WHERE clause in nested query is not improving efficiency - mysql

Question

I have a table where I am running a pretty query intensive RANK window function in MYSQL.
The query looks like this:
SELECT
t1.*,
RANK () OVER (
PARTITION BY
t1.`country`,
t1.`product_id`,
t1.`retrieveDate`,
t1.`retrieveHour`
ORDER BY
t1.`retrieveDatetime` DESC) AS `ranking`
FROM (
SELECT * FROM product_data
WHERE retrieveDate > (CURRENT_DATE() - INTERVAL 1 WEEK)
) t1
WHERE t1.ranking = 1
I originally set up a WHERE filter in the nested query to limit the amount of data to run the actual query but from running EXPLAIN I noticed that no matter what I set with the retrieveDate filter (1 WEEK, 1 MONTH, 2 MONTH), it does not improve performance whatsoever.
This is the EXPLAIN output:
{
"query_block": {
"select_id": 1,
"cost_info": {
"query_cost": "355813.66"
},
"windowing": {
"windows": [
{
"name": "<unnamed window>",
"using_filesort": true,
"filesort_key": [
"`product_id`",
"`country`",
"`retrieveDate`",
"`retrieveHour`",
"`retrieveDatetime` desc"
],
"functions": [
"rank"
]
}
],
"cost_info": {
"sort_cost": "271814.81"
},
"table": {
"table_name": "product_data",
"access_type": "ALL",
"rows_examined_per_scan": 815526,
"rows_produced_per_join": 271814,
"filtered": "33.33",
"cost_info": {
"read_cost": "2446.25",
"eval_cost": "27181.48",
"prefix_cost": "83998.85",
"data_read_per_join": "553M"
},
"used_columns": [
"product_id",
"country",
"category",
"rank",
"primaryCategory",
"primaryCategoryRank",
"retrieveDatetime",
"createdAt",
"updatedAt",
"retrieveDate",
"retrieveHour"
],
"attached_condition": "(`product_data`.`retrieveDate` > <cache>((curdate() - interval 1 week)))"
}
}
}
}
Any thoughts on how to actually improve this query?
I currently have the following indexes set up:
BTree index
product_id, type=varchar, pos=1
country, type=int, pos=2
retrieveDate, type=date, pos=13
retrieveHour, type=int, pos=14
retrieveDatetime, type=datetime, pos=10
as well as one standalone retrieveDatetime index

Answer 1

As far as I can see, in the last line of your SQL t1.ranking = 1 won't run, because ranking isn't a column in the table product_data (unless it coincidentally is at which point you've might've found your problem, lol). Assuming that's just a typo. I'm also going to assume that if you run:
SELECT * FROM product_data WHERE retrieveDate > (CURRENT_DATE() - INTERVAL 1 WEEK)
That it is actually faster than WHERE retrieveDate > (CURRENT_DATE() - INTERVAL 1 MONTH), yeah? If not, then it has nothing to do with your outer query. Can you try it without the subquery at all? I'm wondering whether MySQL is just choosing an bad optimisation strategy because of it.
SELECT
*,
RANK () OVER (
PARTITION BY
`country`,
`product_id`,
`retrieveDate`,
`retrieveHour`
ORDER BY
`retrieveDatetime` DESC) AS `ranking`
FROM product_data
WHERE retrieveDate > (CURRENT_DATE() - INTERVAL 1 WEEK) AND ranking = 1;