Counting distinct multi-column patterns - sql-server-2014

Question

I'm using SQL Server 2014 and i nee some help with a hard query.
I have the following table (MyTable). These columns names are just for the example. They are actually totally different from each other.
id int,
col1 int,
col2 int,
..
..
..
col70 int
For each pairs of sequential columns {(col1, col2), (col2_col3)...(col69_col70)}, i need to calculate the following: The number of different pairs that each values has - col_i is the static column, and col_i+1 is the other one. Each value need to be divided by the total amount of records in the table. For example:
col1 | col2
45 | 789
56 | 345
99 | 234
45 | 789
45 | 222
89 | 678
89 | 345
45 | 789
90 | 234
12 | 567
Calculation:
((45, 789)+(45, 222))/10
(56, 345)/10
(99, 234)/10
(45, 789)+(45, 222)/10
(45, 789)+(45, 222)/10
(89, 678)+(89, 345)/10
(89, 678)+(89, 345)/10
((45, 789)+(45, 222))/10
(90, 234)/10
(12, 567)/10
Output:
col1_col2
0.2
0.1
0.1
0.2
0.2
0.2
0.2
0.2
0.1
0.1
Explanation for the first records:
45 is the value of the static column ,so now i'll check how many different combination we can find with col2:
45 | 789
45 | 789
45 | 222
45 | 789
Total distinct combinations divided by number of records in the table: 2/10 = 0.2
This calculation need for each pairs of sequential columns. Any recommendation? Is there's a smart way to calculate it automatically instead of writing a query with line for each pair?

Answer 1

An example assuming you have a primary key:
create table my_table
(column_id int not null,
column1 int not null,
column2 int not null);
insert into my_table
(column_id, column1, column2)
values
(1, 45,789),
(2, 56,345),
(3, 99,234),
(4, 45,789),
(5, 45,222),
(6, 89,678),
(7, 89,345),
(8, 45,789),
(9, 90,234),
(10, 12,567);
declare #column_a as nvarchar(100) = N'column1';
declare #column_b as nvarchar(100) = N'column2';
declare #result_column as nvarchar(100) = N'column1_2';
declare #sql_string as nvarchar(4000)
set #sql_string =
'select a.column_id,
1.0 * count( distinct b.' + #column_b + ') / (count(a.' + #column_a + ') over ()) as ' + #result_column
+ ' from my_table a
inner join my_table b
on a.' + #column_a + ' = b.' + #column_a +
' group by a.column_id, a.' + #column_a +
' order by a.column_id';
-- print #sql_string;
execute(#sql_string);
If there's no primary key you could use the rownumber() function to create an identifier, but the result order would change. The print command can be useful for checking the dynamic sql string, here commented out.
Putting the dynamic SQL into a stored procedure:
create procedure column_freq #column_a nvarchar(100), #column_b nvarchar(100), #result_column nvarchar(100)
as
begin
declare #sql_string as nvarchar(4000);
set #sql_string =
'select a.column_id,
1.0 * count( distinct b.' + #column_b + ') / (count(a.' + #column_a + ') over ()) as ' + #result_column
+ ' from my_table a
inner join my_table b
on a.' + #column_a + ' = b.' + #column_a +
' group by a.column_id, a.' + #column_a +
' order by a.column_id';
execute(#sql_string);
end;
go
exec column_freq N'column1', N'column2', N'column1_2';
go