CODEWITHSUNDEEP
sqlsql-serverms-accesssequence
THEn
THEn

Reputation: 1938

Checking sequence with SQL query

I have a table orders that keeps all order from all our stores. I wrote a query to check the sequence orders for each store. It looks like that.

select WebStoreID, min(webordernumber), max(webordernumber), count(webordernumber) 
from orders
where ordertype = 'WEB' 
group by WebStoreID

I can check it all orders are present with this query. web ordernumber is number from 1...n.

How can I write query to find missing orders without joining to temporary/different table?

Upvotes: 2

Views: 16894

Answers (4)

Tom Warfield
Tom Warfield

Reputation: 21

If you have the rank() function but not the lag() function (in other words, SQL Server), you can use this (suggested by http://www.sqlmonster.com/Uwe/Forum.aspx/sql-server-programming/10594/Return-gaps-in-a-sequence):

create table test_gaps_in_sequence (x int)
insert into test_gaps_in_sequence values ( 1 )
insert into test_gaps_in_sequence values ( 2 )
insert into test_gaps_in_sequence values ( 4 )
insert into test_gaps_in_sequence values ( 5 )
insert into test_gaps_in_sequence values ( 8 )
insert into test_gaps_in_sequence values ( 9 )
insert into test_gaps_in_sequence values ( 12)
insert into test_gaps_in_sequence values ( 13)
insert into test_gaps_in_sequence values ( 14)
insert into test_gaps_in_sequence values ( 29)

...

 select lower_bound
         , upper_bound
      from (select upper_bound
                 , rank () over (order by upper_bound) - 1 as upper_rank
              from (SELECT x+n as upper_bound
                      from test_gaps_in_sequence
                         , (SELECT 0 n
                            UNION
                            SELECT -1
                           ) T
                     GROUP BY x+n
                    HAVING MAX(n) = -1
                    ) upper_1
            ) upper_2
         , (select lower_bound
                 , rank () over (order by lower_bound) as lower_rank
              from (SELECT x+n as lower_bound
                      from test_gaps_in_sequence
                         , (SELECT 0 n
                            UNION
                            SELECT 1
                           ) T
                     GROUP BY x+n
                    HAVING MIN(n) = 1
                    ) lower_1
            ) lower_2
      where upper_2.upper_rank = lower_2.lower_rank
      order by lower_bound

... or, to include the "outer limits":

select lower_bound
     , upper_bound
  from (select upper_bound
             , rank () over (order by upper_bound) - 1 as upper_rank
          from (SELECT x+n as upper_bound
                  from test_gaps_in_sequence
                     , (SELECT 0 n
                        UNION
                        SELECT -1
                       ) T
                 GROUP BY x+n
                HAVING MAX(n) = -1
                ) upper_1
        ) upper_2
   full join (select lower_bound
             , rank () over (order by lower_bound) as lower_rank
          from (SELECT x+n as lower_bound
                  from test_gaps_in_sequence
                     , (SELECT 0 n
                        UNION
                        SELECT 1
                       ) T
                 GROUP BY x+n
                HAVING MIN(n) = 1
                ) lower_1
        ) lower_2
   on upper_2.upper_rank  = lower_2.lower_rank
     order by coalesce (lower_bound, upper_bound)

Upvotes: 2

Tony Andrews
Tony Andrews

Reputation: 132570

If your database supports analytic functions then you could use a query something like:

select prev+1, curr-1 from
( select webordernumber curr,
         coalesce (lag(webordernumber) over (order by webordernumber), 0) prev
  from   orders
)
where prev != curr-1;

The output will show the gaps e.g.

prev+1 curr-1
------ ------
     3      7

would mean that numbers 3 to 7 inclusive are missing.

Upvotes: 1

Andomar
Andomar

Reputation: 238076

You could join the table on itself to detect rows which have no previous row:

select cur.*
from orders cur
left join orders prev 
    on cur.webordernumber = prev.webordernumber + 1
    and cur.webstoreid = prev.webstoreid
where cur.webordernumber <> 1
and prev.webordernumer is null

This would detect gaps in the 1...n sequence, but it would not detect duplicates.

Upvotes: 6

Alex Martelli
Alex Martelli

Reputation: 881595

I would make an auxiliary table of "all integers from 1 to n" (see http://www.sql-server-helper.com/functions/integer-table.aspx for some ways to make it with a SQL Server function, but since it's something you will need over and over I'd make it into a real table anyway, and with any SQL engine it's easy to make that, just once) then use a nested query, SELECT value FROM integers WHERE value NOT IN (SELECT webordernumber FROM orders) &c. Also see http://www.sqlmag.com/Article/ArticleID/99797/sql_server_99797.html for a problem similar to yours, "detecting gaps in a sequence of numbers".

Upvotes: 4

Related Questions