I have table:
id date default(bool)
1 2015-01-01 0
2 2015-01-02 0
3 2015-01-03 1
4 2015-01-04 1
5 2015-01-05 1
6 2015-01-06 0
7 2015-01-07 0
8 2015-01-08 1
9 2015-01-09 1
10 2015-01-10 0
I want only rows where ordered by date rows change default column from 0 to 1, so in this table rows: 3 and 8.
If id can be used instead of date, use a correlated sub-query to read previous row's value:
select *
from tablename t1
where default = 1
and (select default from tablename t2
where t2.id = t1.id - 1) = 0
Use date instead to find previous row:
select *
from tablename t1
where default = 1
and (select default from tablename t2
where t2.date = (select max(date) from tablename
where date < t1.date)) = 0
Here is another way using dynamic variable
select t.id,t.date,t.`default`
from (
select
t1.*,
#default := if(#prev_default = 0 and t1.`default` = 1,1,0) as def,
#prev_default:=t1.`default`
from (
select * from test order by date
)t1,
(select #prev_default:= 2,#default:=0)r
)t
where t.def = 1 ;
Related
Have a table data structure like below:
id
regid
docid
archived
1
1000
1
0
2
1000
2
0
3
1000
3
0
4
2000
1
0
5
2000
2
0
6
3000
1
0
7
3000
2
0
8
3000
3
0
9
3000
4
0
What I'm trying to do update the archived column to 1 where the docid is less than the max docid, by each regid group.
So I should end up with id's 3, 5 & 9 not being set to 1
Have tried:
update table t1
join (select max(docid) as maxdocid, regid from table) t2 on t1.docid < t2.maxdocid and t1.regid = t2.regid
set t1.archived = 1
But doesn't work, only does the first regid group.
Here's a solution (in MySQL 8.0+) using a CTE:
WITH numbered_table AS (
SELECT id, ROW_NUMBER() OVER (PARTITION BY regid ORDER BY docid DESC) AS rownum
FROM mytable
)
UPDATE mytable JOIN numbered_table USING (id)
SET archived = 1
WHERE rownum > 1
AND archived = 0;
Second solution, if you use an older version of MySQL that doesn't support CTE syntax:
You don't really need to compute the max docid value. If you want to update all rows except for the row with the max docid value, then you can check if a row can be matched to any other row with a greater docid value.
UPDATE mytable AS t1
INNER JOIN mytable AS t2 ON t1.regid = t2.regid AND t1.docid < t2.docid
SET t1.archived = 1
WHERE t1.archived = 0;
This will be true for all rows except the row with the max value. That row will be excluded automatically by the join.
In steps:
Create a query with the MAX value, per docid:
SELECT
ID,
regid,
docid,
(SELECT MAX(docid) FROM t1 te where te.regid=t.regid) as M
FROM t1 t
Join the result, and update:
UPDATE t1
JOIN (
SELECT
ID,
regid,
docid,
(SELECT MAX(docid) FROM t1 te where te.regid=t.regid) as M
FROM t1 t
) x ON t1.id=x.id
SET archived = 1
WHERE t1.docid<x.M AND t1.archived=0;
see: DBFIDDLE
You could try:
update test_tbl t1
set t1.archived = 1
where t1.archived = 0
and t1.id not in ( select t2.id
from (select max(id) as id,
regid,
max(docid)
from test_tbl
group by regid
) as t2
) ;
Result:
id regid docid archived
1 1000 1 1
2 1000 2 1
3 1000 3 0
4 2000 1 1
5 2000 2 0
6 3000 1 1
7 3000 2 1
8 3000 3 1
9 3000 4 0
Demo
Or you can use a LEFT JOIN
update test_tbl t1
left join ( select max(id) as id,
regid,
max(docid) as docid
from test_tbl
group by regid
) as t2 on t1.id=t2.id
set t1.archived = 1
where t1.archived = 0
and t2.id IS NULL
Demo
Use a self join in the update statement:
UPDATE tablename t1
INNER JOIN tablename t2
ON t2.regid = t1.regid AND t2.docid > t1.docid
SET t1.archived = 1;
See the demo.
Suppose, I have a table t1 looking like
id
value1
value2
wk_id
1
2
0
1
1
1
1
2
1
3
0
3
2
2
1
2
2
2
0
3
3
1
0
2
3
2
0
4
3
3
0
5
And I want to sum up the value1 till non-zero value appears on the value2 for first time.
End product must look like this:
id
value1
1
2
2
0
3
6
How to perform this in SQL?
If your MySQL version support window function you can try to use SUM window function with condition aggregate function be a flag to represent your logic (till non-zero value appears on the value2 for first time)
Then do condition aggregate function again.
Query #1
SELECT id,
SUM(CASE WHEN flag = 0 THEN value1 ELSE 0 END) value1
FROM (
SELECT *,
SUM(CASE WHEN value2 = 1 THEN -1 ELSE 0 END) OVER(PARTITION BY ID ORDER BY wk_id) flag
FROM T
) t1
GROUP BY id;
id
value1
1
2
2
0
3
6
View on DB Fiddle
WITH cte AS (
SELECT *,
CASE WHEN SUM(value2) OVER (partition by id ORDER BY wk_id) = 0
THEN SUM(value1) OVER (partition by id ORDER BY wk_id)
ELSE 0
END sum_value1
FROM test
ORDER BY id, wk_id
)
SELECT id, MAX(sum_value1) sum_value1
FROM cte
GROUP BY id;
https://dbfiddle.uk/?rdbms=mysql_8.0&fiddle=0fcdca008e4a821f952de4d608434bcf
One option is a NOT EXISTS clause, looking for the stop row (the first row with value2 = 1).
select id, sum(value1)
from mytable
where not exists
(
select null
from mytable stoprow
where stoprow.id = mytable.id
and stoprow.wk_id <= mytable.wk_id
and stoprow.value2 = 1
)
group by id
order by id;
Aggregating on a calculated rolling total of value2 can also be done via a self-join.
select id
, sum(if(roll_tot_value2=0,value1,0)) as total
from
(
select t1a.id, t1a.wk_id, t1a.value1
, sum(t1b.value2) as roll_tot_value2
from t1 as t1a
join t1 as t1b
on t1b.id = t1a.id
and t1b.wk_id <= t1a.wk_id
group by t1a.id, t1a.wk_id, t1a.value1
) q
group by id;
id
total
1
2
2
0
3
6
Test on db<>fiddle here
You can use a subquery:
select t.id, coalesce(
(select sum(t2.value1) from t1 t2 where t2.value2 = 0 and t2.id = t.id
and (not exists (select 1 from t1 t3 where t3.value2 = 1 and t3.id = t.id)
or t2.wk_id < (select min(t4.wk_id) from t1 t4 where t4.id = t.id and t4.value2 = 1))), 0)
from t1 t group by t.id
Table: mytable
id created_on status purchase_price_without_tax purchase_tax
1 2015-08-07 00:00:00 1 3322.80 5.00
1 2015-07-15 15:16:23 7 3599.00 71.98
1 2015-07-14 15:16:23 2 3599.00 71.98
1 2015-07-16 15:16:23 4 3599.00 91.98
1 2015-07-17 15:16:23 5 3599.00 21.98
here is my query
SELECT * FROM mytable
WHERE
(IF( `status` IN('1','2') , (`status` IN('1','2')), (`status` IN('7')) ) )
AND id=1
But not getting proper result, because I would like it to be like this:
where
IF (`status`=1 OR `status`=2 )
then status IN ('1','2')
ELSE
`status`=7
Note : If status is 1 or 2 then it will left all other status row and select only those rows which have status 1 OR 2 . If there is no rows which have status as 1 or 2, then select only those rows which have status = 7
Use a UNION of two queries that test the two cases.
SELECT *
FROM mytable
WHERE status IN (1, 2)
AND EXISTS (SELECT 1 FROM mytable WHERE status IN (1, 2))
UNION
SELECT *
FROM mytable
WHERE status = 7
AND NOT EXISTS (SELECT 1 FROM mytable WHERE status IN (1, 2))
ID GRP VAL CHK
--- ----- ----- ----
1 1 1 0
2 1 3 0
3 2 7 0
4 2 2 0
5 2 1 0
6 3 5 0
I want to set my CHK field to '1' having maximum value VAL for every group of GRP,
so ID 2,3,6 should be set.
I don't write my trials here, all seems rubbish :)
In MySQL, you can do this using the update/join syntax:
update table t join
(select grp, max(val) as maxval
from table t
group by grp
) tmax
on t.grp = tmax.grp and t.val = tmax.maxval
set t.chk = 1;
I have a column with two columns. one is TIMESTAMP and the other DIGITAL_BIT.
The value digital bit can be either 0 or 1 and changes a few times during the day. Every minute of the day is stored in this table. I would need to read somehow how many times a day this value changed from 0 to 1.
Is it possible to make a query that returns the count of this changes? What I have in mind is something like this:
select * from mytable where digital_bit = 1 and digital_bit (of previous row) = 0 order by timestamp
Can this be done with a query or do i have to process all data in my program?
Thanks
SAMPLE
timestamp | digital_bit
100000 | 0
100001 | 0
100002 | 1
100003 | 1
100004 | 0
100005 | 1
100006 | 0
100007 | 0
100008 | 1
the above should return 3 because for 3 times the value digital passed from 0 to 1. i need to count how often the value digital CHANGES from 0 to 1.
Here you go. This will get you a count of how many times digital_bit switched from 0 to 1 (in your example, this will return 3).
SELECT COUNT(*)
FROM mytable curr
WHERE curr.digital_bit = 1
AND (
SELECT digital_bit
FROM mytable prev
WHERE prev.timestamp < curr.timestamp
ORDER BY timestamp DESC
LIMIT 1
) = 0
SQLFiddle link
(Original answer relied on the timestamps being sequential: e.g. no jumps from 100001 to 100003. Answer has now been updated not to have that restriction.)
IF you have a result once per minte, you can simple join the table with itself, and
use timestamp+1 as well as leftbit != rightbit as join condition.
http://sqlfiddle.com/#!8/791c0/6
ALL Changes:
SELECT
COUNT(*)
FROM
test a
INNER JOIN
test b
ON
a.digital_bit != b.digital_bit
AND b.timestamp = a.timestamp+1;
Changes from 0 to 1
SELECT
COUNT(*)
FROM
test a
INNER JOIN
test b
ON
a.digital_bit = 0 AND
a.digital_bit != b.digital_bit
AND b.timestamp = a.timestamp+1;
Changes from 1 to 0
SELECT
COUNT(*)
FROM
test a
INNER JOIN
test b
ON
a.digital_bit = 1 AND
a.digital_bit != b.digital_bit
AND b.timestamp = a.timestamp+1;
Adapted from: How do I query distinct values within multiple sub record sets
select count(*)
from (select t1.*,
(select digital_bit
from table t2
where t2.timestamp < t1.timestamp
order by timestamp desc LIMIT 1
) as prevvalue
from table t1
) t1
where prevvalue <> digital_bit and digital_bit = 1;
This isn't likely to be efficient with a lot of data, but you can get all the rows and calculate a sequence number for them, then do the same again but with the sequence number offset by 1. Then join the 2 lots together where those calculated sequence numbers match but the first one has a digital bit of 0 and the other a digital bit of 1:-
SELECT COUNT(*)
FROM
(
SELECT mytable.timestamp, mytable.digital_bit, #aCount1:=#aCount1+1 AS SeqCount
FROM mytable
CROSS JOIN (SELECT #aCount1:=1) sub1
ORDER BY timestamp
) a
INNER JOIN
(
SELECT mytable.timestamp, mytable.digital_bit, #aCount2:=#aCount2+1 AS SeqCount
FROM mytable
CROSS JOIN (SELECT #aCount2:=0) sub1
ORDER BY timestamp
) b
ON a.SeqCount = b.SeqCount
AND a.digital_bit = 0
AND b.digital_bit = 1
EDIT - alternative solution and I would be interested to see how this performs. It avoids the need for adding a sequence number and also avoids a correlated sub query:-
SELECT COUNT(*)
FROM
(
SELECT curr.timestamp, MAX(curr2.timestamp) AS MaxTimeStamp
FROM mytable curr
INNER JOIN mytable curr2
ON curr.timestamp > curr2.timestamp
AND curr.digital_bit = 1
GROUP BY curr.timestamp
) Sub1
INNER JOIN mytable curr
ON Sub1.MaxTimeStamp = curr.timestamp
AND curr.digital_bit = 0
As I understood you have one query every minute. So you have no problem with performance.
You can add flag:
timestamp | digital_bit | changed
100000 | 0 | 0
100001 | 0 | 0
100002 | 1 | 1
100003 | 1 | 0
100004 | 0 | 1
100005 | 1 | 1
100006 | 0 | 1
100007 | 0 | 0
100008 | 1 | 1
And make check before insert:
SELECT digital_bit
FROM table
ORDER BY timestamp DESC
LIMIT 1
and if digital_bit is different insert new row with flag.
And then you just can take COUNT of flags:
SELECT COUNT(*)
FROM table
WHERE DATE BETWEEN (start, end)
AND changed = 1
Hope will see in answers better solution.