This is the toughest query I ever made:
http://robertr.pastebin.com/X4bG4pFp
"SELECT `user`.`id` , `user`.`fname` , `user`.`lname` ,
YEAR( `user`.`bday` ) AS `bday_year` , `user`.`class_id` ,
( SELECT `class`.`class_name`
FROM `wp_class_classes` `class`
WHERE `user`.`class_id` = `class`.`id`) AS `class_name`
FROM `wp_class_users` `user`
WHERE MONTH( `bday` ) = $month AND DAY( `bday` ) = $day
OR `user`.`fname` =
( SELECT `name`.`names`
FROM `wp_class_namedays` `name`
WHERE `name`.`day` = '$month.$day'
AND `user`.`fname` = `name`.`names` )
This query grabs data from three different database tables to check if there is someone in the database, who has a party today. And in Latvia we have Name Days too. Anyway, this query works well, and does its job well, but now I want to make it a bit cooler.
I want it to show, who will be having a party next week. You've probably noticed these emails that Facebook sends to you every weekend showing who has a birthday coming up.
But I just can't understand how to get at least that interval?
I remember that PHP has some good functions with which you can find on which day starts month and so on, but maybe here are some bright heart, and willing to help me kick me a bit faster forward.
SELECT
`user`.`id`,
`user`.`fname`,
`user`.`lname` ,
YEAR(`user`.`bday`) AS `bday_year`,
`user`.`class_id`,
(
SELECT
`class`.`class_name`
FROM `wp_class_classes` `class`
WHERE `user`.`class_id` = `class`.`id`
) AS `class_name`,
CASE
WHEN MONTH(`week`.`Date`) = MONTH(`user`.`bday`) AND
DAY(`week`.`Date`) = DAY(`user`.`bday`) THEN 1
ELSE 2
END AS `event_type`
FROM `wp_class_users` `user`
LEFT JOIN `wp_class_namedays` `name` ON `user`.`fname` = `name`.`names`
LEFT JOIN (
SELECT CURDATE() + INTERVAL (1 - DAYOFWEEK(CURDATE())) DAY AS `Date` UNION ALL
SELECT CURDATE() + INTERVAL (2 - DAYOFWEEK(CURDATE())) DAY UNION ALL
SELECT CURDATE() + INTERVAL (3 - DAYOFWEEK(CURDATE())) DAY UNION ALL
SELECT CURDATE() + INTERVAL (4 - DAYOFWEEK(CURDATE())) DAY UNION ALL
SELECT CURDATE() + INTERVAL (5 - DAYOFWEEK(CURDATE())) DAY UNION ALL
SELECT CURDATE() + INTERVAL (6 - DAYOFWEEK(CURDATE())) DAY UNION ALL
SELECT CURDATE() + INTERVAL (7 - DAYOFWEEK(CURDATE())) DAY
) `week`
ON CONCAT(MONTH(`week`.`Date`), '.', DAY(`week`.`Date`)) IN (
CONCAT(MONTH(`user`.`bday`), '.', DAY(`user`.`bday`)),
`name`.`day`
)
WHERE `week`.`Date` IS NOT NULL
The user table is joined with the name day table, and the result set is then compared against the dates of the current week. The final result set lists only those users whose birthdays or name days happen during the week.
If you want to know about the events of, for example, the next week, you can simply change the intervals in the week.Date definitions as 8 - DAYOFWEEK..., 9 - DAYOFWEEK... etc.
One last thing is, instead of the correlated subquery in the select list you could use INNER JOIN, like this:
SELECT
`user`.`id`,
`user`.`fname`,
`user`.`lname` ,
YEAR(`user`.`bday`) AS `bday_year`,
`user`.`class_id`,
`class`.`class_name`
FROM `wp_class_users` `user`
INNER JOIN `wp_class_classes` `class` ON `user`.`class_id` = `class`.`id`
LEFT JOIN `wp_class_namedays` `name` ON ... /* the rest of the above script */
The event_type column as defined above can tell you whether the event is a birthday or not, but it doesn't let you know whether it's both the Birthday and a Name Day for that particular person.
In case you would like to have that distinction, you could change the event_type definition like this:
CASE
WHEN MONTH(`week`.`Date`) = MONTH(`user`.`bday`) AND
DAY(`week`.`Date`) = DAY(`user`.`bday`) THEN 1
ELSE 0
END +
CASE CONCAT(MONTH(`week`.`Date`), '.', DAY(`week`.`Date`))
WHEN `name`.`day` THEN 2
ELSE 0
END AS `event_type`
Now the result of the column would be:
1 – a birthday
2 – a name day
3 – both
Additionally, you could have 'B' instead of 1 and 'N' instead of 2 (and '' instead of 0). The results would be then 'B', or 'N', or 'BN'. Not sure whether + can be used for concatenation, though. If not, put both CASEs into CONCAT().
I'm not sure if I get your query right, but the command SYSDATE() is mentioned in the MySQL docs. You might want to try something like:
... where date = SYSDATE() + 7
(check the syntax, I come from Oracle ;) )
This will get the parties for the next 7 days.
Related
This question already has answers here:
MySQL how to fill missing dates in range?
(6 answers)
MySQL: Select All Dates In a Range Even If No Records Present
(6 answers)
Closed 2 years ago.
I have the following query that returns the dates first_visit starting from today and 7 days back, as well as the visitors hash per day:
SET time_zone= '{$company_timezone}';
SELECT DATE( first_visit ) AS day , COUNT( DISTINCT hash ) AS total
FROM table
WHERE company = 1 and first_visit > SUBDATE( CURDATE( ) , 7 )
GROUP BY day
The flaw with this is that if company = 1 have visitors only today and three days ago, I will get this:
day --------- total
2020-03-08 ----- 30
2020-03-05 ----- 40
leaving out all other dates inbetween.
What I want is to get all the past 7 days, even there are no visitors at all. If there are no visitors, then it should just show 0.
How to edit my query in order to achieve this?
Thank you
Perform an outer join with a derived table that contains desired dates:
select b.date as day, count(distinct hash) as total
from table
right join (select #now := #now - interval 1 day as date from (select #now := curdate()) a, table limit 7) b
on b.date = date(first_visit) and company = 1
group by b.date
This assumes that table has at least 7 rows.
Note: there are two occurrences of table.
If you have data for each day -- but not for that company -- then conditional aggregation is a pretty simply approach:
SELECT DATE( first_visit ) AS day ,
COUNT( DISTINCT CASE WHEN company = 1 THEN hash END ) AS total
FROM table
WHERE first_visit > SUBDATE( CURDATE( ) , 7 )
GROUP BY day;
This only works if all days are represented in your table for some company.
Some solutions involve a table of numbers.
Here is one way with a recursive query, available in MySQL 8.0:
with d as (select 0 n union all select n + 1 where n < 6)
select
current_date - interval n day myday,
count(distinct t.hash) total
from d
left join mytable t
on t.company = 1
and t.first_visit >= current_date - inteval n day
and t.first_visit < current_date - interval (n - 1) day
group by d.n
In earlier version, you can enumerate the numbers as a derived table:
select
current_date - interval n day myday,
count(distinct t.hash) total
from (
select 0 n union all select 1 union all select 2 union all select 3
union all select 4 union all select 5 union all select 6 union all select 7
) d
left join mytable t
on t.compatny = 1
and t.first_visit >= current_date - inteval n day
and t.first_visit < current_date - interval (n - 1) day
group by d.n
I am having a trouble with OR condition inside the SELECT.
I want a simple result if one condition is matched and rest OR condition should not be use.
What i want is:
I have some users shared records and i would like to email them the newest items shared on my website.
For me: Newest Items will be least two days older
Like Today is 9th so i would like to pull all records of 7th. but if i
didn't get any record of 7th then i would like to pull all record of
6th (3 days older from today). if i didn't get any records on 6th then
i would like to pull 1 day older from today.
for all this i have used OR in my SELECT query like this:
SELECT `tg`.* FROM `tblgallery` AS `tg` WHERE (
(tg.added_date BETWEEN '2014-07-07 00:00:00' AND '2014-07-08 00:00:00') OR
(tg.added_date BETWEEN '2014-07-06 00:00:00' AND '2014-07-07 00:00:00') OR
(tg.added_date BETWEEN '2014-07-08 00:00:00' AND '2014-07-09 00:00:00') )
And i have records in my database for dates:
2014-07-06
2014-07-07
and when i run this query it gives me all record of both dates.
But I need to pull only record of 2014-07-07 not of both.(I have mentioned above.)
I know i can do this by using multiple Select and i think that will not be a good idea to request to database again and again.
My Question is : How to pull data from database if the first match is true? and skip all data of rest dates?
OR
Is there any other way to do this?
Please Help
Usually one would just work with LIMIT, which is not applicable here, since there might be many rows per day. What I do is quite similar to LIMIT.
SELECT * FROM (
SELECT
tg.*,
#gn := IF(DATE(tg.added_date) != #prev_date, #gn + 1, #gn) AS my_group_number,
#prev_date := DATE(tg.added_date)
FROM tblgallery tg
, (SELECT #gn := 0, #prev_date := CURDATE()) var_init
ORDER BY FIELD(DATE(tg.added_date), CURDATE() - INTERVAL 1 DAY, CURDATE() - INTERVAL 3 DAY, CURDATE() - INTERVAL 2 DAY) DESC
) sq
WHERE my_group_number = 1;
Here's how it works.
With this line
, (SELECT #gn := 0, #prev_date := CURDATE()) var_init
the variables are initialized.
Then the ORDER BY is important! The FIELD() function sorts the rows from 2 days ago (gets value 3), to 3 days ago (gets value 2), to 1 day ago (gets value 1). Everything else gets value 0.
Then in the SELECT clause the order is also important.
With this line
#gn := IF(DATE(tg.added_date) != #prev_date, #gn + 1, #gn) AS my_group_number,
the variable #gn is incremented when the date of the current row is different from the date of the previous row.
With this line
#prev_date := DATE(tg.added_date)
the date of the current row is assigned to the variable #prev_date. In the line above it still has the value of the previous row.
Now those entries have a 1 in column my_group_number that have the most recent date in the order
2 days ago
3 days ago
yesterday
4 days ago
5 days ago
...
Try this Query:
SELECT GalleryID, PixName, A.added_date
FROM tblGallery A
INNER JOIN (
SELECT added_date FROM tblGallery
WHERE added_date <= DATE_SUB('2014-07-09 00:00:00', interval 2 day)
GROUP BY added_date
ORDER BY added_date DESC
LIMIT 1 ) B
ON A.added_date = B.added_date
See my SQL Fiddle Demo
And even if the date is more than 2 days older it will still work.
See here the Demo below wherein the latest is 4 days older from July 9, 2014
See the 2nd Demo
And if you want the current date instead of literal date like here then you could use CURDATE() function instead. Like one below:
SELECT GalleryID, PixName, A.added_date
FROM tblGallery A
INNER JOIN (
SELECT added_date FROM tblGallery
WHERE added_date <= DATE_SUB(CURDATE(), interval 2 day)
GROUP BY added_date
ORDER BY added_date DESC
LIMIT 1 ) B
ON A.added_date = B.added_date
See 3rd Demo
Well, I'm not being able to solve the multi OR issue but this is how could you get records being added last two days. Change the interval or the CURDATE() in order to fit your needs.
SELECT id, date_added
FROM gallery
WHERE date_added BETWEEN CURDATE() - INTERVAL 2 DAY AND CURDATE()
ORDER BY date_added
Check the SQL Fiddel
It is not about how OR works in MySQL.
I think you are misunderstanding where part by looking at your discussion with #B.T.
It will be executed for each record.
so if one of the record evaluates to false for the first condition then it will evaluate the second condition for that particular record and so on so if any condition evaluates to true by considering all the conditions then that will become part of your result set.
Try this query.
SELECT `tg`.* FROM `tblgallery` AS `tg` WHERE tg.added_date = (
select date (
select distinct(tg.added_date) date from tblgallery as tg
) as t1 order by case
when date between '2014-07-07 00:00:00' AND '2014-07-08 00:00:00'
then 1
when date between '2014-07-06 00:00:00' AND '2014-07-07 00:00:00'
then 2
when date between '2014-07-08 00:00:00' AND '2014-07-09 00:00:00'
then 3
else 4
end limit 1);
Here's what I am doing in this query.
I am getting all the distinct dates.
then I am ordering all the condition in order i.e if first condition is true then 1, if second is true then 2 and so on.
I am limiting the result to 1 so after the order whichever the result is the first row will be selected and which is a date and will be used in the condition.
Note: I have note tested it yes, so you may need to do some changes to the query.
This is my scenario: I have a table that contains events, every event has a field called 'created' with the timestamp in which that event was created. Now I need to sort the events from newest to oldest, but I do not want MySQL to return them all. I need only the latest in a given interval, for example in a range of 24 hours (EDIT: I'd like to have a flexible solution, not only for a 24 hours range, but maybe every few hours). And I only need for the last 10 days. I have achieved that but i'm sure in the most inefficient ways possible, that is, something like that:
$timestamp = time();
for($i = 0; $i < 10; $i++) {
$query = "SELECT * FROM `eventos` WHERE ... AND `created` < '{$timestamp}' ORDER BY `created` DESC LIMIT 1";
$return = $database->query( $query );
if($database->num( $return ) > 0) {
$event = $database->fetch( $return );
$events[] = $event;
$timestamp = $timestamp - 86400;
}
}
I hope I was clear enough. Thanks,
Jesús.
If you have an index with created as the leading column, MySQL may be able to do a reverse scan. If you have a 24 hour period that doesn't have any events, you could be returning a row that is NOT from that period. To make sure you're getting a row in that period, you would really need to include a lower bound on the created column as well, something like this:
SELECT * FROM `eventos`
WHERE ...
AND `created` < FROM_UNIXTIME( {$timestamp} )
AND `created` >= DATE_ADD(FROM_UNIXTIME( {$timestamp} ),INTERVAL -24 HOUR)
ORDER BY `created` DESC
LIMIT 1
I think the big key to performance here is an index with created as the leading column, along with all (or most) of the other columns referenced in the WHERE clause, and making sure that index is used by your query.
If you need a different time interval, down to the second, this approach could be easily generalized.
SELECT * FROM `eventos`
WHERE ...
AND `created` < DATE_ADD(FROM_UNIXTIME({$timestamp}),INTERVAL 0*{$nsecs} SECOND)
AND `created` >= DATE_ADD(FROM_UNIXTIME({$timestamp}),INTERVAL -1*{$nsecs} SECOND)
ORDER BY `created` DESC
LIMIT 1
From your code, it looks like the 24-hour periods are bounded at an arbitrary time... if the time function returns e.g. 1341580800 ('2012-07-06 13:20'), then your ten periods would all be from 13:20 on a given day to 13:20 the following day.
(NOTE: be sure that if your parameter is a unix timestamp integer, that this is being interpreted correctly by the database.)
It might be more efficient to pull the ten rows in a single query. If there is a guarantee that 'timestamp' is unique, then it's possible to craft such a query, but the query text will be considerably more complex than what you have now. We could mess with getting MAX(timestamp_) within each period, and then joining that back to get the row... but that's going to be really messy.
If I were going to try to pull all ten rows I would probably try going with a UNION ALL approach, not very pretty, but it least it could be tuned.
SELECT p0.*
FROM ( SELECT * FROM `eventos` WHERE ...
AND `created` < DATE_ADD(FROM_UNIXTIME({$timestamp}),INTERVAL 0*24 HOUR)
AND `created` >= DATE_ADD(FROM_UNIXTIME({$timestamp}),INTERVAL -1*24 HOUR)
ORDER BY `created` DESC LIMIT 1
) p0
UNION ALL
SELECT p1.*
FROM ( SELECT * FROM `eventos` WHERE ...
AND `created` < DATE_ADD(FROM_UNIXTIME({$timestamp}),INTERVAL -1*24 HOUR)
AND `created` >= DATE_ADD(FROM_UNIXTIME({$timestamp}),INTERVAL -2*24 HOUR)
ORDER BY `created` DESC LIMIT 1
) p1
UNION ALL
SELECT p2.*
FROM ( SELECT * FROM `eventos` WHERE ...
AND `created` < DATE_ADD(FROM_UNIXTIME({$timestamp}),INTERVAL -2*24 HOUR)
AND `created` >= DATE_ADD(FROM_UNIXTIME({$timestamp}),INTERVAL -3*24 HOUR)
ORDER BY `created` DESC LIMIT 1
) p2
UNION ALL
SELECT p3.*
FROM ...
Again, this could be generalized, to pass in a number of seconds as an argument. Replace HOUR with SECOND, and replace the '24' with a bind parameter that has a number of seconds.
It's rather long winded, but it should run okay.
Another really messy and complicated way to get this back in a single result set would be to use an inline view to get the end timestamp for the ten periods, something like this:
SELECT p.period_end
FROM (SELECT DATE_ADD(t.t_,INTERVAL -1 * i.i_* {$nsecs} SECOND) AS period_end
FROM (SELECT FROM_UNIXTIME( {$timestamp} ) AS t_) t
JOIN (SELECT 0 AS i_
UNION ALL SELECT 1
UNION ALL SELECT 2
UNION ALL SELECT 3
UNION ALL SELECT 4
UNION ALL SELECT 5
UNION ALL SELECT 6
UNION ALL SELECT 7
UNION ALL SELECT 8
UNION ALL SELECT 9
) i
) p
And then join that to your table ...
ON `created` < p.period_end
AND `created` >= DATE_ADD(p.period_end,INTERVAL -1 * {$nsecs} SECOND)
And pull back MAX(created) for each period GROUP BY p.period_end, wrap that in an inline view.
And then join that back to your table to get each row.
But that is really, really messy, hard to understand, and not likely to be any faster (or more efficient) than what you are already doing. The most improvement you could make is the time it takes to run 9 of your queries.
Assuming you want the latest (having the greatest created date) event per day for the last 10 days.
so let's get the latest timestamp per day
$today = date('Y-m-d');
$tenDaysAgo = date('Y-m-d', strtotime('-10 day'));
$innerSql = "SELECT date_format(created, '%Y-%m-%d') day, MAX(created) max_created FROM eventos WHERE date_format(created, '%Y-%m-%d') BETWEEN '$today' and '$tenDaysAgo' GROUP BY date_format(created, '%Y-%m-%d')";
Then we can select all the events that match those created dates
$outerSql = "SELECT * FROM eventos INNER JOIN ($innerSql) as A WHERE eventos.created = A.max_created";
I haven't had a chance to test this, but the principles should be sound enough.
If you want to group by some other arbitrary number of hours you would change innerSql:
$fromDate = '2012-07-06' // or if you want a specific time '2012-07-06 12:00:00'
$intervalInHours = 5;
$numberOfIntervals = 10;
$innerSql = "SELECT FLOOR(TIMESTAMPDIFF(HOUR, created, '$fromDate') / $intervalInHours) as grouping, MAX(created) as max_created FROM eventos WHERE created BETWEEN DATE_SUB('$fromDate', INTERVAL ($intervalInHours * $numberOfIntervals) HOUR) AND '$fromDate' GROUP BY FLOOR(TIMESTAMPDIFF(HOUR, created, '$fromDate') / $intervalInHours)";
I'd add another column that is the date(not time) and then use MySQL "group by" to get the most recent for each date.
http://www.tizag.com/mysqlTutorial/mysqlgroupby.php/
This tutorial does just that, but by product type instead of date. This should help!
Do you want all of the events within the 10 days, or just one event per day within the 10 day period?
Either way, consider MySQL's date functions for assistance. It should help you get the date range you want.
Here's one that will get you the first event of the day for the last 10 days.
SELECT *
FROM eventos
WHERE created BETWEEN DATE_SUB(DATE(NOW()), INTERVAL 10 DAY) AND DATE_ADD(DATE(NOW()), INTERVAL 1 DAY)
GROUP BY DATE(created)
ORDER BY MAX(created) DESC
LIMIT 10
Try this:
SELECT *
FROM eventos
WHERE created BETWEEN DATE_SUB(DATE(NOW()), INTERVAL 10 DAY) AND DATE_ADD(DATE(NOW()), INTERVAL 1 DAY)
ORDER BY created DESC
LIMIT 10
i have the following statement:
SELECT
count(rs.rsc_id) as counter
FROM shots as rs
where rsc_rs_id = 345354
AND YEAR(rs.timestamp) = YEAR(DATE_SUB(CURDATE(), INTERVAL 6 MONTH))
GROUP BY DATE_FORMAT(rs.timestamp,'%Y%m')
rs.timestamp is a unix timestamp
Output would be like for each row / month a numeric like '28'
It Works fine, but if i have inconsistent data, like only for the past three month (not for all six month), i get no return from my Database. I would like to have every time there is not data for this month, 0 returned...
any suggestion?
i thought about some case statements, but this seems not so good...
thanks!!
For only 6 months, a date table seems unnecessary, although this looks complicated (it really isn't!)
SELECT DATE_FORMAT(N.PivotDate,'%Y%m'), count(rs.rsc_id) as counter
FROM (
select ADDDATE(CURDATE(), INTERVAL N MONTH) PivotDate
FROM (
select 0 N union all
select 1 union all
select 2 union all
select 3 union all
select 4 union all
select 5 union all
select 6) N) N
LEFT JOIN shots as rs
ON rsc_rs_id = 345354
AND DATE_FORMAT(N.PivotDate,'%Y%m')=DATE_FORMAT(FROM_UNIXTIME(rs.timestamp),'%Y%m')
GROUP BY DATE_FORMAT(N.PivotDate,'%Y%m')
In such cases it's common to use a table of dates with all dates (e.g. from 1/1/1970 to 31/12/2999) and LEFT JOIN your data to that table.
See an example in the answer here: mysql joins tables creating missing dates
If you create a dates table you can use:
SELECT
DATE_FORMAT(d.date,'%Y%m') AS `month`, count(rs.rsc_id) AS `counter`
FROM dates d
LEFT JOIN shots as rs
ON d.date = FROM_UNIXTIME(rs.timestamp)
AND rs.rsc_rs_id = 345354
WHERE d.date > DATE_SUB(CURDATE(), INTERVAL 5 MONTH)
AND d.date < CURDATE()
GROUP BY DATE_FORMAT(d.date,'%Y%m');
I'm looking for a best practice advice how to speed up queries and at the same time to minimize the overhead needed to invoke date/mktime functions. To trivialize the problem I'm dealing with the following table layout:
CREATE TABLE my_table(
id INTEGER PRIMARY KEY NOT NULL AUTO_INCREMENT,
important_data INTEGER,
date INTEGER);
The user can choose to show 1) all entries between two dates:
SELECT * FROM my_table
WHERE date >= ? AND date <= ?
ORDER BY date DESC;
Output:
10-21-2009 12:12:12, 10002
10-21-2009 14:12:12, 15002
10-22-2009 14:05:01, 20030
10-23-2009 15:23:35, 300
....
I don't think there is much to improve in this case.
2) Summarize/group the output by day, week, month, year:
SELECT COUNT(*) AS count, SUM(important_data) AS important_data
FROM my_table
WHERE date >= ? AND date <= ?
ORDER BY date DESC;
Example output by month:
10-2009, 100002
11-2009, 200030
12-2009, 3000
01-2010, 0 /* <- very important to show empty dates, with no entries in the table! */
....
To accomplish option 2) I'm currently running a very costly for-loop with mktime/date like the following:
for(...){ /* example for group by day */
$span_from = (int)mktime(0, 0, 0, date("m", $time_min), date("d", $time_min)+$i, date("Y", $time_min));
$span_to = (int)mktime(0, 0, 0, date("m", $time_min), date("d", $time_min)+$i+1, date("Y", $time_min));
$query = "..";
$output = date("m-d-y", ..);
}
What are my ideas so far? Add additional/ redundant columns (INTEGER) for day (20091212), month (200912), week (200942) and year (2009). This way I can get rid of all the unnecessary queries in the for loop. However I'm still facing the problem to very fastly calculate all dates that doesn't have any equivalent in database. One way to simply move the problem could be to let MySQL do the job and simply use one big query (calculate all the dates/use MySQL date functions) with a left join (the data). Would it be wise to let MySQL take the extra load? Anyway I'm reluctant to use all these mktime/date in the for loop. Since I have complete control over the table layout and code even suggestions with major changes are welcome!
Update
Thanks to Greg I came up with the following SQL query. However it still bugs me to use 50 lines of sql statements - build up with php - that maybe could be done faster and more elegantly otherwise:
SELECT * FROM (
SELECT DATE_ADD('2009-01-30', INTERVAL 0 DAY) AS day UNION ALL
SELECT DATE_ADD('2009-01-30', INTERVAL 1 DAY) AS day UNION ALL
SELECT DATE_ADD('2009-01-30', INTERVAL 2 DAY) AS day UNION ALL
SELECT DATE_ADD('2009-01-30', INTERVAL 3 DAY) AS day UNION ALL
......
SELECT DATE_ADD('2009-01-30', INTERVAL 50 DAY) AS day ) AS dates
LEFT JOIN (
SELECT DATE_FORMAT(date, '%Y-%m-%d') AS date, SUM(data) AS data
FROM test
GROUP BY date
) AS results
ON DATE_FORMAT(dates.day, '%Y-%m-%d') = results.date;
You definitely shouldn't be doing a query inside a loop.
You can group like this:
SELECT COUNT(*) AS count, SUM(important_data) AS important_data, DATE_FORMAT('%Y-%m', date) AS month
FROM my_table
WHERE date BETWEEN ? AND ? -- This should be the min and max of the whole range
GROUP BY DATE_FORMAT('%Y-%m', date)
ORDER BY date DESC;
Then pull these into an array keyed by date and loop over your data range as you are doing (that loop should be pretty light on CPU).
Another idea is not to use string inside the query. Transform the string parameter to datetime, on mysql.
STR_TO_DATE(str,format)
http://dev.mysql.com/doc/refman/5.0/en/date-and-time-functions.html