From: Priya, Data Engineering
Subject: Clean table for web order lines
Analysts keep writing their own FLATTEN over RAW.ORDERS_JSON, and they all get it slightly differently. Please build RIDGELINE.SANDBOX.WEB_ORDER_LINES: one row per order line, with proper types, so they can query it like any table.
Columns: ORDER_ID, PLACED_AT (a timestamp), CUSTOMER_ID, LINE_NO (1 for the first line), SKU, QTY, PRICE. Then we need a load for each new file: this morning's file corrects one existing order and adds a new one, so it has to update and insert.
This project builds a small pipeline the way data engineers do in Snowflake: look at the raw data, write the query that shapes it, save it as a table, use it, and then load changes into it with MERGE.
Steps that build on a query start from what you wrote in the step before. Steps that need the table start with it already built from the model answer, so a slip early on doesn't block the rest.