mysql GROUB BY idea - mysql

Question

I have the following scenario: there are 1 table with books and two couples of tables (HD/IT) with Sales Order and Purchase Order transactions connecting through Sales Order id.
The table structure follows:
CREATE TABLE `books` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`isbn` varchar(100) COLLATE utf8_unicode_ci NOT NULL,
`it_id` int(11) NOT NULL,
`kind` tinyint(4) NOT NULL,
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci;
INSERT INTO `books` (`id`, `isbn`, `it_id`, `kind`) VALUES
(1, '12345', 1, 1),
(2, '12345', 1, 2),
(3, '67890', 2, 1),
(4, '1111111', 2, 2);
CREATE TABLE `porders_hd` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`dt` date NOT NULL,
`so_id` int(11) DEFAULT NULL,
`customer` int(11) NOT NULL,
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci;
INSERT INTO `porders_hd` (`id`, `dt`, `so_id`, `customer`) VALUES
(1, '2017-07-02', 1, 1),
(2, '2017-08-03', NULL, 3);
CREATE TABLE `porders_it` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`hd_id` int(11) NOT NULL,
`isbn` varchar(100) COLLATE utf8_unicode_ci NOT NULL,
`dscr` varchar(100) COLLATE utf8_unicode_ci NOT NULL,
`qty` int(11) NOT NULL,
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci;
INSERT INTO `porders_it` (`id`, `hd_id`, `isbn`, `dscr`, `qty`) VALUES
(1, 1, '12345', 'Book 1', 1),
(2, 2, '1111111', 'Book 2', 1);
CREATE TABLE `sorders_hd` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`dt` date NOT NULL,
`customer` int(11) NOT NULL,
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci;
INSERT INTO `sorders_hd` (`id`, `dt`, `customer`) VALUES
(1, '2017-07-01', 1),
(2, '2017-08-01', 2);
CREATE TABLE `sorders_it` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`hd_id` int(11) NOT NULL,
`isbn` varchar(100) COLLATE utf8_unicode_ci NOT NULL,
`dscr` varchar(100) COLLATE utf8_unicode_ci NOT NULL,
`qty` int(11) NOT NULL,
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci;
INSERT INTO `sorders_it` (`id`, `hd_id`, `isbn`, `dscr`, `qty`) VALUES
(1, 1, '12345', 'Book 1', 1),
(2, 2, '67890', 'Book 2', 1);
In summary there are:
* 1 Sales Order (#1) also existing in the Purchase Order (#1)
* 1 Sales Order (#2) still pending
* 1 Purchase Order (#2) created without a Sales Order
I want to be able to grab all Sales and Purchases Order per book's isbn and the connected SO and PO must be in the same line. The output must be like the one below:
so_id so_date po_id po_date isbn dscr
NULL NULL 2 2017-08-03 1111111 Book 2
1 2017-07-01 1 2017-07-02 12345 Book 1
2 2017-08-01 NULL NULL 67890 Book 3
I tried to grab the rows using a query like the one below:
SELECT
GROUP_CONCAT(so_id) so_id,
GROUP_CONCAT(so_date) so_date,
GROUP_CONCAT(po_id) po_id,
GROUP_CONCAT(po_date) po_date,
isbn,
dscr
FROM (
SELECT
hd.so_id so_id,
NULL so_date,
hd.id po_id,
hd.dt po_date,
bk.isbn,
it.dscr
FROM porders_hd hd,
porders_it it,
books bk
WHERE it.hd_id = hd.id
AND bk.isbn = it.isbn
AND kind = 2
UNION
SELECT
hd.id so_id,
hd.dt so_date,
NULL po_id,
NULL po_date,
bk.isbn,
it.dscr
FROM sorders_hd hd,
sorders_it it,
books bk
WHERE it.hd_id = hd.id
AND bk.isbn = it.isbn
AND kind = 1
) as table1
GROUP BY isbn, so_id, po_id
but since there is info missing I get the following result:
so_id so_date po_id po_date isbn dscr
NULL NULL 2 2017-08-03 1111111 Book 2
1 2017-07-01 NULL NULL 12345 Book 1
1 NULL 1 2017-07-02 12345 Book 1
2 2017-08-01 NULL NULL 67890 Book 3
Any ideas how can I achieve this ?

Answer 1

I think this is what you're after, but I can;t figure out the role of kind from your code. But here is a query that for each books, gets the associated po line item, finds the corresponding so line item and joins the header rows so the dates are available. Note my assumption that a sales order can't exist with a corresponding PO.
SELECT books.isbn, books.descr, sorders_hd.id, sorders_hd.dt, porders_hd.id, porders_hd.dt
FROM book
join porders_it on porders_it.isbn = books.isbn
join porders_hd on porders_hd.id = porders_it.hd_id
left outer join sorders_it on sorders_it.hd_id=porders_hd.so_id and sorders_it.isbn = porders_it.isbn
left outer join sorders_hd on sorders_hd.id = sorders_it.hd_it
You could normalize your tables so that descr need not be repeated, and also use the book.id in the other tables rather than isbn.

Answer 2

I'm adding a new answer because the previous one and the comments are illustrative. Based on that discussion, this requires a FULL OUTER JOIN which must be emulated by UNION ALL in mysql (which may be what OP was attempting originally).
Here is my new code, taking that into account:
SELECT sorders_hd.id as so_id, sorders_hd.dt as so_dt,
porders_hd.id as po_id, porders_hd.dt as po_dt,
books.isbn, porders_it.dscr
from books
left outer join porders_it on porders_it.isbn=books.isbn
join porders_hd on porders_hd.id=porders_it.hd_id
left outer join sorders_it on sorders_it.isbn=books.isbn and sorders_it.hd_id=porders_hd.so_id
left outer join sorders_hd on sorders_hd.id=sorders_it.hd_id
where books.kind=2
UNION ALL
SELECT sorders_hd.id as so_id, sorders_hd.dt as so_dt,
porders_hd.id as po_id, porders_hd.dt as po_dt,
books.isbn, sorders_it.dscr
from books
left outer join sorders_it on sorders_it.isbn=books.isbn
join sorders_hd on sorders_hd.id=sorders_it.hd_id
left outer join porders_it on porders_it.isbn=books.isbn
left outer join porders_hd on porders_hd.id=porders_it.hd_id and porders_hd.so_id=sorders_hd.id
where porders_hd.id is null and books.kind=1;
The output result is:
so_id so_dt po_id po_dt isbn dscr
1 2017-07-01 1 2017-07-02 12345 Book 1
(null) (null) 2 2017-08-03 1111111 Book 2
2 2017-08-01 (null) (null) 67890 Book 2
See SqlFiddle
The "trick" is to use union all with one of the two queries excluding records that linked both sides (to get the 'right' side of the FULL OUTER JOIN)
+1 to OP for providing the DDL and sample data!
I agree that the data model could be reworked, and could be normalized. The existing model still has at least the problem of a duplicate book record when a sales order and purchase order match (one of them is ignored). It seems to me that one improvement would be to have a master book list and include the id (or isbn if that is the primary key) from that table in porders_it and sorders_it, and eliminate the current books table.