Need partition by three field and ranking

问题

I have a query which fetches result by partitioning counterid, status, source as rank.

Using query as

select ch.* except (seqnum, seqnum_s),
           row_number() over (partition by counter_id, status,source, (seqnum_s - seqnum)
                              order by transaction_time
                             ) as ranking
    from (select ch.*,
                 row_number() over (partition by counter_id order,status by transaction_time) as seqnum,
                 row_number() over (partition by counter_id, status,source order by transaction_time) as seqnum_s
          from counter_history ch
         ) ch;

but I need a result as given below.

CounterID   status    transaction_time      source      Rank
121         new       Nov-09-2019 8:32:19   UNDEFINED   1
121         new       Nov-09-2019 8:32:20   UNDEFINED   2
121         CLAIMED   Nov-09-2019 8:32:21   WORKFLOW    1
121         new       Nov-09-2019 8:32:22   WORKFLOW    1
121         CLAIMED   Nov-09-2019 8:32:59   external    1
121         CLAIMED   Nov-09-2019 8:33:00   external    2
121         CLAIMED   Nov-09-2019 8:33:01   external    3
233         new       Nov-09-2019 8:33:32   external    1
233         new       Nov-09-2019 8:49:58   internal    1
233         RESOLVED  Nov-13-2019 6:51:11   internal    1
233         CLAIMED   Nov-11-2019 22:15:52  WORKFLOW    1
233         new       Nov-11-2019 22:15:54  external    1
233         CLAIMED   Nov-11-2019 22:15:55  external    1
156         new       Nov-11-2019 22:15:56  UNDEFINED   1
156         CLAIMED   Nov-11-2019 22:16:24  WORKFLOW    1
156         CLAIMED   Nov-11-2019 22:16:28  external    2
156         RESOLVE   Nov-11-2019 22:16:56  external    1

Kindly help out if there is any issue in the query. using #standardsql Thanks

回答1:

Below is for BigQuery Standard SQL

#standardSQL
WITH `project.dataset.counter_history` AS (
  SELECT 121 CounterID,         'new' status,       'Nov-09-2019 8:32:19' transaction_time,   'UNDEFINED' source UNION ALL
  SELECT 121,         'new',       'Nov-09-2019 8:32:20',   'UNDEFINED'   UNION ALL
  SELECT 121,         'CLAIMED',   'Nov-09-2019 8:32:21',   'WORKFLOW'    UNION ALL
  SELECT 121,         'new',       'Nov-09-2019 8:32:22',   'WORKFLOW'    UNION ALL
  SELECT 121,         'CLAIMED',   'Nov-09-2019 8:32:59',   'external'    UNION ALL
  SELECT 121,         'CLAIMED',   'Nov-09-2019 8:33:00',   'external'    UNION ALL
  SELECT 121,         'CLAIMED',   'Nov-09-2019 8:33:01',   'external'    UNION ALL
  SELECT 233,         'new',       'Nov-09-2019 8:33:32',   'external'    UNION ALL
  SELECT 233,         'new',       'Nov-09-2019 8:49:58',   'internal'    UNION ALL
  SELECT 233,         'RESOLVED',  'Nov-13-2019 6:51:11',   'internal'    UNION ALL
  SELECT 233,         'CLAIMED',   'Nov-11-2019 22:15:52',  'WORKFLOW'    UNION ALL
  SELECT 233,         'new',       'Nov-11-2019 22:15:54',  'external'    UNION ALL
  SELECT 233,         'CLAIMED',   'Nov-11-2019 22:15:55',  'external'    UNION ALL
  SELECT 156,         'new',       'Nov-11-2019 22:15:56',  'UNDEFINED'   UNION ALL
  SELECT 156,         'CLAIMED',   'Nov-11-2019 22:16:24',  'WORKFLOW'    UNION ALL
  SELECT 156,         'CLAIMED',   'Nov-11-2019 22:16:28',  'external'    UNION ALL
  SELECT 156,         'RESOLVE',   'Nov-11-2019 22:16:56',  'external'  
), temp AS (
  SELECT * REPLACE(PARSE_TIMESTAMP('%b-%d-%Y %T', transaction_time) AS transaction_time)
  FROM `project.dataset.counter_history`
)
SELECT * EXCEPT(flag, grp), 
  ROW_NUMBER() OVER(PARTITION BY CounterID, grp ORDER BY transaction_time) ranking 
FROM (
  SELECT *, COUNTIF(flag) OVER(PARTITION BY CounterID ORDER BY transaction_time) grp 
  FROM (
    SELECT *, (status, source) != LAG((status, source)) OVER(PARTITION BY CounterID ORDER BY transaction_time) flag
    FROM temp
  )
)
-- ORDER BY CounterID, transaction_time

As you can see above also takes care of translating transaction_time presented as STRING data type to DATE, so ordering is done properly. without this output will be wrong!!

回答2:

It looks like you are starting over again for each counterid/source/status combination. For that, tweak the query like this:

select ch.* except (seqnum, seqnum_s),
           row_number() over (partition by counter_id, status, source, (seqnum_s - seqnum)
                              order by transaction_time
                             ) as ranking
    from (select ch.*,
                 row_number() over (partition by counter_id order by transaction_time) as seqnum,
                 row_number() over (partition by counter_id, status, source order by transaction_time) as seqnum_s
          from counter_history ch
         ) ch;

The only difference from your query is that status and source are in the second row_number() and the first still counts only by counter_id.

来源：https://stackoverflow.com/questions/59416933/need-partition-by-three-field-and-ranking

标签

sql

google-bigquery