I have data with start date and end date (Say 20th Feb 2018 to 20th Feb 2020), I want to find out the total days in every year inside this range.
For example:
2018 - x days
, 2019 - 365 days
, 2020 - y days etc.
Is there a way I can do in SQL without hardcoding year values?
I tried hardcoding the values and it worked well. But I want a solution without hardcoding year values
I'm not familiar enough with MySql to know if this will port, however here is a tested and confirmed SQL Server solution.
The fiddle link is here for your use.
Given start dates 02/20/2018 and 02/20/2020, the result set is as follows:
Year
periodStart
periodEnd
DaysInPeriod
2018
2018-02-20
2018-12-31
314
2019
2019-01-01
2019-12-31
365
2020
2020-01-01
2020-02-20
51
Declare #StartDate date = '2018-02-20', #EndDate date = '2020-02-20';
WITH x AS (SELECT n FROM (VALUES (0),(1),(2),(3),(4),(5),(6),(7),(8),(9)) v(n)),
Years AS (
SELECT ROW_NUMBER() OVER (ORDER BY (SELECT NULL)) AS Year
FROM x ones, x tens, x hundreds, x thousands)
SELECT Years.Year,
CASE
WHEN Year(#StartDate) = Years.year THEN #StartDate
ELSE DATEFROMPARTS(years.year, 01, 01)
END AS periodStart,
CASE
WHEN Year(#EndDate) = Years.year THEN #EndDate
ELSE DATEFROMPARTS(years.year, 12, 31)
END AS periodEnd,
DATEDIFF(day,
CASE
WHEN Year(#StartDate) = Years.year THEN #StartDate
ELSE DATEFROMPARTS(years.year, 01, 01)
END,
CASE
WHEN Year(#EndDate) = Years.year THEN #EndDate
ELSE DATEFROMPARTS(years.year, 12, 31)
END
) + 1 AS DaysInPeriod
FROM Years
WHERE Years.Year >= Year(#StartDate)
AND Years.Year <= Year(#EndDate)
Using WITH RECURSIVE to create range of dates then we can easly count the number of days for each year using DATEDIFF
WITH RECURSIVE dates AS
(
SELECT min(start_date) as start_date, DATE_FORMAT(min(start_date),'%Y-12-31') as last_day FROM mytable
UNION ALL
SELECT DATE_FORMAT(start_date + INTERVAL 1 YEAR,'%Y-01-01'),
DATE_FORMAT(start_date + INTERVAL 1 YEAR,'%Y-12-31')
FROM dates
WHERE DATE_FORMAT(start_date + INTERVAL 1 YEAR,'%Y-01-01') <= (SELECT MAX(end_date) FROM mytable)
),
cte2 as (
SELECT d.start_date as start_day, if(YEAR(d.start_date) = YEAR(m.end_date), m.end_date, d.last_day) as last_day
FROM dates d, mytable m
)
select *, DATEDIFF(last_day, start_day)+1 as total_days
from cte2;
Demo here
You are looking for the DATEDIFF function.
https://dev.mysql.com/doc/refman/8.0/en/date-and-time-functions.html#function_datediff
DATEDIFF() returns expr1 − expr2 expressed as a value in days from one date to the other. expr1 and expr2 are date or date-and-time expressions.
You are free to specify e.g. "2019-01-01" or "2020-01-01"
as input arguments to DATEDIFF.
You may find it convenient to store several January 1st
dates in a calendar reporting table, if you want SELECT to loop
over several years and report on number of days in each year.
Related
I have a table called months - this contains all 12 months of the calendar, the IDs correspond to the month number.
I will be running a query to retrieve 2 or 3 sequential months from this table, e.g
April & May
June, July, August
December & January
However I want to ensure that whenever December are January and retrieved, that it retrieves them in that order, and not January - December. Here is what I have tried:
SELECT * FROM `months`
WHERE start_date BETWEEN <date1> AND <date2>
ORDER BY
FIELD(id, 12, 1)
This works for December & January, but now when I try to retrieve January & February it does those in the wrong order, i.e "February - January" - I'm guessing because we specified 1 in the ORDER BY as the last value.
Anybody know the correct way to achieve this? As I mentioned this should also work for 3 months, so for example "November, December, January" and "December, January, February" should all be retrieved in that order.
If you want December first, but the other months in order, then:
order by (id = 12) desc, id
MySQL treats booleans as numbers, with "1" for true and "0" for false. The desc puts the 12s first.
EDIT:
To handle the more general case, you can use window functions. Assuming the numbers are consecutive, then the issue is trickier. This will work for 2 and 3-month spans:
order by (case min(id) over () > 1 then id end),
(case when id > 6 1 else 2 end),
id
I'm reluctant to think about a more general solution based only on months. After all, you can just use:
order by start_date
Or, if you have an aggregation query:
order by min(start_date)
to solve the real problem.
This is not "mysql solution" properly :
with cte (id, month) AS (
select id, month from months
union all
select id, month from months
)
, cte1 (id, month, r) as (select id, month, row_number() over() as r from cte )
select * from cte1
where id in (12, 1)
and r >= 12 order by r limit 2 ;
DECLARE
#monthfrom int = 12,
#monthto int = 1;
with months as (select 1 m
union all
select m+1 from months where m<12)
select m
from months
where m in (#monthfrom,#monthto)
order by
case when #monthfrom>#monthto
then
m%12
else
m
end
result:
12
1
Basically in MySQL this can be done the same way:
set #from =12;
set #to =1;
with recursive months(m) as (
select 1 m
union all
select m+1 from months where m<12)
select *
from months
where m in (#from,#to)
order by case when #from>#to then m%12 else m end;
I have a list of periods during a year, and they are the same every year. You can think of it as a Season. They have a startDate and a endDate.
Because there can be Seasons that leap each other, what I need to to is query all the matching Seasons given a date, no matter what year.
As an example:
Season1: from 1st of January to 10th of January
Season2: from 6th of January to 8th of January
Season3: from 11th of January to 20th of January
Given the date 7th of January, I'd need to retrieve the Season1 and Season2.
I've tried converting all dates to the same year, but It doesn't work when the Start Date of a season in "later" than the End Date (for example, there's a period starting on November and ending of February).
Thanks in advance for the help.
Edit, sample data:
StartDate EndDate SeasonId
2000-08-01 2000-08-31 4
2000-12-29 2000-01-02 3
2000-06-01 2000-07-30 3
2000-09-01 2000-09-30 3
2000-01-06 2000-01-08 3
2000-04-07 2000-04-17 3
2000-04-28 2000-05-01 3
2000-06-02 2000-06-05 3
2000-06-23 2000-06-25 3
2000-09-08 2000-09-11 3
2000-09-22 2000-09-25 3
2000-10-12 2000-10-15 3
2000-11-01 2000-11-05 3
2000-12-01 2000-12-10 3
2000-12-22 2000-12-26 3
2000-03-01 2000-05-31 2
2000-10-01 2000-10-31 2
2000-11-01 2000-02-28 1
And I'd need, for example, the season for the date 2000-02-08, and retrieve seasonId = 1, or the date 2000-10-13and retrive seasonId = 3, seasonId = 2
I would do it in 2 'options': (the following SQL assumes you already got rid of the year in the table, and left only month-date format. )
select ... from seasons s where
(s.startDate <= s.endDate and s.startDate <= #mydate and s.endDate >= #mydate) or
(s.startDate > s.endDate and s.startDate >= #mydate and s.endDate <= #mydate)
You could query like this for the Season1:
select * from myTable where (month(myDate) = 1 and DAY(myDate) between 1 and 10)
If you have a season in more than one month, like start date January 20th, and finish date Febrery 10th, you could query this way:
select * from myTable where (month(myDate) = 1 and DAY(myDate) >= 20) or (month(myDate) = 2 and DAY(myDate) <= 10)
UPDATED WITH YOUR UPDATE
It is a little bit tricky, but it should work...
select * from seasons_table
where cast(cast(day(myDate) as char) + '/' + cast(month(myDate) as char) + '/' + '2000' as date) between
cast(cast(day(StartDate) as char) + '/' + cast(month(StartDate) as char) + '/' + '2000' as date) and
cast(cast(day(EndDate) as char) + '/' + cast(month(EndDate) as char) + '/' + '2000' as date)
given tblSeason with columns Id, startdate, enddate and your date as #myDate you would query as
Select Id From tblSeason WHERE #myDate BETWEEN startdate AND enddate
would give list of Id's of the seasons that match.
if you can't work from that, please give more information in your examples as to the structure you are querying and the expected outcome.
*Edit to ignore the year part you could do similar to
Declare #myDate datetime = '2016-10-13'
SELECT [StartDate]
,[EndDate]
,[SeasonId]
FROM [dbo].[Table_1]
where DATEPART(dy, #myDate) >= DATEPART(dy,StartDate)
AND (DATEPART(dy,#myDate) =< DATEPART(dy,EndDate) OR DATEPART(dy,StartDate) > DATEPART(dy,EndDate))
Why are you including the year in the table? That seems strange.
In any case, you only care about the MM-DD format, so use date_format() to convert the values to strings:
select t.*
from t
where (start_date <= end_date and
date_format(#date, '%m-%d') >= date_format(start_date, '%m-%d') and
date_format(#date, '%m-%d') <= date_format(end_date, '%m-%d')
) or
(start_date > end_date and
date_format(#date, '%m-%d') <= date_format(start_date, '%m-%d') and
date_format(#date, '%m-%d') >= date_format(end_date, '%m-%d')
);
The strings are fine for comparison, because you are only looking at the month and day components of the date.
Given the nature of your problem, I would recommend that you store start_date and end_date in a non-date format, such as MM-DD.
I want to show date range of week by passing a variable which contains year and month i.e 2016-07 like this in mysql how could i achieve this:
for eg: if i pass 2016-07 then my output should look like this:
2016-07-03 to 2016-07-10 || 2016-07-11 to 2016-07-17
and so on
There are several parts to solving this problem, when you go to solve it in MySQL.
One is to decide whether your weeks start on Sundays or Mondays. That's a locale-specific business rule. In locales formerly part of the British Empire (USA, India) it's usually Sunday. In other places, it's Monday.
So we'll need a function like this: firstDayOfWeek(date). More about that in a moment.
Once we have that, we'll need a way to get the last day of the month in question. That's easy; it's a built-in MySQL function. LAST_DAY(date)
You have said you'll specify the month in question with a string like 1941-12. We'll need to turn that into a DATE object. That can be done like this:
STR_TO_DATE(CONCAT(`1941-12`, `-01'), '%Y-%m-%d')
We need a virtual table with the integers from 0 to 4. Let's call that table seq_0_to_4. We pick that range of integers because no months have more than five Sundays (or Mondays). More on that later.
OK, these are the conceptual building blocks. Let's use them.
The first day of the week derived from the last day of the month is
SET #month := '1941-12';
SET #first_day_of_month := STR_TO_DATE(CONCAT(`1941-12`, `-01'), '%Y-%m-%d')
SET #seed : =FirstDayOfWeek(LAST_DAY(first_day_of_month));
Then you need four five consecutive week-starting days the last of which is #seed.
SELECT #seed - INTERVAL (7*seq.seq) first_day_of_week
FROM seq_0_to_4
Next you need to limit that to days in your month.
SELECT first_day_of_week,
first_day_of_week + INTERVAL 6 DAY last_day_of_week
FROM (
SELECT #seed - INTERVAL (7*seq.seq) first_day_of_week
FROM seq_0_to_4
) w
WHERE first_day_of_week >= #first_day_of_month
ORDER BY first_day_of_week
That gives you a table of rows, one for each week beginning in the month. If you want to exclude weeks in which the last weekday is in the next month, change your WHERE to
WHERE first_day_of_week >= #first_day_of_month
AND first_day_of_week + INTERVAL 6 DAY <= #seed
Finally, to get the exact string format you specified in your question, wrap that query in this:
SELECT GROUP_CONCAT (
CONCAT(first_day_of_week, ' to ', last_day_of_week)
SEPARATOR ' || '
ORDER BY first_day_of_week)
FROM (
SELECT first_day_of_week,
first_day_of_week + INTERVAL 6 DAY last_day_of_week
FROM (
SELECT #seed - INTERVAL (7*seq.seq) first_day_of_week
FROM seq_0_to_4
) w
WHERE first_day_of_week >= #first_day_of_month
) x
That's it.
I promised to describe FirstDayOfWeek(dt). Here it is.
FROM_DAYS(TO_DAYS(dt) -MOD(TO_DAYS(dt) -1, 7))
It's a bit of a magic spell, but it works. If your weeks start Mondays, it is this.
FROM_DAYS(TO_DAYS(dt) -MOD(TO_DAYS(dt) -2, 7))
I promised to describe seq_0_to_4. If you're using the MariaDB fork of MySQL, it's built in. If you're using the Sun / Oracle fork, you define it like this.
(SELECT 0 AS seq UNION SELECT 1 UNION SELECT 2 UNION SELECT 3 UNION SELECT 4) seq_0_to_4
Putting it all together:
SET #month := '1941-12';
SET #first_day_of_month := STR_TO_DATE(CONCAT(`1941-12`, `-01'), '%Y-%m-%d');
SET #seed := FROM_DAYS(TO_DAYS(LAST_DAY(first_day_of_month))
-MOD(TO_DAYS(LAST_DAY(first_day_of_month)) -1, 7));
SELECT GROUP_CONCAT (
CONCAT(first_day_of_week, ' to ', last_day_of_week)
SEPARATOR ' || '
ORDER BY first_day_of_week)
FROM (
SELECT first_day_of_week,
first_day_of_week + INTERVAL 6 DAY last_day_of_week
FROM (
SELECT #seed - INTERVAL (7*seq.seq) first_day_of_week
FROM (SELECT 0 AS seq UNION SELECT 1 UNION SELECT 2
UNION SELECT 3 UNION SELECT 4
) seq
) w
WHERE first_day_of_week >= #first_day_of_month
) x
It's unreasonably complex (the technical term is a freakin' hairball) to solve your problem in pure MySQL-dialect SQL, but it's possible.
Probably is too late but i figured out this using Common Table expressions in MYSQL 8.0.2
WITH RECURSIVE
Years(y) AS
(
SELECT 2020
UNION ALL
SELECT y + 1 FROM Years WHERE y < 2021
),
Days (d) AS
(
SELECT 1
UNION ALL
SELECT d + 1 FROM Days WHERE d < 366
)
SELECT
y AS Year,
MONTH(MakeDate(y,d)) AS Month,
WEEK(MakeDate(y,d))+1 -WEEK(TIMESTAMPADD(MONTH,MONTH(MakeDate(y,d))-1,MakeDate(y,1))) AS Week,
Min(MakeDate(y,d)) AS StartDate,
timestampadd(second,-1,timestampadd(day,1,MAx(MakeDate(y,d)))) AS EndDate
FROM Years,Days
WHERE Year(MakeDate(y,d)) <= y
GROUP BY y, MONTH(MakeDate(y,d)),WEEK(MakeDate(y,d))+1 -WEEK(TIMESTAMPADD(MONTH,MONTH(MakeDate(y,d))-1,MakeDate(y,1)))
ORDER BY 1,2,3
I need to query data with count and sum by multiple date ranges and I am looking for a faster query than what I am doing now.
I have a transaction table with a date and amount. I need to present a table with a count of transactions and total amount by date ranges of today, yesterday, this week, last week, this month, last month. Currently I am doing sub queries, is there a better way?
select
(select count(date) from transactions where date between ({{today}})) as count_today,
(select sum(amount) from transactions where date between ({{today}})) as amount_today,
(select count(date) from transactions where date between ({{yesterday}})) as count_yesterday,
(select sum(amount) from transactions where date between ({{yesterday}})) as amount_yesterday,
(select count(date) from transactions where date between ({{thisweek}})) as count_thisweek,
(select sum(amount) from transactions where date between ({{thisweek}})) as amount_thisweek,
etc...
Is there a better way?
although you have a marked solution, I have another that will probably simplify your query even further using MySQL variables so you don't have to mis-type / calculate dates and such...
Instead of declaring variables up front, you can do them inline as a select statement, then use them as if they were columns in another table. Since it is created as a single row, there is no Cartesian result. First the query, then I'll describe the computations on it.
select
sum( if( t.date >= #today AND t.date < #tomorrow, 1, 0 )) as TodayCnt,
sum( if( t.date >= #today AND t.date < #tomorrow, amount, 0 )) as TodayAmt,
sum( if( t.date >= #yesterday AND t.date < #today, 1, 0 )) as YesterdayCnt,
sum( if( t.date >= #yesterday AND t.date < #today, amount, 0 )) as YesterdayAmt,
sum( if( t.date >= #FirstOfWeek AND t.date < #EndOfWeek, 1, 0 )) as WeekCnt,
sum( if( t.date >= #FirstOfWeek AND t.date < #EndOfWeek, amount, 0 )) as WeekAmt
from
transations t,
( select #today := curdate(),
#yesterday := date_add( #today, interval -1 day ),
#tomorrow := date_add( #today, interval 1 day ),
#FirstOfWeek := date_add( #today, interval +1 - dayofweek( #today) day ),
#EndOfWeek := date_add( #FirstOfWeek, interval 7 day ),
#minDate := least( #yesterday, #FirstOfWeek ) ) sqlvars
where
t.date >= #minDate
AND t.date < #EndOfWeek
Now, the dates. Since the #variables are prepared in sequence, you can think of it as an inline program to set the variables. Since they are a pre-query, they are done first and available for the duration of the rest of the query as previously stated. So to start, I am working with whatever "curdate()" is which gets the date portion only without respect to time. From that, subtract 1 day (add -1) to get the beginning of yesterday. Add 1 day to get Tomorrow. Then, the first of the week is whatever the current date is +1 - the actual day of week (you will see shortly). Add 7 days from the first of the week to get the end of the week. Finally, get whichever date is the LEAST between a yesterday (which COULD exist at the end of the prior week), OR the beginning of the week.
Now look at today for example... Feb 23rd.
Sun Mon Tue Wed Thu Fri Sat Sun
21 22 23 24 25 26 27 28
Today = 23
Yesterday = 22
Tomorrow = 24
First of week = 23 + 1 = 24 - 3rd day of week = 21st
End of Week = 21st + 7 days = 28th.
Why am I doing a cutoff of the dates stripping times? To simplify the SUM() condition for >= AND <. If I stated some date = today, what if your transactions were time-stamped. Then you would have to extract the date portion only to qualify. By this approach, I can say that "Today" count and amount is any date >= Feb 23 at 12am midnight AND < Feb 24th 12 am midnight. This is all time inclusive Feb 23rd up to 11:59:59pm hence LESS than Feb 24th (tomorrow).
Similar consideration for yesterday is all inclusive UP TO but not including whatever "today" is. Similarly for the week range.
Finally the WHERE clause is looking for the earliest date as the range so it does not have to run through the entire database of transactions to the end.
Lastly, if you ever wanted the counts and totals for a prior week / period, whatever, you could just extrapolate and change
#today := '2015-01-24'
and the computations will be AS IF the query was run ON THAT DATE.
Similar if you cared to alter such as for a month, you could compute the first of the month to the first of a following month for MONTHLY totals.
Hope you enjoy this flexible solution to you.
Yes, you can use aggregate functions on conditional expressions, like so:
SELECT SUM(IF(date between ({{today}})), 1, 0) AS count_today
, SUM(IF(date between ({{today}})), amount, 0) AS amount_today
, ...
I have a MySQL database containing discounts. A simplified version looks like this:
id | start (UNIX timestamp) | end (UNIX timestamp)
45 | 1384693200 | 1398992400
68 | 1386018000 | 1386277200
263 | 1388530800 | 1391209200
A discount can last a few days, a few months, or even a few years. I'm looking for a way to select a unique list of months where (future) discounts are valid.
If there is:
a discount which starts in november 2013 and ends in april 2014
a discount which starts in december 2013 and ends in the same month
a discount which starts in january 2014 and ends one month later
a discount which starts in june 2014 and ends the same month
The output should be:
- December (2013)
- January (2014)
- February (2014)
- March (2014)
- April (2014)
- June (2014)
November 2013 is not shown because it is in the past. May 2014 is not shown because there is no discount in that month.
Can somebody help?
Thanks in advance!
Create a table containing a sequence of numbers from 0 to a number of month you could ever require, and join this table to your table.
This is example how to get a list of years+months separately for each id
SELECT id,
year( start + interval x month ) year,
month( start + interval x month ) month
FROM
numbers n
JOIN
(
SELECT id,
from_unixtime( start ) start,
from_unixtime( end ) end
FROM Table1
) q
ON n.x <= period_diff( date_format( q.end, '%Y%m' ),date_format( q.start, '%Y%m' ))
ORDER BY id, year, month ;
Demo --> http://www.sqlfiddle.com/#!9/d7cfc/4
If you want to combine years+months for all id, skip id column and use GROUP BY
SELECT year( start + interval x month ) year,
month( start + interval x month ) month
FROM
numbers n
JOIN
(
SELECT id,
from_unixtime( start ) start,
from_unixtime( end ) end
FROM Table1
) q
ON n.x <= period_diff( date_format( q.end, '%Y%m' ),date_format( q.start, '%Y%m' ))
GROUP BY year, month
ORDER BY year, month ;
If you want to skip past years and months, add WHERE year >= current year AND month >= current month, this is a trivial change. Also add another WHERE end < current-unix-time in the subquery to filter out unwanted past rows.