获取会话数据中的滚动订单计数

Question

我有以下table

一位客户在一次会话中购买了两次。

我的目标是为 table 的每一行分配一个订单计数器。

为了达到这个目标，我使用滞后函数来调用最后一个 order_id 和最后一个 order_timestamp:

SELECT 
lag(event_timestamp) over (partition by session_id order by
ecom_data.order_id) as prev_order_timestamp,  
lag(ecom_data.order_id)
over (partition by session_id order by event_timestamp) as
prev_order_number 
From table

我想要的输出是这样的：

问题，我没有得到之前的订单时间。相反，我从上一个事件中得到 event_timestamp。

我的第二个挑战是我不知道如何分配 order_count。我想要的输出是这样的：

理想情况下，这个订单计数应该像真实数据集中那样滚动我不知道每个会话总共有多少订单。可以有 0 - 无限订单。

你能帮忙吗？

谢谢！

Answer 1


### create sample table (helps to introduce these in your questions)
WITH
  base AS (
  SELECT
    'A' AS client_id,
    1 AS session_id,
    'pv' AS event_name,
    NULL AS order_id,
    DATETIME("2022-05-12 10:17:41") AS event_timestamp
  UNION ALL
  SELECT
    'A' AS client_id,
    1 AS session_id,
    'ts' AS event_name,
    NULL AS order_id,
    DATETIME("2022-05-12 10:17:42") AS event_timestamp
  UNION ALL
  SELECT
    'A' AS client_id,
    1 AS session_id,
    'pv' AS event_name,
    NULL AS order_id,
    DATETIME("2022-05-12 10:27:14") AS event_timestamp
  UNION ALL
  SELECT
    'A' AS client_id,
    1 AS session_id,
    'atc' AS event_name,
    NULL AS order_id,
    DATETIME("2022-05-12 10:27:15") AS event_timestamp
  UNION ALL
  SELECT
    'A' AS client_id,
    1 AS session_id,
    'p' AS event_name,
    123 AS order_id,
    DATETIME("2022-05-12 10:30:47") AS event_timestamp
  UNION ALL
  SELECT
    'A' AS client_id,
    1 AS session_id,
    'pv' AS event_name,
    NULL AS order_id,
    DATETIME("2022-05-12 10:30:50") AS event_timestamp
  UNION ALL
  SELECT
    'A' AS client_id,
    1 AS session_id,
    'pv' AS event_name,
    NULL AS order_id,
    DATETIME("2022-05-12 10:31:01") AS event_timestamp
  UNION ALL
  SELECT
    'A' AS client_id,
    1 AS session_id,
    'atc' AS event_name,
    NULL AS order_id,
    DATETIME("2022-05-12 10:31:20") AS event_timestamp
  UNION ALL
  SELECT
    'A' AS client_id,
    1 AS session_id,
    'ts' AS event_name,
    NULL AS order_id,
    DATETIME("2022-05-12 10:31:22") AS event_timestamp
  UNION ALL
  SELECT
    'A' AS client_id,
    1 AS session_id,
    'rv' AS event_name,
    NULL AS order_id,
    DATETIME("2022-05-12 10:31:32") AS event_timestamp
  UNION ALL
  SELECT
    'A' AS client_id,
    1 AS session_id,
    'pv' AS event_name,
    NULL AS order_id,
    DATETIME("2022-05-12 10:31:35") AS event_timestamp
  UNION ALL
  SELECT
    'A' AS client_id,
    1 AS session_id,
    'pv' AS event_name,
    NULL AS order_id,
    DATETIME("2022-05-12 10:32:49") AS event_timestamp
  UNION ALL
  SELECT
    'A' AS client_id,
    1 AS session_id,
    'p' AS event_name,
    456 AS order_id,
    DATETIME("2022-05-12 10:33:35") AS event_timestamp
  UNION ALL
  SELECT
    'A' AS client_id,
    1 AS session_id,
    'pv' AS event_name,
    NULL AS order_id,
    DATETIME("2022-05-12 10:33:48") AS event_timestamp
  UNION ALL
  SELECT
    'A' AS client_id,
    1 AS session_id,
    'tv' AS event_name,
    NULL AS order_id,
    DATETIME("2022-05-12 10:33:50") AS event_timestamp
  UNION ALL
  SELECT
    'B' AS client_id,
    1 AS session_id,
    'pv' AS event_name,
    NULL AS order_id,
    DATETIME("2022-05-12 10:31:50") AS event_timestame
  UNION ALL
  SELECT
    'B' AS client_id,
    1 AS session_id,
    'p' AS event_name,
    123 AS order_id,
    DATETIME("2022-05-12 10:33:50") AS event_timestame
  UNION ALL
  SELECT
    'C' AS client_id,
    1 AS session_id,
    'pv' AS event_name,
    NULL AS order_id,
    DATETIME("2022-05-12 11:13:50") AS event_timestame
  UNION ALL
  SELECT
    'C' AS client_id,
    1 AS session_id,
    'pv' AS event_name,
    NULL AS order_id,
    DATETIME("2022-05-12 11:33:50") AS event_timestame),

  prev_order1 AS (
  SELECT
    *,
    LAG(order_id) OVER (PARTITION BY client_id ORDER BY event_timestamp) AS prev_order_number1
  FROM
    base),

  ### filling in order number using your requested output
  prev_order2 AS (
  SELECT
    *,
    MAX(prev_order_number1) OVER(partition by client_id ORDER BY event_timestamp ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) AS prev_order_number2
  FROM
    prev_order1
  ORDER BY
    event_timestamp )

### inserting order_counter logic
SELECT
  *,
  DENSE_RANK() OVER(partition by client_id ORDER BY prev_order_number2) - 1 AS order_counter
FROM
  prev_order2

考虑边缘情况，也许你想按其他维度进行分区，例如 client_id 与总数 table（如你现在所拥有的）。我以 client_id = B 为例。

获取会话数据中的滚动订单计数

Get a rolling order count into session data

google-bigquery