MYSQL在工会中选择与别名相同的列名

Question

I have a simple MYSQL query that unions two tables:

SELECT * FROM (
    SELECT col1 AS col1A FROM table1
    UNION 
    SELECT col1 AS col1B FROM table2
) AS t WHERE col1A <> col1B

I have a column called col1 in both tables and I need to select only rows that have a different value of that column so I select them as aliases. When I run this query I got:

Unknown column 'col1B' in 'where clause'

Table1 data:

col1
----
test

Table2 data:

col1
----
test

The query should return no rows as each value in col1 in table1 is equal to each value in col1 in table2 instead it returns that col1 in table2 is unknown although I select it as an alias

Answer 1

I think you need to look up the appropriate usage of UNION . It will return all results from first query combined with all results from the second query. This results in a single dataset, with a single column (not col1 and col2), just col1 in this case.

Assuming you're trying to get all records in table1 that don't exist in table2, you can use NOT EXISTS :

SELECT col1 
FROM table1 t1 
WHERE NOT EXISTS (
    SELECT 1
    FROM table2 t2
    WHERE t1.col1 = t2.col1
    )

Answer 2

Why Error 1054 is being returned by OP query

The error that's being returned is because the name assigned to a column from the result of a UNION is taken from the first SELECT.

You can observe this by running a simple example:

 SELECT 1 AS one 
  UNION
 SELECT 2 AS two 

The resultset returned by that query will contain a single column, the name assigned to the column will be one , the column name from the first SELECT. This explains why you are getting the error from your query.

---

One way to return rows with no match

To return values of col1 from table1 which do not match any value in the col1 column from table2 ...

one option to use an anti-join pattern...

    SELECT t1.col1
      FROM table1 t1
      LEFT
      JOIN table2 t2
        ON t2.col1 = t1.col1
     WHERE t2.col1 IS NULL

The LEFT JOIN operation returns all rows from table1, along with any "matching" rows found in table2. The "trick" is the predicate in the WHERE clause... any "matching" rows from table2 will have a non-NULL value in col1. So, if we exclude all of the rows where we found a match, we're left with rows from table1 that didn't have a match.

If we want to get rows from table2 that don't have a "matching" row in table1 , we can do the same thing, just flipping the order of the tables.

If we combine the two sets, but only want a "distinct" list of "not matched" values, we can use the UNION set operator:

    SELECT t1.col1
      FROM table1 t1
      LEFT
      JOIN table2 t2
        ON t2.col1 = t1.col1
     WHERE t2.col1 IS NULL
     UNION 
    SELECT s2.col1
      FROM table2 s2
      LEFT
      JOIN table1 s1
        ON s1.col1 = s2.col1
     WHERE s1.col1 IS NULL

--

Finding out which table the non-matched value is from

Sometimes, we want to know which query returned the value; we can get that by including a literal value as a discriminator in each query.

    SELECT 'table1' AS src
         , t1.col1
      FROM table1 t1
      LEFT
      JOIN table2 t2
        ON t2.col1 = t1.col1
     WHERE t2.col1 IS NULL
     UNION 
    SELECT 'table2' AS src
         , s2.col1
      FROM table2 s2
      LEFT
      JOIN table1 s1
        ON s1.col1 = s2.col1
     WHERE s1.col1 IS NULL
    ORDER BY 2

---

A different (usually less performant) approach to finding non-matching rows

An entirely different approach, to returning an equivalent result, would be do something like this:

SELECT q.col1
  FROM ( SELECT 't1' AS src, t1.col1 FROM table1 t1 
          UNION
         SELECT 't2' AS src, t2.col1 FROM table2 t2
       ) q
 GROUP BY q.col1
HAVING COUNT(DISTINCT q.src) < 2
ORDER BY q.col1

(The inline view q will be "materialized" as a derived table, so this approach can be expensive for large sets, and this approach won't take advantage of indexes on col1 to perform the matching.) One other small difference between this and the anti-join approach: this will omit a col1 value of NULL if a NULL exists in both tables. Aside from that, the resultset is equivalent.