Correct way to select from two tables in SQL Server with no common field to join on

Question

Back in the old days, I used to write select statements like this:

SELECT 
table1.columnA, table2.columnA

FROM
table1, table2

WHERE
table1.columnA = 'Some value'

However I was told that having comma separated table names in the "FROM" clause is not ANSI92 compatible. There should always be a JOIN statement.

This leads to my problem.... I want to do a comparison of data between two tables but there is no common field in both tables with which to create a join. If I use the 'legacy' method of comma separated table names in the FROM clause (see code example), then it works perfectly fine. I feel uncomfortable using this method if it is considered wrong or bad practice.

Anyone know what to do in this situation?

Extra Info:

Table1 contains a list of locations in 'geography' data type Table2 contains a different list of 'geography' locations

I am writing select statement to compare the distances between the locations. As far I know you cant do a JOIN on a geography column??

Answer 1

You can (should) use CROSS JOIN. Following query will be equivalent to yours:

SELECT 
   table1.columnA
 , table2.columnA
FROM table1 
CROSS JOIN table2
WHERE table1.columnA = 'Some value'

or you can even use INNER JOIN with some always true conditon:

FROM table1 
INNER JOIN table2 ON 1=1

Answer 2

Cross join will help to join multiple tables with no common fields.But be careful while joining as this join will give cartesian resultset of two tables. QUERY:

SELECT 
   table1.columnA
 , table2,columnA
FROM table1 
CROSS JOIN table2

Alternative way to join on some condition that is always true like

SELECT 
   table1.columnA
 , table2,columnA
FROM table1 
INNER JOIN table2 ON 1=1

But this type of query should be avoided for performance as well as coding standards.

Answer 3

A suggestion - when using cross join please take care of the duplicate scenarios. For example in your case:

• Table 1 may have >1 columns as part of primary keys(say table1_id, id2, id3, table2_id)
• Table 2 may have >1 columns as part of primary keys(say table2_id, id3, id4)

since there are common keys between these two tables (i.e. foreign keys in one/other) - we will end up with duplicate results. hence using the following form is good:

WITH data_mined_table (col1, col2, col3, etc....) AS
SELECT DISTINCT col1, col2, col3, blabla
FROM table_1 (NOLOCK), table_2(NOLOCK))
SELECT * from data_mined WHERE data_mined_table.col1 = :my_param_value