So, I’ve been wrestling with this script that’s supposed to work with a database. The database has these columns: open, high, low, close volume, and unix_timestamp. I’m trying to add some preprocessed indicators to the script, but it’s throwing this traceback error at me.

I really need to sort this out because I’m moving at the end of the semester and I want to have everything ready before then. Any help or ideas would be super appreciated!

These are the traceback errors:

Traceback (most recent call last):   File "C:UsersPycharmProjectspythonProject2processing 2.py", line 369, in process_data
    df = calculate_cycles(df, open_arr, high_arr, low_arr, close_arr)   File "C:UsersPycharmProjectspythonProject2processing 2.py", line 55, in calculate_cycles
    cdl_inside_df = ta.cdl_inside(open_arr, high_arr, low_arr, close_arr)   File "C:UsersAppDataLocalProgramsPythonPython310libsite-packagespandas_tacandlescdl_inside.py", line 16, in cdl_inside
    inside = (high.diff() < 0) & (low.diff() > 0) AttributeError: 'NoneType' object has no attribute 'diff' Error processing data: 'NoneType' object has no attribute 'diff' Loading data from D:New folderdataNew folder - CopyNew folderoutputNew folderdata_1.parquet

This is the script

def load_data(file_path):
    try:
        print(f"Loading data from {file_path}")
        # Load data using Dask
        df = dd.read_parquet(file_path, engine='pyarrow', blocksize=500, columns=['open', 'high', 'low', 'close',
                                                                                  'volume'])
        # Convert columns to numeric, handling errors by coercing
        for col in ['open', 'high', 'low', 'close']:
            df[col] = df[col].astype('float32')

        # Fill NaN values with the mean of the column
        for col in ['open', 'high', 'low', 'close']:
            df[col] = df[col].fillna(df[col].mean())

        # Convert to NumPy ndarrays
        open_arr = df['open'].compute().to_numpy()
        high_arr = df['high'].compute().to_numpy()
        low_arr = df['low'].compute().to_numpy()
        close_arr = df['close'].compute().to_numpy()
        volume_arr = df['volume'].compute().to_numpy()

        return df, open_arr, high_arr, low_arr, close_arr, volume_arr
    except Exception as e:
        print(f"Error loading data from {file_path}: {e}")
        traceback.print_exc()
        return None, None, None, None, None, None


# Function to process each file
def calculate_cycles(df, open_arr, high_arr, low_arr, close_arr):
    if open_arr is not None and high_arr is not None and low_arr is not None and close_arr is not None:
        ebsw_df = ta.ebsw(close_arr)
        cdl_doji_df = ta.cdl_doji(open_arr, high_arr, low_arr, close_arr)
        cdl_inside_df = ta.cdl_inside(open_arr, high_arr, low_arr, close_arr)
        cdl_pattern_df = ta.cdl_pattern(open_arr, high_arr, low_arr, close_arr)

        df['ebsw_EBSW'] = ebsw_df
        df['CDL_DOJI_CDL_DOJI'] = cdl_doji_df
        df['CDL_INSIDE_CDL_INSIDE'] = cdl_inside_df
        df['CDL_PATTERN_CDL_PATTERN'] = cdl_pattern_df

        return df
    else:
        return df

def process_data(file_path):
    try:
        df, open_arr, high_arr, low_arr, close_arr, volume_arr = load_data(file_path)

        df = calculate_cycles(df, open_arr, high_arr, low_arr, close_arr)

        # Fill NaN values with 0
        df = df.fillna(0)
        return df
    except Exception as e:
        print(f"Error processing data: {e}")
        traceback.print_exc()
        return None


if __name__ == '__main__':
    # Directory to walk
    file_path = r"D:New folderdataNew folder - CopyNew folderoutputNew folder"
    # Prepare a list of files to process
    parquet_files = [os.path.join(root, file)
                     for root, dirs, files in os.walk(file_path)
                     for file in files if file.endswith('.parquet')]
    # Process files in chunks of 100
    chunk_size = 2

    for i in range(0, len(parquet_files), chunk_size):
        chunk_files = parquet_files[i:i + chunk_size]
        # Process files in parallel for each chunk
        with ThreadPoolExecutor(32) as executor:
            for file_path in chunk_files:
                df = process_data(file_path)
                if df is not None:
                    df.to_parquet(file_path.replace('.parquet', '_processed.parquet'), compression="zstd",
                                  engine="pyarrow",
                                  index=True)
        gc.collect()  # Force garbage collection
        # Sleep for 10 seconds to avoid rate limiting
        time.sleep(5)