Pandas DataFrame Properties

Prerequisites

• A Custom Python Connection

Required Python Functions

The following table describes the functions available when creating a new Python Read Connector utilizing Pandas DataFrame interface. Create a Python Read Connector using the information below and these step-by-step instructions.

🚧

Out of Memory Exception

Because Pandas DataFrames are loaded into processing memory, large amounts of data can result in an out of memory exception.

Recursive list_objects

Metadata is a Python dictionary that defines a partition. Metadata is used in both list_objects and read_bytes. To trigger the recursive behavior within in list_objects and create partitions, set is_prefix to True. If a previously created partition is not recalled when generating list_objects, all previous partition metadata will be deleted.

🚧

When constructing your Python code, list_objects must return the partition metadata for all the partitions you expect to be in the component.

Example Pandas DataFrame Code

The following code example describes reading a spreadsheet for Google Sheets.

# This example reads a spreadsheet from Google Sheets to explain the functions to implement

import pandas as pd
from typing import Dict, Any, Iterator


def context(credentials) -> Dict[str, Any]:
   """
     Sets up the context for reading and listing data from data source.
     This is where the Python Connection information will be passed through. 
     Avoid opening a database connection. 
     """
  service_account_info = json.loads(credentials)

  creds = service_account.Credentials.from_service_account_info(service_account_info, scopes=SCOPES)
  g_sheet = build('sheets', 'v4', credentials=creds)
  drive = build('drive', 'v3', credentials=creds)

  return {
      'g_sheet_client': g_sheet,
      'drive_client': drive,
  }

def list_objects(context: Dict[str, Any], metadata) -> Iterator[Dict[str, Any]]:

  yield {'name': 'example_id', 'fingerprint': 'fingerprint', 'is_prefix': False}


def read_pandas_dataframe(context: Dict[str, Any], metadata) -> pd.DataFrame:
  """
    # Returns a Pandas DataFrame.
    """
  data = [['Scott', 50], ['Jeff', 45], ['Thomas', 54], ['Ann', 34]]
  # Create the pandas DataFrame
  return pd.DataFrame(data, columns=['Name', 'Age'])

👍

External Keys and Secrets Management for Credentials

To use external keys or a secrets manager, set the def context(credentials: str) where str is the name you assigned the credential. Any connector code interface will receive external key as a result of that function call.

Parsers

Ascend natively parses Pandas DataFrames and does not require additional parser configurations.

Code Fingerprint Strategy

The Code Fingerprint Strategy controls Ascend's DataAware function. Ascend read's the code and the configuration, and calculates the hash of the component. If the code changes in any way, the data must be reprocessed. Currently, the only available setting is Automatic content-based fingerprint.

Schema Generation

The Schema Generation reads the first few rows of the dataset and generates a schema just as with other Ascend Read Connectors. However, in order to sample the dataset, schema generation will run the code inserted into the Connector Code Interface across the entire dataset.