Solution for subquery in MySQL WHERE IN - mysql

Question

I am trying to solve WHERE subquery or to find different solution.
What I am trying to achieve is based on this query:
SELECT c.orig_point_id,
(SELECT attempts
FROM
(SELECT
orig_carrier_id,
orig_point_id,
term_point_id,
term_route,
currency_sell,
is_special,
COUNT(*) AS attempts
FROM cdr
WHERE 1=1
AND start_time >= '2016-10-01 0:00:00'
AND start_time <= '2016-10-31 23:59:59'
GROUP BY orig_carrier_id, currency_sell) AS c0
WHERE c0.orig_carrier_id=3
AND c0.currency_sell="USD"
LIMIT 1) AS attempts,
(SELECT SPLIT(clear_number) as array
FROM
(SELECT
COUNT(*) as total,
clear_number,
orig_carrier_id,
currency_sell
FROM `cdr`
WHERE `start_time`>='2016-10-01 00:00:00'
AND start_time <= '2016-10-31 23:59:59'
GROUP BY `clear_number`
ORDER BY total DESC) AS c0
WHERE c0.orig_carrier_id=3
AND c0.currency_sell="USD"
LIMIT 1) AS splitted_number
FROM cdr AS c
GROUP BY c.orig_carrier_id, c.currency_sell;
SPLIT is a function. Query in that section finds a number(most frequent) and function splits it in ex. 12345,1234,123,12,1. Problem comes when i try to use that as IN subquery. When used directly mysql says functionality not supported. Looks like query is too complex.
When i alias subquery as a workaround it returns NULL, so workaround doesn't work and i believe its returning NULL because of the same reason that its not feasible.
SELECT
CONCAT_WS(" - ",country,region) AS route_name
FROM numbering_plan_external
WHERE
prefix IN(
SELECT array
FROM
(SELECT SPLIT(clear_number) as array
FROM
(SELECT
COUNT(*) as total,
clear_number,
orig_carrier_id,
currency_sell
FROM `cdr`
WHERE `start_time`>='2016-10-01 00:00:00'
AND start_time <= '2016-10-31 23:59:59'
GROUP BY `clear_number`
ORDER BY total DESC) AS c0
WHERE c0.orig_carrier_id=3
AND c0.currency_sell="USD"
LIMIT 1) AS splitted_number)
ORDER BY prefix DESC LIMIT 1) AS top_route
Am I doing anything wrong here, or is there different approach to achieve this. I can leave just split number and later through PHP find the route. It will require lots of queries depending on the results and I am trying to avoid it if possible.
Thanks in advance guys.
Some sample data
CREATE TABLE IF NOT EXISTS `numbering_plan_external` (
`id` int(11) NOT NULL,
`country` varchar(255) NOT NULL,
`region` varchar(255) DEFAULT NULL,
`prefix` varchar(50) NOT NULL,
`is_mobile` tinyint(1) NOT NULL DEFAULT '0',
`last_updated` timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP,
`min_sale_price_currency` char(3) CHARACTER SET latin1 COLLATE latin1_general_ci DEFAULT NULL,
`min_sale_price_amount` decimal(10,4) DEFAULT NULL
) ENGINE=InnoDB AUTO_INCREMENT=14004 DEFAULT CHARSET=latin1;
INSERT INTO `numbering_plan_external`
(`id`, `country`, `region`,`prefix`, `is_mobile`, `last_updated`, `min_sale_price_currency`, `min_sale_price_amount`)
VALUES
(13047, 'Tunisia', '', '216', 0, '2016-02-17 12:30:44', NULL, NULL),
(13048, 'Tunisia', 'Mobile (ORANGE)', '2165', 1, '2016-02-17 12:30:44', NULL, NULL),
(13049, 'Tunisia', 'Mobile (ORASCOM)', '2162', 1, '2016-02-17 12:30:44', NULL, NULL),
(13050, 'Tunisia', 'Mobile (TUNTEL)', '21640', 1, '2016-02-17 12:30:44', NULL, NULL),
(13051, 'Tunisia', 'Mobile (TUNTEL)', '21641', 1, '2016-02-17 12:30:44', NULL, NULL),
(13052, 'Tunisia', 'Mobile (TUNTEL)', '2169', 1, '2016-02-17 12:30:44', NULL, NULL);
CREATE TABLE IF NOT EXISTS `cdr` (
`id` int(11) NOT NULL,
`orig_carrier_id` int(11) NOT NULL,
`orig_point_id` int(11) NOT NULL,
`term_carrier_id` int(11) NOT NULL,
`term_point_id` int(11) NOT NULL,
`clear_number` varchar(100) COLLATE latin1_general_ci NOT NULL,
`is_special` tinyint(1) NOT NULL DEFAULT '0',
`start_time` datetime NOT NULL,
`currency_sell` char(3) COLLATE latin1_general_ci NOT NULL
) ENGINE=InnoDB AUTO_INCREMENT=16385 DEFAULT CHARSET=latin1 COLLATE=latin1_general_ci;
INSERT INTO `cdr`
(`id`, `orig_carrier_id`, `orig_point_id`, `term_carrier_id`, `term_point_id`, `clear_number`, `is_special`, `start_time`, `currency_sell`) VALUES
(1, 3, 5, 0, 0, '21658502507', 0, '2016-10-17 00:02:04', 'USD'),
(2, 3, 5, 0, 0, '21658502507', 0, '2016-10-17 00:02:04', 'USD'),
(3, 3, 5, 0, 0, '21658502507', 0, '2016-10-17 00:03:56', 'USD'),
(4, 3, 5, 0, 0, '21658502507', 0, '2016-10-17 00:09:28', 'USD'),
(5, 3, 5, 0, 0, '21658502507', 0, '2016-10-17 00:16:35', 'USD');

Answer 1

IN considers values as a whole. Whatever your SPLIT() is doing, even though it returns a "csv", that entire list is considered one SINGLE value, e.g. it'll be parsed/executed as the equivalent
WHERE foo IN ('12345,1234,...')
WHERE foo='12345,1234,...'
instead of these
WHERE foo IN ('12345', '1234', '123', ...)
WHERE foo='12345' OR foo='1234' OR ...
You could try using MySQL's find_in_set() instead, which basically does what you want.

Answer 2

Using your sample data with this query:
SELECT
cdr.orig_point_id
, count(cdr.*) attempts
, group_concat(distinct npe.region) regions
FROM cdr
INNER JOIN numbering_plan_external npe
ON cdr.clear_number like concat(npe.prefix,'%') COLLATE latin1_general_ci
AND npe.region <> ''
WHERE cdr.orig_carrier_id=3
AND cdr.currency_sell='USD'
AND cdr.start_time >= '2016-10-01'
AND cdr.start_time < '2016-11-01'
GROUP BY
cdr.orig_point_id
;
Result:
| orig_point_id | attempts | regions |
|---------------|----------|----------------|
| 5 | 5 | Mobile (ORANGE)|
The join between those tables involves comparing the prefix to the starting characters of the clear_number. However you have a collation conflict so you need to specify the collation being used. Using LIKE is not the most efficient style of join condition and it could lead to performance issues as it doesn't make use of indexes. However it does demonstrate that a logical join does exist and that you do not need that split function (which is not good for a join either by the way).
I have left the remainder of my earlier question available for reference:
Query:
SELECT
orig_point_id
, count(*) attempts
FROM cdr
WHERE orig_carrier_id=3
AND currency_sell='USD'
AND start_time >= '2016-10-01'
AND start_time < '2016-11-01'
GROUP BY
orig_point_id
Results:
| orig_point_id | attempts |
|---------------|----------|
| 5 | 5 |
extracted for Original Query:
SELECT c.orig_point_id,
(SELECT attempts
FROM
(SELECT
orig_carrier_id,
orig_point_id,
term_point_id,
/* term_route, */
currency_sell,
is_special,
COUNT(*) AS attempts
FROM cdr
WHERE 1=1
AND start_time >= '2016-10-01 0:00:00'
AND start_time <= '2016-10-31 23:59:59'
GROUP BY orig_carrier_id, currency_sell) AS c0
WHERE c0.orig_carrier_id=3
AND c0.currency_sell="USD"
LIMIT 1) AS attempts
FROM cdr AS c
GROUP BY c.orig_carrier_id, c.currency_sell
Results:
| orig_point_id | attempts |
|---------------|----------|
| 5 | 5 |
Hopefully you can see that you do not need as much complexity in your query as you do now. I suspect that if we know more about the "expected result" we might be able to do it without the split function.