SQL 雪花 - 在一张表中比较 2 个日期

Question

I have the following table

card id | txn id | date | legit
 xys       txq      Mar.1  1
 xys       txs      Mar.8  0
 xys       txt      Mar.12 0
 xyb       txl      Mar.13 0
 xyb       txp      Mar.14 1

I want to be able to compare the date between the first record and any future records for that card id where that future record's legit = 0, and if the first record is within 10 days, show that record. (note this would also need to include the first record if legit = 0).

If there is a better way to think about this problem, I'm all ears. I'm pretty sure dateddiff will come in use here but I'm not sure how to make the comparison between the 1st record and any future record.

Example output would be:

card id | txn id | date | legit
xys       txs      Mar.8  0
xyb       txl     Mar.13  0

WHERE legit = 0...

Answer 1

There might be a more efficient way to do this, but I think this gets you what you are asking for:

WITH x AS (
    SELECT *
    FROM (VALUES 
          ('xys','txq','2020-03-01'::DATE,1),
          ('xys','txs','2020-03-08'::DATE,0),
          ('xys','txt','2020-03-12'::DATE,0),
          ('xyb','txl','2020-03-13'::DATE,0),
          ('xyb','txp','2020-03-14'::DATE,1)) as y (card_id, txn_id, date_, legit)
    )
SELECT card_id,
    CASE WHEN legit = 0 THEN txn_id
         WHEN DATEDIFF('d',date_next,date_) <= 10 THEN txn_id_next
         ELSE txn_id END as txn_id,
    CASE WHEN legit = 0 THEN date_
         WHEN DATEDIFF('d',date_next,date_) <= 10 THEN date_next
         ELSE date_ END as date_,   
    CASE WHEN legit = 0 THEN legit
         WHEN DATEDIFF('d',date_next,date_) <= 10 THEN legit_next
         ELSE legit END as legit
FROM (
  SELECT *,
      ROW_NUMBER() OVER (PARTITION BY card_id ORDER BY date_) as row_num,
      LEAD(txn_id) OVER (PARTITION BY card_id ORDER BY date_) as txn_id_next,
      LEAD(date_) OVER (PARTITION BY card_id ORDER BY date_) as date_next,
      LEAD(legit) OVER (PARTITION BY card_id ORDER BY date_) as legit_next
  FROM x
  QUALIFY row_num = 1
  );

This first gets the first record and the "next record" and moves them to a single line, and then the CASE statements determine whether to use the first record or the "next" record based on the rules that I think you specified.

Answer 2

I want to be able to compare the date between the first record and any future records for that card id where that future record's legit = 0, and if the first record is within 10 days, show that record.

If I understand correctly, you can use window functions:

select t.*
from (select t.*, min(date) over (partition by cardid) as min_date
      from t
     ) t
where legit = 0 and date < min_date + interval '10 day';