如何比较文本随时间的变化？

Question

我不太清楚如何捕捉随着时间的推移发生的变化。我有以下内容：

db<>小提琴

CREATE TABLE term_changes (load_id int, merchant_id int, load_date date, terms varchar(50));

INSERT INTO term_changes (load_id, merchant_id, load_date, terms) VALUES 
  (1, 1, '2023-01-05', 'Roses are red'),
  (2, 2, '2023-01-05', 'Roses are blue'),
  (3, 1, '2023-01-06', 'Roses are red'),
  (4, 2, '2023-01-06', 'Roses are blue'),
  (5, 1, '2023-01-07', 'Roses are violet'),
  (6, 2, '2023-01-07', 'Roses are blue'),
  (7, 1, '2023-01-08', 'Roses are violet'),
  (8, 2, '2023-01-08', 'Roses are yellow');

WITH t1 AS (SELECT load_id, merchant_id, load_date, MD5(terms) AS terms
            FROM term_changes
            ORDER BY merchant_id, load_id),
t2 AS (SELECT load_id, 
              merchant_id, 
              load_date, 
              terms, 
              LAG(load_id, 1) OVER (PARTITION BY merchant_id 
                                    ORDER BY load_id) AS prev_load_id
       FROM t1)
SELECT *
FROM   t2 
       JOIN t1 ON t1.load_id = t2.prev_load_id
                  AND t1.merchant_id = t2.merchant_id
                  AND t1.terms != t2.terms

哪个回报

加载_id	商户id	加载日期	条款	prev_load_id	加载_id	商户id	加载日期	条款
5	1	2023-01-07	84df2c2124ad3fc5c8cdf76ce1d7f3e3	3	3	1	2023-01-06	6becb043847fefb01e7989034cbdb136
8	2	2023-01-08	90b64ad67da829652ee622e0695748fc	6	6	2	2023-01-07	0e3eb8ff97b31e8874f7b51c23f242a2

我之后的输出通过merchant_id捕获当前和

更改

值：

加载_id	商户id	加载日期	条款
1	1	2023-01-05	玫瑰是红色的
5	1	2023-01-07	玫瑰是紫罗兰色的
2	2	2023-01-05	玫瑰是蓝色的
8	2	2023-01-08	玫瑰是黄色的

• 使用

  load_id

  ，因为每个日期可能有多个条目

Answer 1

您可以使用此查询：

WITH t1 AS (
  SELECT 
    load_id, 
    merchant_id, 
    load_date, 
    terms, 
    LAG(terms) OVER (PARTITION BY merchant_id ORDER BY load_date) AS prev_terms
  FROM term_changes
),
t2 AS (
  SELECT 
    load_id, 
    merchant_id, 
    load_date, 
    terms 
  FROM t1 
  WHERE prev_terms IS NULL OR prev_terms != terms
)
SELECT *
FROM t2
ORDER BY merchant_id, load_date;

小提琴

Answer 2

看起来我只需要更改为

LEFT JOIN

移动谓词：

CREATE TABLE term_changes (load_id int, merchant_id int, load_date date, terms varchar(50));

INSERT INTO term_changes (load_id, merchant_id, load_date, terms) VALUES 
  (1, 1, '2023-01-05', 'Roses are red'),
  (2, 2, '2023-01-05', 'Roses are blue'),
  (3, 1, '2023-01-06', 'Roses are red'),
  (4, 2, '2023-01-06', 'Roses are blue'),
  (5, 1, '2023-01-07', 'Roses are violet'),
  (6, 2, '2023-01-07', 'Roses are blue'),
  (7, 1, '2023-01-08', 'Roses are violet'),
  (8, 2, '2023-01-08', 'Roses are yellow'),
  (9, 3, '2023-01-05', 'Roses are green'),
  (10, 3, '2023-01-06', 'Roses are green'),
  (11, 3, '2023-01-07', 'Roses are green'),
  (12, 3, '2023-01-08', 'Roses are green');


WITH t1 AS (SELECT load_id, 
                   merchant_id, 
                   load_date, 
                   terms, 
                   MD5(terms) AS terms_hashed
            FROM term_changes
            ORDER BY merchant_id, load_id),
t2 AS (SELECT load_id, 
              merchant_id, 
              load_date, 
              terms, 
              terms_hashed, 
              LAG(load_id, 1) OVER (PARTITION BY merchant_id 
                                    ORDER BY load_id) AS prev_load_id
       FROM t1)
SELECT t2.load_id,
       t2.merchant_id,
       t2.load_date,
       t2.terms
FROM   t2 
       LEFT JOIN t1 ON t1.load_id = t2.prev_load_id
                       AND t1.merchant_id = t2.merchant_id
WHERE  ( t1.terms_hashed != t2.terms_hashed
         OR t2.prev_load_id IS NULL )

加载_id	商户id	加载日期	条款
1	1	2023-01-05	玫瑰是红色的
5	1	2023-01-07	玫瑰是紫罗兰色的
2	2	2023-01-05	玫瑰是蓝色的
8	2	2023-01-08	玫瑰是黄色的
9	3	2023-01-05	玫瑰是绿色的

小提琴

Answer 3

你让事情变得比需要的复杂得多。

小提琴

select load_id, merchant_id, load_date, terms, unchanged
from (
    select load_id, merchant_id, load_date, terms,
        coalesce(terms = lag(terms) over (partition by merchant_id order by load_id), 0) unchanged
    from term_changes
) term_changes_plus
where not unchanged
order by merchant_id, load_id