Question

I have two tables x and y

Table x has 24 columns each of which holds a different Id value

Table Y has an Id value column and a description

I need to write a procedure, query or view that returns the descriptions for each of the 24 Id values contained in table x in the best performing way possible.

I’ve written a view that calls a function 24 times. The function returns the description based on the Id provided.
Whilst this works it doesn’t perform particularly well.

Is there a technique I should use where this number of descriptions are required from a single table?

Here is the definition for table x (with non relevant columns removed for clarity)

    [DefinitiveHLATypeId] [int] IDENTITY(1,1) NOT NULL,
[PersonId] [int] NOT NULL,
[A_X] [int] NULL,
[A_Y] [int] NULL,
[B_X] [int] NULL,
[B_Y] [int] NULL,
[Bw_X] [int] NULL,
[Bw_Y] [int] NULL,
[C_X] [int] NULL,
[DRB1_X] [int] NULL,
[DRB1_Y] [int] NULL,
[DRB3_X] [int] NULL,
[DRB3_Y] [int] NULL,
[DRB4_X] [int] NULL,
[DRB4_Y] [int] NULL,
[DRB5_X] [int] NULL,
[DRB5_Y] [int] NULL,
[DQA_X] [int] NULL,
[DQA_Y] [int] NULL,
[DQB_X] [int] NULL,
[DQB_Y] [int] NULL,
[DPA1_X] [int] NULL,
[DPA1_Y] [int] NULL,
[DPB1_X] [int] NULL,
[DPB1_Y] [int] NULL

Here is the definition for table y (with non relevant columns removed for clarity)

    [AntigenId] [int] IDENTITY(1,1) NOT NULL,
[AntigenDescription] [varchar](2000) NOT NULL

The relationship between the two tables is between the _X & _Y columns in table x and the AntigenId column in table Y

I need to return the Antigen Description for each of the _X & _Y columns in table x.

Answer 1

Without seeing any sample data in your tables, if you have a table that have 24 columns of data that you need to compare to another table there are a few ways to do this.

First, you could perform multiple joins on the table for each column similar to this:

select *
from tablex x
left join tabley y1
  on x.col1 = y1.id
left join tabley y2
  on x.col2 = y2.id --- add more joins

I don’t know if joining the tables 24 times would be the most efficient, so it might be easier to perform an UNPIVOT on the table with 24 columns and join on that result:

select x.value,
  y.description
from 
(
  select value, col
  from tablex
  unpivot
  (
    value
    for col in (col1, col2, col3, col4, col5)
  ) unpiv
) x
left join tabley y
  on x.value = y.id

See SQL Fiddle with Demo

The UNPIVOT is the same as using a UNION ALL on tablex to transform the data from multiple columns into rows of data which will make it easier to join on:

select *
from
(
  select col1 value, 'col1' col
  from tablex
  union all
  select col2 value, 'col2' col
  from tablex
  union all
  select col3 value, 'col3' col
  from tablex
  union all
  select col4 value, 'col4' col
  from tablex
  union all
  select col5 value, 'col5' col
  from tablex
) x
left join tabley y
  on x.value = y.id

Based on your edit, your query would be similar to this:

select x.value,
  y.AntigenDescription
from 
(
  select DefinitiveHLATypeId, PersonId, value, col
  from tablex
  unpivot
  (
    value
    for col in (A_X, A_Y, B_X, B_Y, Bw_X, Bw_Y,
                    C_X, DRB1_X, DRB1_Y, DRB3_X, DRB3_Y,
                    DRB4_X, DRB4_Y, DRB5_X, DRB5_Y, DQA_X, DQA_Y,
                    DQB_X, DQB_Y, DPA1_X, DPA1_Y,
                    DPB1_X, DPB1_Y
  ) unpiv
) x
left join tabley y
  on x.value = y.AntigenId

Note: In the long term my suggestion would be to redesign the tablex because having a table structured in this way will make it difficult to access the data.