How can I group an (unknown) number of rows into a single row where the set columns determine the grouping?
For example, shift
Ref Name
You might pivot the table using row_number() as a source of column names:
select *
from
(
select ref,
name,
link,
row_number() over (partition by ref, name order by link) rn
from table1
) s
pivot (min (link) for rn in ([1], [2], [3], [4])) pvt
Simply extend the list of numbers if you have more rows.
Live test is @ Sql Fiddle.