在T-SQL中为每个组选择最短的列

Question

So my table has an NVARCHAR(MAX) column where each entry (path) is prefixed with an identifier, for example, with N identifiers:

id-1\path\to\stuff
id-1\different\path
...
id-2\path\to\stuff
id-2\different\path
...
id-N\path\to\stuff
id-N\different\path
...

I'm having trouble coming up with a query to select one row per identifier, and each row would be chosen by having the shortest path in it's identifier group, including the other columns in the row.

So the result of the query would be N rows total, one row per identifier, with the rows being chosen on the basis of the shortest overall path length.

I feel like I'm missing something obvious, but I'm not sure what it is.

Answer 1

You need to count the backslashes. You can do that with len() and replace() . So:

select t.*
from  (select t.*, row_number() over (partition by id order by len) as seqnum
       from t cross apply
            ( values (len(path) - len(replace(path, '/', ''))) ) v(len)
      ) t
where seqnum = 1;

The cross apply is just a convenient way of naming an expression for the query. It calculates the number of slashes ("length of path") by comparing the lengths of the path with and without slashes. The row_number() then orders the rows for eah id based on len , with the shortest len getting a value of 1 .