mysql: convert_tz ignored in where clause comparisons?

Question

So I have a healthy table of hourly stats where the day and time are split rather than kept as a singular datetime:

+-------+-------------+------------+----------+
| cakes | pies        | day        | hour     |
+-------+-------------+------------+----------+
|     1 |          28 | 2012-02-21 |       20 |
|     0 |          14 | 2012-02-21 |       21 |
|     1 |          15 | 2012-02-21 |       22 |
|     1 |          11 | 2012-02-21 |       23 |
|     0 |           7 | 2012-02-22 |        0 |
|     0 |           9 | 2012-02-22 |        1 |
|     0 |           5 | 2012-02-22 |        2 |
|     0 |           8 | 2012-02-22 |        3 |
|     1 |          11 | 2012-02-22 |        4 |
|     0 |          11 | 2012-02-22 |        5 |
|     0 |          12 | 2012-02-22 |        6 |
|     1 |          19 | 2012-02-22 |        7 |
|     0 |          26 | 2012-02-22 |        8 |
|     0 |          20 | 2012-02-22 |        9 |
|     0 |          24 | 2012-02-22 |       10 |
|     2 |          26 | 2012-02-22 |       11 |
|     1 |          22 | 2012-02-22 |       12 |
|     1 |          24 | 2012-02-22 |       13 |
|     1 |          32 | 2012-02-22 |       14 |
|     0 |          25 | 2012-02-22 |       15 |
|     2 |          20 | 2012-02-22 |       16 |
|     0 |          24 | 2012-02-22 |       17 |
|     1 |          24 | 2012-02-22 |       18 |
|     0 |          15 | 2012-02-22 |       19 |
+-------+-------------+------------+----------+

I want to do timezone-sensitive selects from this table so people around the world can know about my cake or pie quotas at any hour.

Here's what my query would be without timezone conversion:

select cakes, pies, day, hour, 
    str_to_date(concat(day,' ',hour,':00'), '%Y-%m-%d %H:%i:%s') 'this' from stats where 
    str_to_date(concat(day,' ',hour,':00'), '%Y-%m-%d %H:%i:%s') between 
    str_to_date('2012-02-21 20:00:00', '%Y-%m-%d %H:%i:%s') and 
    str_to_date('2012-02-21 23:00:00', '%Y-%m-%d %H:%i:%s');    

..which returns the first four rows of the dataset above:

+-------+-------------+------------+----------+---------------------+
| cakes | pies        | day        | hour     | this                |
+-------+-------------+------------+----------+---------------------+
|     1 |          28 | 2012-02-21 |       20 | 2012-02-21 20:00:00 |
|     0 |          14 | 2012-02-21 |       21 | 2012-02-21 21:00:00 |
|     1 |          15 | 2012-02-21 |       22 | 2012-02-21 22:00:00 |
|     1 |          11 | 2012-02-21 |       23 | 2012-02-21 23:00:00 |
+-------+-------------+------------+----------+---------------------+

So far so good. Now I need to make this timezone-sensitive. Say my server's in California and someone in New Zealand during daylight savings is trying to access my cake and pie stats from 2012-02-21 20:00:00 through 2012-02-21 23:00:00:

select cakes, pies, day, hour, 
    str_to_date(convert_tz(concat(day,' ',hour,':00'), '+13:00','-8:00'), '%Y-%m-%d %H:%i:%s') 'this' from stats where
    str_to_date(convert_tz(concat(day,' ',hour,':00'), '+13:00','-8:00'), '%Y-%m-%d %H:%i:%s') between 
    str_to_date(convert_tz('2012-02-21 20:00:00', '+13:00','-8:00') , '%Y-%m-%d %H:%i:%s') and 
    str_to_date(convert_tz('2012-02-21 23:00:00', '+13:00','-8:00') , '%Y-%m-%d %H:%i:%s');

But here's where it gets weird:

+-------+-------------+------------+----------+---------------------+
| cakes | pies        | day        | hour     | this                |
+-------+-------------+------------+----------+---------------------+
|     1 |          28 | 2012-02-21 |       20 | 2012-02-20 23:00:00 |
|     0 |          14 | 2012-02-21 |       21 | 2012-02-21 00:00:00 |
|     1 |          15 | 2012-02-21 |       22 | 2012-02-21 01:00:00 |
|     1 |          11 | 2012-02-21 |       23 | 2012-02-21 02:00:00 |
+-------+-------------+------------+----------+---------------------+   

It SEEMS to think it's looking up timezone-adjusted values (the "this" column), but the stats returned are the exact same as those from the non-timezoned query! What's going on here?

Answer 1

You're converting all the dates to compare (both input dates and dates from the database) using convert_tz so no surprises you get the same data back. I think what you probably want is

SELECT cakes, pies, day, hour, 
    CONVERT_TZ(CONCAT(day,' ',hour,':00'), '+13:00','-8:00') AS 'this' 
FROM stats 
WHERE 
    CONVERT_TZ(CONCAT(day,' ',hour,':00'), '+13:00','-8:00')
        BETWEEN 
            '2012-02-21 20:00:00' 
        AND '2012-02-21 23:00:00'

But why are you storing the date and time separately? It would make your queries far easier to store them together as a DATETIME field. Also you don't need to use STR_TO_DATE on values that are already MySQL DATETIME values.