Data Engineering

ETL Pipeline Pattern

from datetime import datetime
from dataclasses import dataclass

@dataclass
class PipelineResult:
    records_extracted: int
    records_transformed: int
    records_loaded: int
    errors: list[str]
    duration_seconds: float

class OrderPipeline:
    def __init__(self, source_db, warehouse_db):
        self.source = source_db
        self.warehouse = warehouse_db

    def extract(self, since: datetime) -> list[dict]:
        query = """
            SELECT o.*, c.name as customer_name, c.segment
            FROM orders o
            JOIN customers c ON o.customer_id = c.id
            WHERE o.updated_at > %s
        """
        return self.source.fetch_all(query, [since])

    def transform(self, records: list[dict]) -> list[dict]:
        transformed = []
        for record in records:
            transformed.append({
                "order_id": record["id"],
                "customer_name": record["customer_name"],
                "segment": record["segment"].upper(),
                "total_amount": float(record["total"]),
                "order_date": record["created_at"].date(),
                "fiscal_quarter": get_fiscal_quarter(record["created_at"]),
                "is_high_value": float(record["total"]) > 1000,
                "loaded_at": datetime.utcnow(),
            })
        return transformed

    def load(self, records: list[dict]) -> int:
        return self.warehouse.upsert_batch(
            table="fact_orders",
            records=records,
            conflict_keys=["order_id"],
            batch_size=5000,
        )