The source report identified a hard SQL question about a rolling sum but did not preserve its schema or exact output. The following is a self-contained practice reconstruction of that topic. You have a PostgreSQL table: ```text orders ------ order_id BIGINT PRIMARY KEY seller_id BIGINT NOT NULL ordered_at TIMESTAMPTZ NOT NULL status TEXT NOT NULL -- 'completed', 'cancelled', or 'refunded' amount NUMERIC NOT NULL -- nonnegative order amount ``` Write one PostgreSQL query that returns a daily revenue series for every seller with at least one completed order. For each such seller: - Start the series on that seller's earliest UTC completed-order date. - End the series on that seller's latest UTC completed-order date. - Include every calendar date in between, even when the seller has no completed orders that day. - `daily_revenue` is the sum of `amount` for completed orders on that UTC date; use `0` on a missing date. - `rolling_7d_revenue` is the sum of `daily_revenue` for the current calendar date and the previous six calendar dates. Before a seller has seven dates of history, use all dates available since that seller's series began. - Cancelled and refunded orders contribute nothing. Return exactly these columns: - `order_date` - `seller_id` - `daily_revenue` - `rolling_7d_revenue` The window is based on calendar dates, not the previous seven order rows. Preserve numeric precision and order the final result by `seller_id` ascending and `order_date` ascending.