ruby-on-rails - Postgresql 使用原始 sql 填充缺失的日期

Question

我有下表(简化如下):

订单:

<id: 1, shipping: 6.0, price: 20.0>
<id: 2, shipping: 10.0, price: 30.0>
<id: 3, shipping: 7.0, price: 12.0>
<id: 4, shipping: 5.0, price: 0.0> #0 dollars because it was updated after return

销售:

<id: 1, order_id: 1, price:10.0, qty:2, date: "2020-06-01T01:16:15-04:00">
<id: 2, order_id: 1, price:9.0, qty: 1, date: "2020-06-01T01:16:15-04:00">
<id: 3, order_id: 2, price:15.0, qty:2, date: "2020-06-01T01:23:53-04:00">
<id: 4, order_id: 3, price:4.0, qty: 1, date: "2020-06-01T20:28:18-04:00">
<id: 5, order_id: 3, price:4.0, qty: 2, date: "2020-06-01T20:31:15-04:00">
<id: 6, order_id: 4, price:29.0, qty:1, date: "2020-06-03T20:16:15-04:00">

退款:

<id: 1, order_id: 1, qty:1, amount: 9.0, date: "2020-06-01T01:23:15-04:00">
<id: 2, order_id: 4, qty:1, amount: 29.0, date: "2020-06-04T03:34:53-04:00">

我正在编写原始sql来计算运费(即sum(orders.shipping))、总订单(即COUNT(DISTINCTorders.id))和净销售额(即 sales.price * sales.qty - COALESCE(refunds.refund_amount, 0))按天分组。搜索将采用格式为 YYYY-MM-DDThh24:mi:ss 的 min_date 和 max_date 来过滤销售或退款不在日期范围内。我遇到的问题是使用 generate_series 添加表中不存在的所有日期，并将值全部设置为 0。因此，如果 min_date = 2020-06-01T00，则为示例响应:00:00 和 max_date = 2020-06-05T23:59:59 类似于:

"2020-06-01": {shipping: 6, total_orders: 3, net_sales: 62.0},
"2020-06-02": {shipping: 0, total_orders: 0, net_sales: 0}, --> newly added
"2020-06-03": {shipping: 5, total_orders: 1, net_sales: 29},
"2020-06-04": {shipping: 0, total_orders: 1, net_sales: -29.0},
"2020-06-05": {shipping: 0, total_orders: 0, net_sales: 0} --> newly added.

谁能帮我得到上面想要的结果。我看过一些例子，但我无法让它适合我的场景。谢谢!

Answer 1

我认为这会做你想要的:

select 
    d.dt, 
    o.shipping,
    s.total_orders,
    coalesce(s.sales_amount, 0) - coalesce(r.refound_amount, 0) net_sales
from generate_series(?::timestamp, ?::timestamp, interval '1 day') d(dt)
left join lateral (
    select 
        count(distinct order_id) total_orders,
        sum(price * quantity) sales_amount,
        array_agg(order_id) order_ids
    from sales s
    where s.date >= d.dt and s.date < d.dt + interval '1 day'
) s on true
left join lateral (
    select sum(o.shipping) shipping
    from orders o
    where o.id = any(s.order_ids)
) o on true
left join lateral (
    select sum(r.amount) refound_amount
    from refunds r
    where r.order_id = any(s.order_ids)
) r on true

查询首先生成给定间隔内的所有日期(? 代表两个日期参数)。

然后，我们使用带有聚合查询的横向联接来获取该期间内发生的所有销售的信息。另一个稍后加入带来与第一个横向加入选择的order_id相对应的运费，另一个带来相应的退款。