Pipeline tests on data that still joins

Live sample — synthetic, runnable now

These rows come from the published Retail & E-commerce bundle: invented records, generated from a fixed seed, with every key intact.

Masked by the real engine with a fixed public sample key: SHP000000001 becomes SHP098977906 in both storefront.shoppers.shopper_id and orders.orders.shopper_ref, so the join still holds after masking.

Run it yourself

Check referential integrity between orders and order lines in the Retail & E-commerce bundle. Python 3, standard library only.

import csv, io, urllib.request

BASE = "https://www.datanivra.com/downloads/packs/retail-ecommerce/2.0.0/synthetic"


def table(name):
    with urllib.request.urlopen(f"{BASE}/{name}.csv") as response:
        return list(csv.DictReader(io.StringIO(response.read().decode("utf-8"))))


orders = {o["order_number"] for o in table("orders.orders")}
lines = table("orders.order_lines")
dangling = [line["line_id"] for line in lines if line["order_number"] not in orders]
print(f"{len(lines)} order lines, {len(orders)} orders, {len(dangling)} dangling")

Expected output:

11 order lines, 5 orders, 0 dangling

Full tested example: Relational subsetting without orphan rows. Or open the synthetic playground for this pack — no account.