Combine two SQL queries which use Date Intervals, Group and SUM - mysql

First off, thank you in advance.
I have 2 queries that I am working with as I would like to compare last week to the previous week...
Getting Last Week
SELECT keyword,
SUM(users_desktop) AS desktop_last_week
FROM js
WHERE country='US'
AND day >= curdate() - INTERVAL DAYOFWEEK(curdate())+6 DAY
AND day < curdate() - INTERVAL DAYOFWEEK(curdate())-1 DAY
GROUP BY keyword
ORDER BY desktop_last_week DESC
LIMIT 10;
Getting Previous Week
SELECT keyword,
SUM(users_desktop) AS desktop_previous_week
FROM js
WHERE country='US'
AND day >= curdate() - INTERVAL DAYOFWEEK(curdate())+12 DAY
AND day < curdate() - INTERVAL DAYOFWEEK(curdate())-7 DAY
GROUP BY keyword
ORDER BY desktop_previous_week DESC
LIMIT 10;
What I would like to do is combine these queries so that I can then ORDER BY the division of desktop_last_week/desktop_this_week to find keywords that are trending up (ie: searched a lot more this week than last week)
Any idea on how to combine these together?

This is called conditional aggregation. Use where clauses from your existing queries to do it.
SELECT keyword,
Sum(CASE
WHEN day >= Curdate() - INTERVAL Dayofweek(Curdate())+6 day
AND day < Curdate() - INTERVAL Dayofweek(Curdate())-1 day THEN
users_desktop
ELSE 0
end) AS desktop_last_week,
Sum(CASE
WHEN day >= Curdate() - INTERVAL Dayofweek(Curdate())+12 day
AND day < Curdate() - INTERVAL Dayofweek(Curdate())-7 day THEN
users_desktop
ELSE 0
end) AS desktop_prev_last_week
FROM js
WHERE country = 'US'
GROUP BY keyword
ORDER BY desktop_prev_last_week / desktop_this_week DESC
LIMIT 10

Use Union!
You can get 2 Queries to one with that. If you want to order that then put it into a subquery

You could expand the where clause to contain both time ranges and move the logic of differentiating between them to case expressions inside the sums. By the way, using the between operator instead of a >= and < pair would make the query itself a tad easier to read:
SELECT keyword,
SUM(CASE WHEN day BETWEEN
(CURDATE() - INTERVAL DAYOFWEEK(curdate()) + 6 DAY) AND
(CURDATE() - INTERVAL DAYOFWEEK(curdate()) - 1 DAY)
THEN users_desktop
END) AS desktop_last_week,
SUM(CASE WHEN day BETWEEN
(CURDATE() - INTERVAL DAYOFWEEK(curdate()) + 12 DAY) AND
(CURDATE() - INTERVAL DAYOFWEEK(curdate()) - 7 DAY)
THEN users_desktop
END) AS desktop_previous_week
FROM js
WHERE country = 'US'
GROUP BY keyword
ORDER BY desktop_last_week DESC, desktop_previous_week DESC
LIMIT 10;

Related

How DO I fetch last Week and current week

How do I fetch last week data from monday time (00:00:01) and end on sunday time (23:59:59)...
same as this current week from monday time (00:00:01) and end on sunday time (23:59:59)
WHat I tried!
$query = "SELECT users.name,count(*) as count,
campaign.campaign_name,
campaign.payout_cost*count(*) as totalPrice
FROM users
JOIN transactions on users.uid=transactions.uid
JOIN campaign on campaign.campaign_id_id=transactions.campaign_id
WHERE uid=$uid
AND `date` >= DATE_SUB(DATE(NOW()), INTERVAL DAYOFWEEK(NOW())+6 DAY)
AND `date` < DATE_SUB(DATE(NOW()), INTERVAL DAYOFWEEK(NOW())-1 DAY)
GROUP BY campaign.campaign_name_name ";
You are on the right track by avoiding functions like week() on the column -- that just messes up the optimizer. On the other hand, the uid parameter should be passed as a parameter rather than munging the query string.
You want to use the weekday() function because you want weeks to start on a Monday. Just some arcaneness of MySQL: weekday() returns 0 for Monday whereas dayofweek() returns 2 for Monday.
So, the logic for the current week would be:
date >= curdate() - interval weekday(curdate()) day and
date < curdate() + interval 7 - weekday(curdate()) day
For last week, this would be:
date >= curdate() - interval 7 + weekday(curdate()) day and
date < curdate() + interval - weekday(curdate()) day
Notes that curdate() (or current_date) returns the current date with no time component, so no date() is required.
Couple of ways to do it...
select data from tableName
where date between date_sub(now(),INTERVAL 1 WEEK) and now();
select data FROM tableName
wherdate >= curdate() - INTERVAL DAYOFWEEK(curdate())+6 DAY
AND date < curdate() - INTERVAL DAYOFWEEK(curdate())-1 DAY
You can use WEEK() function, which returns the week number for a given date, by adding
AND WEEK(date-INTERVAL 1 DAY) = WEEK(NOW()) - 1 to get current week's data starting from monday upto sunday,
and
AND WEEK(date-INTERVAL 1 DAY) = WEEK(NOW()) - 2 for the previous week's data
into the WHERE condition after WHERE uid=$uid
such as
$query = "SELECT c.campaign_name,
COUNT(*) as total_count,
SUM(c.payout_cost) as total_payout
FROM transactions t
JOIN campaign c
ON c.campaign_id = t.campaign_id
WHERE uid = $uid
AND WEEK(date - INTERVAL 1 DAY) = WEEK(NOW()) - 1
GROUP BY c.campaign_name ";
and replace WEEK(NOW()) - 1 with WEEK(NOW()) - 2, also
Demo

Counting all rows in column with two different date conditions

I'm trying to turn two count queries with date conditions (the ones below) into one query.
SELECT COUNT(*) as yesterday FROM orders WHERE DATE(timedate) = DATE(NOW() - INTERVAL 1 DAY)
SELECT COUNT(*) as yesterday FROM orders WHERE DATE(timedate) = DATE(NOW() - INTERVAL 2 DAY)
Following the advice of another answer I created the following, but that doesn't seem to work syntax-wise, and I'm not quite sure why. Is there another way to do this? I can't find a similar question on this
SELECT
SUM(IF(DATE(timedate) = DATE(NOW() - INTERVAL 1 DAY))) AS testcount1,
SUM(IF(DATE(timedate) = DATE(NOW() - INTERVAL 2 DAY))) AS testcount2
FROM
orders
You're missing the output values for the IF expression. Also you should use CURRENT_DATE() so you don't need to convert to a DATE:
SELECT
SUM(IF(DATE(timedate) = CURRENT_DATE() - INTERVAL 1 DAY, 1, 0)) AS testcount1,
SUM(IF(DATE(timedate) = CURRENT_DATE() - INTERVAL 2 DAY, 1, 0)) AS testcount2
FROM
orders
Note that MySQL treats boolean expressions as 1 (true) or 0 (false) in a numeric context, so you can actually SUM the expression without needing the IF:
SELECT
SUM(DATE(timedate) = CURRENT_DATE() - INTERVAL 1 DAY) AS testcount1,
SUM(DATE(timedate) = CURRENT_DATE() - INTERVAL 2 DAY) AS testcount2
FROM
orders
You want conditional aggregation. I would phrase the query as follows:
SELECT
SUM(
timedate >= CURRENT_DATE - INTERVAL 1 DAY
and timedate < CURRENT_DATE
) AS testcount1,
SUM(
timedate >= CURRENT_DATE - INTERVAL 2 DAY
and timedate < CURRENT_DATE- INTERVAL 1 DAT
) AS testcount2
FROM orders
Details:
this uses a nice feature of MySQL, that evaluates false/true conditions as 0/1 in numeric context
no date functions are applied on the timedate column : instead, we do litteral date comparisons. This is much more efficient, since the database can possibly take advantage of an index on the datetime column
You might also want to add a WHERE clause to the query:
WHERE
timedate >= CURRENT_DATE - INTERVAL 2 day
AND timedate< CURRENT_DATE

Get number of entries per multiple date intervals using single query

SELECT COUNT(*) FROM `table` WHERE `datetime` > SUBDATE(NOW(), INTERVAL 1 DAY)
This will get number of entries during last day. But is it possible to get number of entries for multiple intervals without having to send variation of this query multiple times (INTERVAL 1 DAY, INTERVAL 1 WEEK, INTERVAL 1 MONTH, ...)?
You need CASE WHEN expression to accomplish that.
SELECT
COUNT(CASE WHEN DATE(`datetime`) >= CURDATE() - INTERVAL 1 DAY AND DATE(`datetime`) < CURDATE() THEN 1 END) AS lastDay,
COUNT(CASE WHEN DATE(`datetime`) >= CURDATE() - INTERVAL 7 DAY AND DATE(`datetime`) < CURDATE() THEN 1 END ) AS lastSevenDays,
COUNT(*) AS lastThirtyDays
FROM `table`
WHERE
DATE(`datetime`) >= CURDATE() - INTERVAL 30 DAY
How to use CASE WHEN expression
Note: If your requirement is to get result of last day, last 7 days and last 30 days then go with this query.
EDIT:
If you have an index on datetime field then the above query will fail to use that index. Please use the query given below in order to utilize the index on datetime.
SELECT
COUNT(CASE WHEN DATE(`datetime`) >= CURDATE() - INTERVAL 1 DAY AND DATE(`datetime`) < CURDATE() THEN 1 END) AS lastDay,
COUNT(CASE WHEN DATE(`datetime`) >= CURDATE() - INTERVAL 7 DAY AND DATE(`datetime`) < CURDATE() THEN 1 END ) AS lastSevenDays,
COUNT(*) AS lastThirtyDays
FROM `table`
WHERE
`datetime` >= (NOW() - INTERVAL 30 DAY - INTERVAL HOUR(NOW()) HOUR - INTERVAL MINUTE(NOW()) MINUTE - INTERVAL SECOND(NOW()) SECOND)

How do I retrieve data for the previous month in SQL

I want to get data for the dates between 2015-05-01 and 2015-06-01 using SQL.
Please help me with the query.
The query I used is:
select *,count(id) as multiple_visitors
from table1
where id=123
and (date(server_time) between (CURDATE() - INTERVAL 31 DAY) AND CURDATE())
group by user_id having count(id)>1
You can do this with month() and year():
where month(server_time) = month(curdate() - interval 1 month) and
year(server_time) = year(curdate() - interval 1 month)
However, I recommend a slightly more complex expression:
where server_time >= date_sub(date_sub(curdate(), interval - day(curdate()) + 1 day), interval 1 month) and
server_time < date_sub(curdate(), interval - day(curdate()) + 1 day)
The advantage is that there are no functions on server_time, so the database engine can use an index, if appropriate.
As a note: the expression date_sub(curdate(), interval - day(curdate()) + 1 day) gets midnight on the first day of the month.
Try using "WHERE" with MONTH(date).
Like this:
SELECT * FROM Table
WHERE MONTH(date) = 1

comparing dates by month and year in mysql

I have a table containing data about events and festivals with following columns recording their start and end dates.
Start_Date
End_Date
date format is in YYYY-MM-DD. I need to fetch event details with the following condition.
Need to fetch all events which start with a current month and there end dates can be anything say currentDate+next30days.
I am clear about end date concept. but not sure how I can fetch data whose start dates are in a current month.
For this, I need to compare current year and current month against the Start_Date column in my database.
Can anyone help me to point out as how I can do that?
select * from your_table
where year(Start_Date) = year(curdate())
and month(Start_Date) = month(curdate())
and end_date <= curdate() + interval 30 day
I don't like either of the other two answers, because they do not let the optimizer use an index on start_date. For that, the functions need to be on the current date side.
So, I would go for:
where start_date >= date_add(curdate(), interval 1 - day(curdate()) day) and
start_date < date_add(date_add(curdate(), interval 1 - day(curdate()) day), interval 1 month)
All the date functions are on curdate(), which does not affect the ability of MySQL to use an index in this case.
You can also include the condition on end_date:
where (start_date >= date_add(curdate(), interval 1 - day(curdate()) day) and
start_date < date_add(date_add(curdate(), interval 1 - day(curdate()) day), interval 1 month)
) and
end_date <= date_add(curdate(), interval 30 day)
This can still take advantage of an index.
DateTime functions are your friends:
SELECT
*
FROM
`event`
WHERE
(MONTH(NOW()) = MONTH(`Start_Date`))
AND
(`End_Date` <= (NOW() + INTERVAL 30 DAY))
AND
(YEAR(NOW()) = YEAR(`Start_Date`))
Comparing the year and month separately feels messy. I like to contain it in one line. I doubt it will make a noticeable difference in performance, so its purely personal preference.
select * from your_table
where LAST_DAY(Start_Date) = LAST_DAY(curdate())
and end_date <= curdate() + interval 30 day
So all I'm doing is using the last_day function to check the last day of the month of each date and then comparing this common denominator. You could also use
where DATE_FORMAT(Start_Date ,'%Y-%m-01') = DATE_FORMAT(curdate(),'%Y-%m-01')