pattern contains a dot (schema.table pattern)
+        if '.' in pattern:
+            schema_pattern, table_pattern = pattern.split('.', 1)
+            for schema, table in result:
+                if fnmatch.fnmatch(schema, schema_pattern) and fnmatch.fnmatch(table, table_pattern):
+                    if schema not in matched:
+                        matched[schema] = []
+                    matched[schema].append(table)
+        else:
+            # Pattern matches only table names
+            for schema, table in result:
+                if fnmatch.fnmatch(table, pattern):
+                    if schema not in matched:
+                        matched[schema] = []
+                    matched[schema].append(table)
+        
+        return matched
+    except:
+        return {}
+
+
+def get_stats(duckrun_instance, source: str = None, detailed = False):
     """
     Get comprehensive statistics for Delta Lake tables.
     
     Args:
         duckrun_instance: The Duckrun connection instance
-        source: Can be one of:
+        source: Optional. Can be one of:
+               - None: Use all tables in the connection's schema (default)
                - Table name: 'table_name' (uses main schema in DuckDB)
                - Schema.table: 'schema.table_name' (specific table in schema, if multi-schema)
                - Schema only: 'schema' (all tables in schema, if multi-schema)
+               - Wildcard pattern: '*.summary' (matches tables across all schemas)
+        detailed: Optional. Controls the level of detail in statistics:
+                 - False (default): Aggregated table-level stats (total rows, file count, 
+                   row groups, average row group size, file sizes, VORDER status)
+                 - True: Row group level statistics with compression details, row group sizes,
+                   and parquet metadata
     
     Returns:
-        Arrow table with statistics including total rows, file count, row groups, 
-        average row group size, file sizes, VORDER status, and timestamp
+        DataFrame with statistics based on detailed parameter:
+        - If detailed=False: Aggregated table-level summary
+        - If detailed=True: Granular file and row group level stats
     
     Examples:
         con = duckrun.connect("tmp/data.lakehouse/test")
         
-        # Single table in main schema (DuckDB uses 'main', not 'test')
-        stats = con.get_stats('price_today')
+        # All tables in the connection's schema (aggregated)
+        stats = con.get_stats()
+        
+        # Single table with detailed row group statistics
+        stats_detailed = con.get_stats('price_today', detailed=True)
         
         # Specific table in different schema (only if multi-schema enabled)
         stats = con.get_stats('aemo.price')
         
         # All tables in a schema (only if multi-schema enabled)
         stats = con.get_stats('aemo')
+        
+        # Wildcard pattern across all schemas (only if multi-schema enabled)
+        stats = con.get_stats('*.summary')
     """
     timestamp = datetime.now().strftime('%Y-%m-%d %H:%M:%S')
     
@@ -93,8 +150,31 @@ def get_stats(duckrun_instance, source: str):
     duckdb_schema = "main"
     url_schema = duckrun_instance.schema  # This is from the connection URL path
     
+    # If source is not provided, default to all tables in the connection's schema
+    if source is None:
+        source = url_schema
+    
+    # Check if source contains wildcard characters
+    if '*' in source or '?' in source:
+        # Wildcard pattern mode - only valid if multi-schema is enabled
+        if not duckrun_instance.scan_all_schemas:
+            raise ValueError(f"Wildcard pattern '{source}' not supported. Connection was made to a specific schema '{url_schema}'. Enable multi-schema mode to use wildcards.")
+        
+        matched_tables = _match_tables_by_pattern(duckrun_instance, source)
+        
+        if not matched_tables:
+            raise ValueError(f"No tables found matching pattern '{source}'")
+        
+        # Flatten the matched tables into a list with schema info
+        tables_with_schemas = []
+        for schema, tables in matched_tables.items():
+            for table in tables:
+                tables_with_schemas.append((schema, table))
+        
+        print(f"Found {len(tables_with_schemas)} tables matching pattern '{source}'")
+        
     # Parse the source and validate existence
-    if '.' in source:
+    elif '.' in source:
         # Format: schema.table - only valid if multi-schema is enabled
         schema_name, table_name = source.split('.', 1)
         
@@ -105,46 +185,45 @@ def get_stats(duckrun_instance, source: str):
         if not _table_exists(duckrun_instance, schema_name, table_name):
             raise ValueError(f"Table '{table_name}' does not exist in schema '{schema_name}'")
         
-        list_tables = [table_name]
+        tables_with_schemas = [(schema_name, table_name)]
     else:
         # Could be just table name or schema name
         if duckrun_instance.scan_all_schemas:
             # Multi-schema mode: DuckDB has actual schemas
             # First check if it's a table in main schema
             if _table_exists(duckrun_instance, duckdb_schema, source):
-                list_tables = [source]
-                schema_name = duckdb_schema
+                tables_with_schemas = [(duckdb_schema, source)]
             # Otherwise, check if it's a schema name
             elif _schema_exists(duckrun_instance, source):
                 schema_name = source
                 list_tables = _get_existing_tables_in_schema(duckrun_instance, source)
                 if not list_tables:
                     raise ValueError(f"Schema '{source}' exists but contains no tables")
+                tables_with_schemas = [(schema_name, tbl) for tbl in list_tables]
             else:
                 raise ValueError(f"Neither table '{source}' in main schema nor schema '{source}' exists")
         else:
             # Single-schema mode: tables are in DuckDB's main schema, use URL schema for file paths
             if _table_exists(duckrun_instance, duckdb_schema, source):
                 # It's a table name
-                list_tables = [source]
-                schema_name = url_schema  # Use URL schema for file path construction
+                tables_with_schemas = [(url_schema, source)]
             elif source == url_schema:
                 # Special case: user asked for stats on the URL schema name - list all tables
                 list_tables = _get_existing_tables_in_schema(duckrun_instance, duckdb_schema)
-                schema_name = url_schema  # Use URL schema for file path construction
                 if not list_tables:
                     raise ValueError(f"No tables found in schema '{url_schema}'")
+                tables_with_schemas = [(url_schema, tbl) for tbl in list_tables]
             else:
                 raise ValueError(f"Table '{source}' does not exist in the current context (schema: {url_schema})")
     
     # Use the existing connection
     con = duckrun_instance.con
     
-    print(f"Processing {len(list_tables)} tables: {list_tables}")
+    print(f"Processing {len(tables_with_schemas)} tables from {len(set(s for s, t in tables_with_schemas))} schema(s)")
     
     successful_tables = []
-    for idx, tbl in enumerate(list_tables):
-        print(f"[{idx+1}/{len(list_tables)}] Processing table '{tbl}'...")
+    for idx, (schema_name, tbl) in enumerate(tables_with_schemas):
+        print(f"[{idx+1}/{len(tables_with_schemas)}] Processing table '{schema_name}.{tbl}'...")
         # Construct lakehouse path using correct ABFSS URL format (no .Lakehouse suffix)
         table_path = f"{duckrun_instance.table_base_url}{schema_name}/{tbl}"
         
@@ -171,8 +250,18 @@ def get_stats(duckrun_instance, source: str):
                     print(f"Warning: Could not convert RecordBatch for table '{tbl}': Unexpected type {type(add_actions)}")
                     xx = {}
             
-            # Check if VORDER exists
-            vorder = 'tags.VORDER' in xx.keys()
+            # Check if VORDER exists - handle both formats:
+            # 1. Flattened format: 'tags.VORDER' or 'tags.vorder' in keys
+            # 2. Nested format: check in 'tags' dict for 'VORDER' or 'vorder'
+            vorder = False
+            if 'tags.VORDER' in xx.keys() or 'tags.vorder' in xx.keys():
+                vorder = True
+            elif 'tags' in xx.keys() and xx['tags']:
+                # Check nested tags dictionary (tags is a list of dicts, one per file)
+                for tag_dict in xx['tags']:
+                    if tag_dict and ('VORDER' in tag_dict or 'vorder' in tag_dict):
+                        vorder = True
+                        break
             
             # Calculate total size
             total_size = sum(xx['size_bytes']) if xx['size_bytes'] else 0
@@ -187,6 +276,7 @@ def get_stats(duckrun_instance, source: str):
                 con.execute(f'''
                     CREATE OR REPLACE TEMP TABLE tbl_{idx} AS
                     SELECT 
+                        '{schema_name}' as schema,
                         '{tbl}' as tbl,
                         'empty' as file_name,
                         0 as num_rows,
@@ -199,21 +289,36 @@ def get_stats(duckrun_instance, source: str):
                 ''')
             else:
                 # Get parquet metadata and create temp table with compression info
-                con.execute(f'''
-                    CREATE OR REPLACE TEMP TABLE tbl_{idx} AS
-                    SELECT 
-                        '{tbl}' as tbl,
-                        fm.file_name,
-                        fm.num_rows,
-                        fm.num_row_groups,
-                        CEIL({total_size}/(1024*1024)) as size,
-                        {vorder} as vorder,
-                        COALESCE(STRING_AGG(DISTINCT pm.compression, ', ' ORDER BY pm.compression), 'UNCOMPRESSED') as compression,
-                        '{timestamp}' as timestamp
-                    FROM parquet_file_metadata({delta}) fm
-                    LEFT JOIN parquet_metadata({delta}) pm ON fm.file_name = pm.file_name
-                    GROUP BY fm.file_name, fm.num_rows, fm.num_row_groups
-                ''')
+                if detailed == True:
+                    # Detailed mode: Include ALL parquet_metadata columns
+                    con.execute(f'''
+                        CREATE OR REPLACE TEMP TABLE tbl_{idx} AS
+                        SELECT 
+                            '{schema_name}' as schema,
+                            '{tbl}' as tbl,
+                            {vorder} as vorder,
+                            pm.*,
+                            '{timestamp}' as timestamp
+                        FROM parquet_metadata({delta}) pm
+                    ''')
+                else:
+                    # Aggregated mode: Original summary statistics
+                    con.execute(f'''
+                        CREATE OR REPLACE TEMP TABLE tbl_{idx} AS
+                        SELECT 
+                            '{schema_name}' as schema,
+                            '{tbl}' as tbl,
+                            fm.file_name,
+                            fm.num_rows,
+                            fm.num_row_groups,
+                            CEIL({total_size}/(1024*1024)) as size,
+                            {vorder} as vorder,
+                            COALESCE(STRING_AGG(DISTINCT pm.compression, ', ' ORDER BY pm.compression), 'UNCOMPRESSED') as compression,
+                            '{timestamp}' as timestamp
+                        FROM parquet_file_metadata({delta}) fm
+                        LEFT JOIN parquet_metadata({delta}) pm ON fm.file_name = pm.file_name
+                        GROUP BY fm.file_name, fm.num_rows, fm.num_row_groups
+                    ''')
             
         except Exception as e:
             error_msg = str(e)
@@ -237,6 +342,7 @@ def get_stats(duckrun_instance, source: str):
                     con.execute(f'''
                         CREATE OR REPLACE TEMP TABLE tbl_{idx} AS
                         SELECT 
+                            '{schema_name}' as schema,
                             '{tbl}' as tbl,
                             'empty' as file_name,
                             0 as num_rows,
@@ -261,21 +367,36 @@ def get_stats(duckrun_instance, source: str):
                         filenames.append(table_path + "/" + filename)
                     
                     # Use parquet_file_metadata to get actual parquet stats with compression
-                    con.execute(f'''
-                        CREATE OR REPLACE TEMP TABLE tbl_{idx} AS
-                        SELECT 
-                            '{tbl}' as tbl,
-                            fm.file_name,
-                            fm.num_rows,
-                            fm.num_row_groups,
-                            0 as size,
-                            false as vorder,
-                            COALESCE(STRING_AGG(DISTINCT pm.compression, ', ' ORDER BY pm.compression), 'UNCOMPRESSED') as compression,
-                            '{timestamp}' as timestamp
-                        FROM parquet_file_metadata({filenames}) fm
-                        LEFT JOIN parquet_metadata({filenames}) pm ON fm.file_name = pm.file_name
-                        GROUP BY fm.file_name, fm.num_rows, fm.num_row_groups
-                    ''')
+                    if detailed == True:
+                        # Detailed mode: Include ALL parquet_metadata columns
+                        con.execute(f'''
+                            CREATE OR REPLACE TEMP TABLE tbl_{idx} AS
+                            SELECT 
+                                '{schema_name}' as schema,
+                                '{tbl}' as tbl,
+                                false as vorder,
+                                pm.*,
+                                '{timestamp}' as timestamp
+                            FROM parquet_metadata({filenames}) pm
+                        ''')
+                    else:
+                        # Aggregated mode: Original summary statistics
+                        con.execute(f'''
+                            CREATE OR REPLACE TEMP TABLE tbl_{idx} AS
+                            SELECT 
+                                '{schema_name}' as schema,
+                                '{tbl}' as tbl,
+                                fm.file_name,
+                                fm.num_rows,
+                                fm.num_row_groups,
+                                0 as size,
+                                false as vorder,
+                                COALESCE(STRING_AGG(DISTINCT pm.compression, ', ' ORDER BY pm.compression), 'UNCOMPRESSED') as compression,
+                                '{timestamp}' as timestamp
+                            FROM parquet_file_metadata({filenames}) fm
+                            LEFT JOIN parquet_metadata({filenames}) pm ON fm.file_name = pm.file_name
+                            GROUP BY fm.file_name, fm.num_rows, fm.num_row_groups
+                        ''')
                 
                 print(f"   ✓ Successfully processed '{tbl}' using DuckDB fallback with parquet metadata")
             except Exception as fallback_error:
@@ -291,30 +412,44 @@ def get_stats(duckrun_instance, source: str):
         # No tables were processed successfully - return empty dataframe
         print("⚠️  No tables could be processed successfully")
         import pandas as pd
-        return pd.DataFrame(columns=['tbl', 'total_rows', 'num_files', 'num_row_group', 
-                                     'average_row_group', 'file_size_MB', 'vorder', 'compression', 'timestamp'])
+        if detailed == True:
+            return pd.DataFrame(columns=['schema', 'tbl', 'vorder', 'timestamp'])
+        else:
+            return pd.DataFrame(columns=['schema', 'tbl', 'total_rows', 'num_files', 'num_row_group', 
+                                         'average_row_group', 'file_size_MB', 'vorder', 'compression', 'timestamp'])
     
     # Union all successfully processed temp tables
     union_parts = [f'SELECT * FROM tbl_{i}' for i in successful_tables]
     union_query = ' UNION ALL '.join(union_parts)
     
-    # Generate final summary
-    final_result = con.execute(f'''
-        SELECT 
-            tbl,
-            SUM(num_rows) as total_rows,
-            COUNT(*) as num_files,
-            SUM(num_row_groups) as num_row_group,
-            CAST(CEIL(SUM(num_rows)::DOUBLE / NULLIF(SUM(num_row_groups), 0)) AS INTEGER) as average_row_group,
-            MIN(size) as file_size_MB,
-            ANY_VALUE(vorder) as vorder,
-            STRING_AGG(DISTINCT compression, ', ' ORDER BY compression) as compression,
-            ANY_VALUE(timestamp) as timestamp
-        FROM ({union_query})
-        WHERE tbl IS NOT NULL
-        GROUP BY tbl
-        ORDER BY total_rows DESC
-    ''').df()
+    # Generate final summary based on detailed flag
+    if detailed == True:
+        # Detailed mode: Return ALL parquet_metadata columns
+        final_result = con.execute(f'''
+            SELECT *
+            FROM ({union_query})
+            WHERE tbl IS NOT NULL
+            ORDER BY schema, tbl, file_name, row_group_id, column_id
+        ''').df()
+    else:
+        # Aggregated mode: Original summary statistics
+        final_result = con.execute(f'''
+            SELECT 
+                schema,
+                tbl,
+                SUM(num_rows) as total_rows,
+                COUNT(*) as num_files,
+                SUM(num_row_groups) as num_row_group,
+                CAST(CEIL(SUM(num_rows)::DOUBLE / NULLIF(SUM(num_row_groups), 0)) AS INTEGER) as average_row_group,
+                MIN(size) as file_size_MB,
+                ANY_VALUE(vorder) as vorder,
+                STRING_AGG(DISTINCT compression, ', ' ORDER BY compression) as compression,
+                ANY_VALUE(timestamp) as timestamp
+            FROM ({union_query})
+            WHERE tbl IS NOT NULL
+            GROUP BY schema, tbl
+            ORDER BY total_rows DESC
+        ''').df()
     
     return final_result
 

{duckrun-0.2.16.dev2.dist-info → duckrun-0.2.19.dev5.dist-info}/METADATA

@@ -1,7 +1,7 @@
 Metadata-Version: 2.4
 Name: duckrun
-Version: 0.2.16.dev2
-Summary: Lakehouse task runner powered by DuckDB for Microsoft Fabric
+Version: 0.2.19.dev5
+Summary: Helper library for Fabric Python using duckdb, arrow and delta_rs (orchestration, queries, etc.)
 Author: mim
 License: MIT
 Project-URL: Homepage, https://github.com/djouallah/duckrun