Enhanced error handling and debugging for nemaquant.py execution

app.py

@@ -107,12 +107,27 @@ def process_images():
             print(f"Weights file exists: {os.path.exists(str(WEIGHTS_FILE))}")
             print(f"Weights file size: {os.path.getsize(str(WEIGHTS_FILE)) if os.path.exists(str(WEIGHTS_FILE)) else 'File not found'} bytes")
             
+            # Additional debugging before running the command
+            print(f"Current working directory: {os.getcwd()}")
+            print(f"Output directory permissions: {oct(os.stat(str(job_output_dir)).st_mode)[-3:] if os.path.exists(str(job_output_dir)) else 'Directory not found'}")
+            print(f"Executable Python: {sys.executable}")
+            
             # Run the script, capture output and errors
             # Timeout might be needed for long processes on shared infrastructure like HF Spaces
-            result = subprocess.run(cmd, capture_output=True, text=True, check=True, timeout=300) # 5 min timeout
+            result = subprocess.run(cmd, capture_output=True, text=True, check=False, timeout=300) # 5 min timeout, don't raise exception
+            
+            # Get script return code and output
+            print(f"NemaQuant script return code: {result.returncode}")
+            print(f"NemaQuant stdout: {result.stdout}")
+            print(f"NemaQuant stderr: {result.stderr}")
+            
             status_log = f"NemaQuant Output:\n{result.stdout}\nNemaQuant Errors:\n{result.stderr}"
             print(status_log) # Log script output
             
+            # Check if the command failed
+            if result.returncode != 0:
+                raise subprocess.CalledProcessError(result.returncode, cmd, output=result.stdout, stderr=result.stderr)
+            
             # Check output directory after command runs
             print(f"Output directory exists: {os.path.exists(str(job_output_dir))}")
             if os.path.exists(str(job_output_dir)):
@@ -120,10 +135,63 @@ def process_images():
             
             # Check output file
             print(f"Output CSV exists: {os.path.exists(str(output_csv))}")
+            
+            # If CSV doesn't exist, try to check if there are any other CSVs created in the output directory
+            if not output_csv.exists() and os.path.exists(str(job_output_dir)):
+                csv_files = [f for f in os.listdir(str(job_output_dir)) if f.endswith('.csv')]
+                if csv_files:
+                    print(f"Found other CSV files in output directory: {csv_files}")
+                    # Try using the first found CSV instead
+                    output_csv = job_output_dir / csv_files[0]
+                    print(f"Using alternate CSV file: {output_csv}")
+                else:
+                    # No CSV files found - create a fallback CSV file manually
+                    # This is a workaround in case nemaquant.py's to_csv() fails due to permissions
+                    try:
+                        print("No CSV found - trying to create a fallback CSV file")
+                        # Look for annotated images to get filenames and create a basic CSV
+                        annotated_files = [f for f in os.listdir(str(job_output_dir)) if '_annotated' in f]
+                        if annotated_files:
+                            # Create a simple CSV with the filenames found
+                            fallback_csv = job_output_dir / f"{job_id}_fallback_results.csv"
+                            with open(fallback_csv, 'w') as f:
+                                f.write("filename,num_eggs,fallback\n")
+                                for img_file in annotated_files:
+                                    # Extract original filename by removing '_annotated'
+                                    orig_filename = img_file.replace('_annotated', '')
+                                    f.write(f"{orig_filename},0,true\n")
+                            output_csv = fallback_csv
+                            print(f"Created fallback CSV at {fallback_csv}")
+                    except Exception as csv_error:
+                        print(f"Error creating fallback CSV: {csv_error}")
 
             # --- Parse Results ---
             if not output_csv.exists():
-                 raise FileNotFoundError(f"Output CSV not found at {output_csv}")
+                 error_message = f"Output CSV not found at {output_csv}"
+                 print(error_message)
+                 # Check if nemaquant.py exists and try to debug file paths
+                 nemaquant_script = APP_ROOT / 'nemaquant.py'
+                 nemaquant_exists = nemaquant_script.exists()
+                 error_message += f"\nNemaQuant script exists: {nemaquant_exists}"
+                 if nemaquant_exists:
+                     # Try to inspect the nemaquant script to understand its file I/O
+                     try:
+                         with open(nemaquant_script, 'r') as f:
+                             script_content = f.read()
+                             # Look for common file output patterns
+                             if 'to_csv(' in script_content:
+                                 error_message += "\nNemaQuant script contains 'to_csv()' calls for file output"
+                             if 'open(' in script_content and 'w' in script_content:
+                                 error_message += "\nNemaQuant script contains file opening in write mode"
+                     except Exception as e:
+                         error_message += f"\nCould not read nemaquant.py: {str(e)}"
+
+                 # Check if any files were written at all
+                 if os.path.exists(str(job_output_dir)):
+                     files_in_output = os.listdir(str(job_output_dir))
+                     error_message += f"\nFiles in output directory: {files_in_output}"
+                 
+                 raise FileNotFoundError(error_message)
 
             df = pd.read_csv(output_csv)
             # Expect columns like 'filename', 'num_eggs' (based on nemaquant.py)
@@ -224,6 +292,29 @@ if __name__ == '__main__':
     print(f"Weights file exists: {os.path.exists(str(WEIGHTS_FILE))}")
     if os.path.exists(str(WEIGHTS_FILE)):
         print(f"Weights file size: {os.path.getsize(str(WEIGHTS_FILE))} bytes")
+    
+    # Check if running in Docker or Hugging Face Space container
+    in_container = os.path.exists('/.dockerenv') or os.environ.get('SPACE_ID')
+    print(f"Running in container: {in_container}")
+    if in_container:
+        print("Container environment detected:")
+        print(f"User running process: {os.getuid()}:{os.getgid()}")
+        try:
+            # Check permissions for key directories
+            for path in [UPLOAD_FOLDER, RESULT_FOLDER]:
+                stat_info = os.stat(path)
+                print(f"Permissions for {path}: {oct(stat_info.st_mode)[-3:]}")
+                print(f"Owner for {path}: {stat_info.st_uid}:{stat_info.st_gid}")
+        except Exception as e:
+            print(f"Error checking permissions: {e}")
+    
+    print("---------------------------------------")
+
+    # Check for nemaquant.py script
+    nemaquant_script = APP_ROOT / 'nemaquant.py'
+    print(f"NemaQuant script exists: {nemaquant_script.exists()}")
+    if nemaquant_script.exists():
+        print(f"NemaQuant script permissions: {oct(os.stat(nemaquant_script).st_mode)[-3:]}")
     print("---------------------------------------")
 
     app.run(debug=True, host='0.0.0.0', port=7860) # Port 7860 is common for HF Spaces