stateful-data-processor 3.4.1

Resumable, checkpointed item processing with graceful interrupts.

stateful-data-processor

Resumable, checkpointed item processing with graceful interrupts — subclass and go.

A tiny utility for long-running, restart-safe loops: process items, persist state, resume exactly where you stopped or when an exception is raised, and handle SIGINT/SIGTERM cleanly.

• Install: pip install stateful-data-processor
• Why: Skip rework after crashes/interrupts; keep logic in a single subclass.
• Good for: Batch jobs, ETL steps, scraping, “process a big list with restarts”.

Quick start (60 seconds)

import time
from stateful_data_processor.file_rw import FileRW
from stateful_data_processor.processor import StatefulDataProcessor

class MyDataProcessor(StatefulDataProcessor):

 def process_item(self, item, iteration_index: int, delay: float):
     ''' item and iteration_index are automatically supplied by the framework.
      iteration_index may or may not be used.
     '''
     self.data[item] = item ** 2  # Example processing: square the item
     time.sleep(delay) # Simulate long processing time

# Example usage
file_rw = FileRW('data.json')
processor = MyDataProcessor(file_rw)

items_to_process = [1, 2, 3, 4, 5]
processor.run(items=items_to_process, delay=1.5) # Ctrl+C anytime; rerun to resume.

---

stateful-data-processor is a utility designed to handle large amounts of data incrementally. It allows you to process data step-by-step, saving progress to avoid data loss in case of interruptions or errors. The processor can be subclassed to implement custom data processing logic.

Features

• Incremental & resumable — process large datasets in chunks and pick up exactly where you left off.
• State persisted to disk — saves progress to a file so restarts are fast and reliable.
• Graceful shutdown — handles SIGINT/SIGTERM (e.g., Ctrl+C) and saves state before exiting.
• Crash-safe — catches exceptions, saves current progress, and lets you restart without losing work.
• Automatic logging — a logger is created for you if you don’t inject one.
• Skip completed work — automatically avoids already processed items on restart.
• Easy to extend — subclass to implement custom processing logic.
• Reprocess cached items — optionally revisit items already stored to explore alternative processing strategies.

Problem

Processing massive datasets is slow, brittle, and easy to interrupt. You need a way to:

• Iterate through items one-by-one and save progress to disk as you go.
• Resume exactly where you left off after crashes, timeouts, restarts, or upgrades.
• Gracefully interrupt with SIGINT/SIGTERM (e.g., Ctrl+C) and persist state before exiting.
• Subclass cleanly to provide your own process_data and process_item logic.
• Avoid rework by skipping already-processed items—or intentionally reprocess cached items to explore alternatives.

In short: incremental processing with safety, resumability, and extensibility built in.

Solution

StatefulDataProcessor provides a resilient, incremental pipeline for large datasets:

• Incremental processing: Iterate through big inputs in manageable chunks (e.g., from a JSON source) without starting over.
• Persistent state: Progress and results are stored in a dictionary on disk; the processor tracks the current position.
• Graceful interruption: Handles SIGINT/SIGTERM (e.g., Ctrl+C) and saves state before exiting.
• Subclass-first design: Implement your own logic by overriding process_item (required) and process_data (optional).
• Per-item execution: run(**kwargs) forwards all arguments to process_item, iterating over items and processing one at a time.
• Unique keys: Results are keyed by each item’s unique label, so items must be unique.
• Customizable workflow: Override process_data to pre/post-process items, filter, batch, or enrich as needed.

Usage

Example usage in a large project:

alphaspread analysis of nasdaq symbols

filter ranging stocks

xtb to yfinance symbol conversion

Example: Passing extra arguments via a subclass

File: processors.py

from typing import Any, Optional
from stateful_data_processor.processor import StatefulDataProcessor

class GenericAnalyzer(StatefulDataProcessor):
    """
    Parent processor that expects an extra kwarg: `payload`.
    In a real project this could be HTML, JSON, text, bytes, etc.
    """

    def process_item(self, item: str, payload: Optional[Any], iteration_index: int):
        # Use the extra arg however you like
        self.logger.info(f"[{iteration_index}] Processing {item}; has_payload={payload is not None}")

        # Do minimal "work" for the README: store something derived from the payload
        result = {
            "item": item,
            "payload_preview": str(payload)[:40],  # keep it tiny for docs
            "payload_length": len(str(payload)) if payload is not None else 0,
        }

        # Persist per-item result; the base class handles saving/resuming
        self.data[item] = result

File: run_example.py

# run_example.py
from datetime import date
from typing import Any, List

from stateful_data_processor.file_rw import JsonFileRW
from processors import GenericAnalyzer

def build_payload(item: str) -> Any:
    """
    Stand-in for I/O or computation (e.g., HTTP GET, DB read, cache lookup).
    Keep it simple for the README.
    """
    return f"payload for {item}"

class UrlAnalyzer(GenericAnalyzer):
    """
    Child processor that *adds* the extra argument (`payload`)
    and forwards it to the parent via super().
    """
    def process_item(self, item: str, iteration_index: int):
        payload = build_payload(item)
        # Forward both the original item and the extra kwarg to the parent
        super().process_item(item=item, payload=payload, iteration_index=iteration_index)

if __name__ == "__main__":
    items: List[str] = ["AAPL", "MSFT", "GOOGL", "NVDA"]

    # Results are saved incrementally; reruns resume from where you stopped.
    out_file = JsonFileRW(f"./demo-analysis-{date.today()}.json")

    analyzer = UrlAnalyzer(json_file_writer=out_file)
    analyzer.run(items=items)

    # Access in-memory results if needed
    data = analyzer.data
    print(f"Processed {len(data)} items. Output file: {out_file.path}")

---

Example: Passing extra arguments via a subclass

Sometimes your per-item logic needs more than just the item itself (e.g., a pre-fetched blob, metadata, cached JSON). You can add any keyword args you want to your processor’s process_item signature, and then supply them from a subclass.

processors.py

# processors.py
from typing import Any, Optional
from stateful_data_processor.processor import StatefulDataProcessor

class GenericAnalyzer(StatefulDataProcessor):
    """
    Parent processor that expects an extra kwarg: `payload`.
    In a real project this could be HTML, JSON, text, bytes, etc.
    """

    def process_item(self, item: str, payload: Optional[Any], iteration_index: int):
        # Use the extra arg however you like
        self.logger.info(f"[{iteration_index}] Processing {item}; has_payload={payload is not None}")

        # Do minimal "work" for demonstrational purpose: store something derived from the payload
        result = {
            "item": item,
            "payload_preview": str(payload)[:40],  # keep it tiny for docs
            "payload_length": len(str(payload)) if payload is not None else 0,
        }

        # Persist per-item result; the base class handles saving/resuming
        self.data[item] = result

run_example.py

# run_example.py
from datetime import date
from typing import Any, List

from stateful_data_processor.file_rw import JsonFileRW
from processors import GenericAnalyzer

def build_payload(item: str) -> Any:
    """
    Stand-in for I/O or computation (e.g., HTTP GET, DB read, cache lookup).
    Keep it simple for the README.
    """
    return f"payload for {item}"

class UrlAnalyzer(GenericAnalyzer):
    """
    Child processor that *adds* the extra argument (`payload`)
    and forwards it to the parent via super().
    """
    def process_item(self, item: str, iteration_index: int):
        payload = build_payload(item)
        # Forward both the original item and the extra kwarg to the parent
        super().process_item(item=item, payload=payload, iteration_index=iteration_index)

if __name__ == "__main__":
    items: List[str] = ["AAPL", "MSFT", "GOOGL", "NVDA"]

    # Results are saved incrementally; reruns resume from where you stopped.
    out_file = JsonFileRW(f"./demo-analysis-{date.today()}.json")

    analyzer = UrlAnalyzer(json_file_writer=out_file)
    analyzer.run(items=items)

    # Access in-memory results if needed
    data = analyzer.data
    print(f"Processed {len(data)} items. Output file: {out_file.path}")

What this demonstrates

• The parent class GenericAnalyzer defines process_item(self, item, payload, iteration_index) — here, payload is an extra kwarg (analogous to the soup argument in your original GurufocusAnalyzer).

• The child class UrlAnalyzer overrides process_item(self, item, iteration_index), computes the extra data (payload = build_payload(item)), and then forwards it with:

  super().process_item(item=item, payload=payload, iteration_index=iteration_index)
  

• You can add any keyword arguments you need to the parent’s process_item (e.g., html, row, features, raw_bytes, context, etc.), and supply them from subclasses that know how to build/fetch them.

• All the usual benefits still apply: incremental processing, state persisted to disk, resumability after crashes or Ctrl+C, and a simple subclassing model.

Releasing

git tag x.y
tox
tox -e docs
tox -e build
tox -e publish -- --repository pypi --verbose

Note

This project has been set up using PyScaffold 4.5. For details and usage information on PyScaffold see https://pyscaffold.org/.