Mysql – How to split a list of number into ranges with a fixed interval with SQL

MySQLpostgresql

Let's say I have a table like this

I want to calculate the frequency ( How many times that product exists in that price range ), in intervals of "50"

So eventually it will give me a table like

Interval for range will be lets pretend a fixed 50
We don't know highest and lowest price of these each products.

So I will run the query and it will give a table as shown above.

I got a query like this but it didn't work. It gives this error ( I cannot edit server config )

This is the query

select product, count(*) as frequency,
       floor(price / 50) as range_start, floor(price / 50) + 50 as range_end
from t 
group by product, floor(price / 50)
order by product, min(price)

This is the error

 [42000][1055] Expression #4 of SELECT list is not in GROUP BY clause and contains nonaggregated column 'price' which is not functionally dependent on columns in GROUP BY clause; this is incompatible with sql_mode=only_full_group_by

After that I tried this ( changed floor(price / 50) to floor(min(price) / 50)

select product, count(*) as frequency,
       floor(min(price) / 50) as range_start, floor(min(price) / 50) + 50 as range_end
from t 
group by product, floor(price / 50)
order by product, min(price)

This query worked but it gives ranges like

Best Answer

select product, count(*) as frequency,
       50 * floor(price / 50) as range_start,
       50 * (floor(price / 50) + 1) as range_end
group by product, 50 * floor(price / 50), 50 * (floor(price / 50) + 1)
order by product, min(price)

Related Solutions

Mysql – Selecting minimum value using a subquery

Modifying slightly your second query, will give you both the merchant id and the lowest price (over all products that pass the conditions - I guess that's what you want):

SELECT p.p_m_id, MIN(p_price) AS min_p_price 
FROM tgmp_affiliates ga 
JOIN tgmp_prices p 
    ON ga.a_code = p.p_gtin 
        AND ga.a_code > '' 
JOIN tgmp_merchants m 
    ON m.m_id = p.p_m_id 
WHERE ga.site_id = '34' 
    AND p.site_id = '34' 
    AND ga.a_parent = '25573' 
    AND p.p_type = 'games' 
    AND m.m_hide = 0 
GROUP BY p.p_m_id ;

Then you can join this - as a derived table - to all the tables that you need data from in the results:

SELECT
    m.*, p.*, ga.*                     -- whatever columns you want  
FROM tgmp_affiliates ga 
JOIN tgmp_prices p 
    ON ga.a_code = p.p_gtin 
        AND ga.a_code > '' 
JOIN tgmp_merchants m 
    ON m.m_id = p.p_m_id 
JOIN
      ( SELECT p.p_m_id, MIN(p_price) AS p_price 
        FROM tgmp_affiliates ga 
        JOIN tgmp_prices p 
            ON ga.a_code = p.p_gtin 
                AND ga.a_code > '' 
        JOIN tgmp_merchants m 
            ON m.m_id = p.p_m_id 
        WHERE ga.site_id = '34' 
            AND p.site_id = '34' 
            AND ga.a_parent = '25573' 
            AND p.p_type = 'games' 
            AND m.m_hide = 0 
        GROUP BY p.p_m_id 
      ) AS tmp
    ON  tmp.p_m_id = p.p_m_id 
    AND tmp.p_price = p.p_price
WHERE ga.site_id = '34' 
    AND p.site_id = '34' 
    AND ga.a_parent = '25573' 
    AND p.p_type = 'games' 
ORDER BY p.p_price ;

Mysql – Storing prices with price history and different specificity (global, user group, single user) in MySQL database

There are a couple of ways you could separate the current values from the historic ones: you could simply include a boolean field that is true for the latest price and filter on that, or you could keep current prices in a separate table. Both options involve a little extra work to maintain integrity but would make a query for current prices more efficient. This doesn't remove the need for the sub-query to pick the first row that matches the user/group/NULL condition, though fewer rows will be scanned in processing both the inner and outer queries.

The following might be more efficient as a straight query if only the current prices are in the base table (add the is_latest=true check or your current date filter to the where and on clauses if you keep the historic values in the same table):

SELECT CASE WHEN usermatch.id IS NOT NULL THEN usermatch.field1 WHEN groupmatch.id IS NOT NULL THEN groupmatch.field1 ELSE nullmatch.field1 END AS field1
     , CASE WHEN usermatch.id IS NOT NULL THEN usermatch.field2 WHEN groupmatch.id IS NOT NULL THEN groupmatch.field2 ELSE nullmatch.field2 END AS field2
     (... and so on ...)
     , CASE WHEN usermatch.id IS NOT NULL THEN usermatch.fieldN WHEN groupmatch.id IS NOT NULL THEN groupmatch.fieldN ELSE nullmatch.fieldN END AS fieldN
FROM   base_prices AS nullmatch
LEFT OUTER JOIN 
       base_prices AS groupmatch ON groupmatch.id=nullmatch.id AND nullmatch.referent_id = @GroupIDHere
LEFT OUTER JOIN 
       base_prices AS usermatch ON usermatch.id=nullmatch.id AND nullmatch.referent_id = @UserIDHere
WHERE nullmatch.referent_id IS NULL

This will only work if you always have a row for each product where referent_id is null, otherwise you could try:

SELECT CASE WHEN usermatch.id IS NOT NULL THEN usermatch.field1 WHEN groupmatch.id IS NOT NULL THEN groupmatch.field1 ELSE nullmatch.field1 END AS field1
     , CASE WHEN usermatch.id IS NOT NULL THEN usermatch.field2 WHEN groupmatch.id IS NOT NULL THEN groupmatch.field2 ELSE nullmatch.field2 END AS field2
     (... and so on ...)
     , CASE WHEN usermatch.id IS NOT NULL THEN usermatch.fieldN WHEN groupmatch.id IS NOT NULL THEN groupmatch.fieldN ELSE nullmatch.fieldN END AS fieldN
FROM   users
CROSS JOIN 
       base_prices AS nullmatch
LEFT OUTER JOIN 
       base_prices AS groupmatch ON groupmatch.id=nullmatch.id AND nullmatch.referent_id = users.groupid
LEFT OUTER JOIN 
       base_prices AS usermatch ON usermatch.id=nullmatch.id AND nullmatch.referent_id = users.id
WHERE users.id = @UserIDHere
AND   nullmatch.referent_id IS NULL

In both cases be careful using this construct in a view because further filtering by the fields brought out via those case statements will not be able to use any indexes you may have defined.

Obviously the case statement per field is a bit of faf but it should stop the query runner searching the table once for every row returned as it might with the sub-query arrangement.

In other databases you might be able to use windowing functions (like ROW_NUMBER) to do this more conveniently and efficiently, but mySQL does not support those to my knowledge.

Best Answer

Related Solutions

Mysql – Selecting minimum value using a subquery

Mysql – Storing prices with price history and different specificity (global, user group, single user) in MySQL database

Related Question