I need help regarding JOIN query in mysql - mysql

Question

I have started learning MySQL and I'm having a problem with JOIN.
I have two tables: purchase and sales
purchase
--------------
p_id date p_cost p_quantity
---------------------------------------
1 2014-03-21 100 5
2 2014-03-21 20 2
sales
--------------
s_id date s_cost s_quantity
---------------------------------------
1 2014-03-21 90 9
2 2014-03-22 20 2
I want these two tables to be joined where purchase.date=sales.date to get one of the following results:
Option 1:
p_id date p_cost p_quantity s_id date s_cost s_quantity
------------------------------------------------------------------------------
1 2014-03-21 100 5 1 2014-03-21 90 9
2 2014-03-21 20 2 NULL NULL NULL NULL
NULL NULL NULL NULL 2 2014-03-22 20 2
Option 2:
p_id date p_cost p_quantity s_id date s_cost s_quantity
------------------------------------------------------------------------------
1 2014-03-21 100 5 NULL NULL NULL NULL
2 2014-03-21 20 2 1 2014-03-21 90 9
NULL NULL NULL NULL 2 2014-03-22 20 2
the main problem lies in the 2nd row of the first result. I don't want the values
2014-03-21, 90, 9 again in row 2... I want NULL instead.
I don't know whether it is possible to do this. It would be kind enough if anyone helps me out.
I tried using left join
SELECT *
FROM sales
LEFT JOIN purchase ON sales.date = purchase.date
output:
s_id date s_cost s_quantity p_id date p_cost p_quantity
1 2014-03-21 90 9 1 2014-03-21 100 5
1 2014-03-21 90 9 2 2014-03-21 20 2
2 2014-03-22 20 2 NULL NULL NULL NULL
but I want 1st 4 values of 2nd row to be NULL

Answer 1

Since there are no common table expressions or full outer joins to work with, the query will have some duplication and instead need to use a left join unioned with a right join;
SELECT p_id, p.date p_date, p_cost, p_quantity,
s_id, s.date s_date, s_cost, s_quantity
FROM (
SELECT *,(SELECT COUNT(*) FROM purchase p1
WHERE p1.date=p.date AND p1.p_id<p.p_id) rn FROM purchase p
) p LEFT JOIN (
SELECT *,(SELECT COUNT(*) FROM sales s1
WHERE s1.date=s.date AND s1.s_id<s.s_id) rn FROM sales s
) s
ON s.date=p.date AND s.rn=p.rn
UNION
SELECT p_id, p.date p_date, p_cost, p_quantity,
s_id, s.date s_date, s_cost, s_quantity
FROM (
SELECT *,(SELECT COUNT(*) FROM purchase p1
WHERE p1.date=p.date AND p1.p_id<p.p_id) rn FROM purchase p
) p RIGHT JOIN (
SELECT *,(SELECT COUNT(*) FROM sales s1
WHERE s1.date=s.date AND s1.s_id<s.s_id) rn FROM sales s
) s
ON s.date=p.date AND s.rn=p.rn
An SQLfiddle to test with.

Answer 2

In a general sense, what you're looking for is called a FULL OUTER JOIN, which is not directly available in MySQL. Instead you only get LEFT JOIN and RIGHT JOIN, which you can UNION together to get essentially the same result. For a very thorough discussion on this subject, see Full Outer Join in MySQL.
If you need help understanding the different ways to JOIN a table, I recommend A Visual Explanation of SQL Joins.
The way this is different from a regular FULL OUTER JOIN is that you're only including any particular row from either table at most once in the JOIN result. The problem being, if you have one purchase record and two sales records on a particular day, which sales record is the purchase record associated with? What is the relationship you're trying to represent between these two tables?
It doesn't sound like there's any particular relationship between purchase and sales records, except that some of them happened to take place on the same day. In which case, you're using the wrong tool for the job. If all you want to do is display these tables side by side and line the rows up by date, you don't need a JOIN at all. Instead, you should SELECT each table separately and do your formatting with some other tool (or manually).

Answer 3

Here's another way to get the same result, but the EXPLAIN for this is horrendous; and performance with large sets is going to be atrocious.
This is essentially two queries UNIONed together. The first query is essentially "purchase LEFT JOIN sales", the second query is essentially "sales ANTI JOIN purchase".
Because there is no foreign key relationship between the two tables, other than rows matching on date, we have to "invent" a key we can join on; we use user variables to assign ascending integer values to each row within a given date, so we can match row 1 from purchase to row 1 from sales, etc.
I wouldn't normally generate this type of result using SQL; it's not a typical JOIN operation, in the sense of how we traditionally join tables.
But, if I had to produce the specified resultset using MySQL, I would do it like this:
SELECT p.p_id
, p.p_date
, p.p_cost
, p.p_quantity
, s.s_id
, s.s_date
, s.s_cost
, s.s_quantity
FROM ( SELECT #pl_i := IF(pl.date = #pl_prev_date,#pl_i+1,1) AS i
, #pl_prev_date := pl.date AS p_date
, pl.p_id
, pl.p_cost
, pl.p_quantity
FROM purchase pl
JOIN ( SELECT #pl_i := 0, #pl_prev_date := NULL ) pld
ORDER BY pl.date, pl.p_id
) p
LEFT
JOIN ( SELECT #sr_i := IF(sr.date = #sr_prev_date,#sr_i+1,1) AS i
, #sr_prev_date := sr.date AS s_date
, sr.s_id
, sr.s_cost
, sr.s_quantity
FROM sales sr
JOIN ( SELECT #sr_i := 0, #sr_prev_date := NULL ) srd
ORDER BY sr.date, sr.s_id
) s
ON s.s_date = p.p_date
AND s.i = p.i
UNION ALL
SELECT p.p_id
, p.p_date
, p.p_cost
, p.p_quantity
, s.s_id
, s.s_date
, s.s_cost
, s.s_quantity
FROM ( SELECT #sl_i := IF(sl.date = #sl_prev_date,#sl_i+1,1) AS i
, #sl_prev_date := sl.date AS s_date
, sl.s_id
, sl.s_cost
, sl.s_quantity
FROM sales sl
JOIN ( SELECT #sl_i := 0, #sl_prev_date := NULL ) sld
ORDER BY sl.date, sl.s_id
) s
LEFT
JOIN ( SELECT #pr_i := IF(pr.date = #pr_prev_date,#pr_i+1,1) AS i
, #pr_prev_date := pr.date AS p_date
, pr.p_id
, pr.p_cost
, pr.p_quantity
FROM purchase pr
JOIN ( SELECT #pr_i := 0, #pr_prev_date := NULL ) prd
ORDER BY pr.date, pr.p_id
) p
ON p.p_date = s.s_date
AND p.i = s.i
WHERE p.p_date IS NULL
ORDER BY COALESCE(p_date,s_date),COALESCE(p_id,s_id)