MySQL Getting results for a specific user - mysql

I have the following query which works as expected.
SELECT T.*, #curRank := #curRank + 1 AS rank
FROM ( SELECT e.guid,
e.name,
(SELECT COUNT(ev.event_vote_id)
FROM event_vote ev
WHERE ev.event_uid = s.guid) AS votes
FROM event e
) as T
CROSS JOIN (SELECT #curRank := 0) r
ORDER BY votes DESC
It returns the guid, name, vote count and rank of all events.
However I want to make it specific to a certain user by linking the user_event table using something like the following:
JOIN user_event ON t.guid = ue.event_uid
WHERE ue.user_uid = 'abc123'
However i'm unsure on where to put this or if this.
I have the following query as a start but it returns exactly the opposite to expected i.e. every event not belonging to the user.
SELECT t.*
FROM user_event ue
JOIN ( SELECT e.guid,
e.name,
e.ownerId,
e.thumbnailSrc,
#curRank := #curRank + 1 AS rank,
( SELECT COUNT(ev.event_vote_id)
FROM event_vote ev
WHERE ev.event_uid = e.guid) AS votes
FROM event e, (SELECT #curRank := 0) r
) AS t
ON t.guid = ue.event_uid
WHERE ue.user_uid = 'abc123'
ORDER BY rank ASC
expected results
list of all of the events:
guid | name | votes | rank
def test2 2 1 (user1)
abc test1 1 2 (user2)
ghi test3 0 3 (user1)
jkl test4 0 4 (user3)
what the query should return for user 1 (user1 guid being abc123)
guid | name | votes | rank
def test2 2 1
ghi test3 0 3

Let's forget about rank for a moment, if you want only event from one specific user you do:
SELECT e.*, ue.*, ( ... ) as votes
FROM event e
JOIN user_event ue
ON e.guid = ue.event_uid
WHERE ue.user_uid = 'abc123'
And over that result you can do ranking.
SELECT T.*, #curRank := #curRank + 1 AS rank
FROM ( ... previous query ... ) as T
CROSS JOIN (SELECT #curRank := 0) r
ORDER BY votes DESC
EDIT:
to filter for a single user you need create another subquery.
calculate votes
calculate ranks
filter user
So query become:
SELECT *
FROM ( SELECT T.*, #curRank := #curRank + 1 AS rank
FROM ( SELECT e.*, ue.*, ( ... ) as votes
FROM event e
JOIN user_event ue
ON e.guid = ue.event_uid ) as T
CROSS JOIN (SELECT #curRank := 0) r
ORDER BY votes DESC
) as ranked_result
WHERE ranked_result.user_uid = 'abc123'

Related

Multiple Alias, one Column

I am trying to give multiple Aliases to the same column, basically, i want these two queries to be one:
SELECT name AS singlePeople FROM People
JOIN ID FROM Numbers
ON People.ID=Numbers.ID
WHERE People.isMarried=f;
SELECT name AS marriedPeople FROM People
JOIN ID FROM Numbers
ON People.ID=Numbers.ID
WHERE People.isMarried=t;
I want my results to look like:
singlePeople marriedPeople
------------- --------------
Bob Kelly John SMith
John Adams
Is this sufficient?
SELECT (CASE WHEN p.isMarried THEN 'Married' ELSE 'Single' END) as which,
name
FROM People p JOIN
Numbers n
ON p.ID = n.ID;
If not, you can do this with variables:
select max(case when not ismarried then name end) as single,
max(case when ismarried then name end) as married
from (select name, p.ismarried,
(#rn := if(#i = ismarried, #rn + 1,
if(#i := ismarried, 1, 1)
)
) as seqnum
from people p join
numbers n
on p.id = n.id cross join
(select #i := NULL, #rn := 0) params
order by ismarried
) pn
group by rn;

Select first N messages each user receives

I have a table that stores messages sent to users, the layout is as follows
id (auto-incrementing) | message_id | user_id | datetime_sent
I'm trying to find the first N message_id's that each user has received, but am completely stuck. I can do it easily on a per-user basis (when defining the user ID in the query), but not for all users.
Things to note:
Many users can get the same message_id
Message ID's aren't sent sequentially (i.e. we can send message 400 before message 200)
This is a read only mySQL database
EDIT: On second thought I removed this bit but have added it back in since someone was kind enough to work on it
The end goal is to see what % of users opened one of the first N messages they received.
That table of opens looks like this:
user_id | message_id | datetime_opened
This is an untested answer to the original question (with 2 tables and condition on first 5):
SELECT DISTINCT user_id
FROM (
SELECT om.user_id,
om.message_id,
count(DISTINCT sm2.message_id) messages_before
FROM opened_messages om
INNER JOIN sent_messages sm
ON om.user_id = sm.user_id
AND om.message_id = sm.message_id
LEFT JOIN sent_messages sm2
ON om.user_id = sm2.user_id
AND sm2.datetime_sent < sm.datetime_sent
GROUP BY om.user_id,
om.message_id
HAVING messages_before < 5
) AS base
The subquery joins in sm2 to count the number of preceding messages that were sent to the same user, and then the having clause makes sure that there are fewer than 5 earlier messages sent. As for the same user there might be multiple messages (up to 5) with that condition, the outer query only lists the unique users that comply to the condition.
To get the first N (here 2) messages, try
SELECT
user_id
, message_id
FROM (
SELECT
user_id
, message_id
, id
, (CASE WHEN #user_id != user_id THEN #rank := 1 ELSE #rank := #rank + 1 END) AS rank,
(CASE WHEN #user_id != user_id THEN #user_id := user_id ELSE #user_id END) AS _
FROM (SELECT * FROM MessageSent ORDER BY user_id, id) T
JOIN (SELECT #cnt := 0) c
JOIN (SELECT #user_id := 0) u
) R
WHERE rank < 3
ORDER BY user_id, id
;
which uses a RANK substitute, derived from #Seaux response to Does mysql have the equivalent of Oracle's “analytic functions”?
To extend this to your original question, just add the appropriate calculation:
SELECT
COUNT(DISTINCT MO.user_id) * 100 /
(SELECT COUNT(DISTINCT user_id)
FROM (
SELECT
user_id
, message_id
, id
, (CASE WHEN #user_id != user_id THEN #rank := 1 ELSE #rank := #rank + 1 END) AS rank,
(CASE WHEN #user_id != user_id THEN #user_id := user_id ELSE #user_id END) AS _
FROM (SELECT * FROM MessageSent ORDER BY user_id, id) T
JOIN (SELECT #cnt := 0) c
JOIN (SELECT #user_id := 0) u
) R2
WHERE rank < 3
) AS percentage_who_read_one_of_the_first_messages
FROM MessageOpened MO
JOIN
(SELECT
user_id
, message_id
FROM (
SELECT
user_id
, message_id
, id
, (CASE WHEN #user_id != user_id THEN #rank := 1 ELSE #rank := #rank + 1 END) AS rank,
(CASE WHEN #user_id != user_id THEN #user_id := user_id ELSE #user_id END) AS _
FROM (SELECT * FROM MessageSent ORDER BY user_id, id) T
JOIN (SELECT #cnt := 0) c
JOIN (SELECT #user_id := 0) u
) R
WHERE rank < 3) MR
ON MO.user_id = MR.user_id
AND MO.message_id = MR.message_id
;
With no CTEs in MySQL, and being in a read-only database - I see no way around having the above query twice in the statement.
See it in action: SQL Fiddle.
Please comment if and as this requires adjustment / further detail.

Mysql query counter x/y

It's possible to create a query that return the x/y number of records?
Eg.
I have table like this
ID | id_user | id_event
23 | 3 | 1
24 | 3 | 1
25 | 3 | 1
26 | 4 | 2
27 | 4 | 2
I will return something that looks like this:
Event
id_user 3 -> **1/3**
id_user 3 -> **2/3**
id_user 3 -> **3/3**
id_user 4 -> **1/2**
id_user 4 -> **2/2**
Any suggestion is appreciated!
Try this
SET #id_event := 0;
SELECT CONCAT('id_user ', id_user ,'->','**', (#id_event := #id_event + 1) ,'/', id_user ,** ) from table
This is probably a duplicate to this question.
SELECT CONCAT('id_user ',id_user,' -> **',rank,'/',group_total,'**') FROM (
SELECT id,
group_total,
CASE id_user
WHEN #id_user THEN
CASE id_event
WHEN #id_event THEN #rowno := #rowno + 1
ELSE #rowno := 1
END
ELSE #rowno :=1
END AS rank,
#id_user := id_user AS id_user,
#id_event := id_event AS id_event
FROM event_table
JOIN (SELECT id_user, id_event, COUNT(*) group_total FROM event_table GROUP BY id_user, id_event) t USING (id_user, id_event)
JOIN (SELECT #rowno := 0, #id_user := 0, #id_event := 0) r
ORDER BY id_user, id_event
) c;
Assuming you want output like this:
id_user < id_user > ** serial number of event related to this user / total events related to this user **
You can accomplish such result by the following query:
SELECT
CONCAT('id_user ',UE.id_user,' -> **',IF(#userID = UE.id_user, #eventNumber := #eventNumber + 1, #eventNumber := 1),'/',t.totalEvents,'**') AS output,
#userID := UE.id_user
FROM (SELECT #userID := -1, #eventNumber := 1) var,user_events UE
INNER JOIN
(
SELECT
id_user,
COUNT(id_event) totalEvents
FROM user_events
GROUP BY id_user
) AS t
ON UE.id_user = t.id_user
ORDER BY UE.id_user;
SQL FIDDLE DEMO
More:
SQL FIDDLE DEMO 2
This particular fiddle returns only the desired output column whereas the first fiddle contains one extra column
I played a little bit and that would be my solution:
SELECT id, id_user, id_event, if(#n = a.id_event, #c:=#c+1, if(#n:=a.id_event, #c:=1, #c:=1)) as count, (SELECT count(*) from TABLE b WHERE a.id_user = b.id_user) as total, from TABLE a join (SELECT #n:= "", #c:=1) c
It just have two if conditions for counting a #c up if #n and id_user matches if not #n become id_user and #c is 1 again. The join is for initialize the var in the same query.
Thx to that question, i found the answer to a questions that i asked 4 days ago.

What's the SQL idiom for zipping — in the functional sense — two queries?

For example, if I have a set of classes and a set of classrooms, and I want to pair the two up with some arbitrary pairing:
> SELECT class_name FROM classes ORDER BY class_name
Calculus
English
History
> SELECT room_name FROM classrooms ORDER BY room_name
Room 101
Room 102
Room 201
I'd like to "zip" them like this:
> SELECT class_name FROM classes ORDER … ZIP SELECT room_name FROM classrooms ORDER …
Calculus | Room 101
English | Room 102
History | Room 201
Currently I'm dealing with MySQL… but possibly — optimistically? — there is a reasonably standards compliant way to do this?
One way to do it in MySql
SELECT c.class_name, r.room_name
FROM
(
SELECT class_name, #n := #n + 1 rnum
FROM classes CROSS JOIN (SELECT #n := 0) i
ORDER BY class_name
) c JOIN
(
SELECT room_name, #m := #m + 1 rnum
FROM classrooms CROSS JOIN (SELECT #m := 0) i
ORDER BY room_name
) r
ON c.rnum = r.rnum
Output:
| CLASS_NAME | ROOM_NAME |
-------------|-----------|
| Calculus | Room 101 |
| English | Room 102 |
| History | Room 201 |
Here is SQLFIddle demo
Same thing in Postgres will look like
SELECT c.class_name, r.room_name
FROM
(
SELECT class_name,
ROW_NUMBER() OVER (ORDER BY class_name) rnum
FROM classes
) c JOIN
(
SELECT room_name,
ROW_NUMBER() OVER (ORDER BY room_name) rnum
FROM classrooms
) r
ON c.rnum = r.rnum
Here is SQLFiddle demo
And in SQLite
SELECT c.class_name, r.room_name
FROM
(
SELECT class_name,
(SELECT COUNT(*)
FROM classes
WHERE c.class_name >= class_name) rnum
FROM classes c
) c JOIN
(
SELECT room_name,
(SELECT COUNT(*)
FROM classrooms
WHERE r.room_name >= room_name) rnum
FROM classrooms r
) r
ON c.rnum = r.rnum
Here is SQLFiddle demo
This is a form of join, but you need to create the join key. Alas, though, this requires a full outer join, because you do not know which list is longer.
So, you can do this by using variables to enumerate the rows and then using union all and group by to get the values:
select max(case when which = 'class' then name end) as class_name,
max(case when which = 'room' then name end) as room_name
from ((SELECT class_name as name, #rnc := #rnc + 1 as rn, 'class' as which
FROM classes cross join
(select #rnc := 0) const
ORDER BY class_name
) union all
(select room_name, #rnr := #rnr + 1 as rn, 'room'
from classrooms cross join
(select #rnr := 0) const
ORDER BY room_name
)
) t
group by rn;

query to add incremental field based on GROUP BY

Have a table photos
photos.id
photos.user_id
photos.order
A) Is it possible via a single query to group all photos by user and then update the order 1,2,3..N ?
B) added twist, what if some of the photos already have an order value associated? Make sure that the new photos.order never gets repeated and fills in ant orders lower or higher than those existing (as best as possible)
My only thought is just to run a script on this and loop through it and re'order' everything?
photos.id int(10)
photos.created_at datetime
photos.order int(10)
photos.user_id int(10)
Right now data may look like this
user_id = 1
photo_id = 1
order = NULL
user_id = 2
photo_id = 2
order = NULL
user_id = 1
photo_id = 3
order = NULL
the desired result would be
user_id = 1
photo_id = 1
order = 1
user_id = 2
photo_id = 2
order = 1
user_id = 1
photo_id = 3
order = 2
A)
You can use a variable that increments with each row and resets with each user_ID to get the row count.
SELECT ID,
User_ID,
`Order`
FROM ( SELECT #r:= IF(#u = User_ID, #r + 1,1) AS `Order`,
ID,
User_ID,
#u:= User_ID
FROM Photos,
(SELECT #r:= 1) AS r,
(SELECT #u:= 0) AS u
ORDER BY User_ID, ID
) AS Photos
Example on SQL Fiddle
B)
My First solution was to just add Order to the sorting that adds the row number, therefore anything with an Order Gets sorted by its order first, but this only works if your ordering system has no gaps and starts at 1:
SELECT ID,
User_ID,
RowNumber AS `Order`
FROM ( SELECT #r:= IF(#u = User_ID, #r + 1,1) AS `RowNumber`,
ID,
User_ID,
#u:= User_ID
FROM Photos,
(SELECT #i:= 1) AS r,
(SELECT #u:= 0) AS u
ORDER BY User_ID, `Order`, ID
) AS Photos
ORDER BY `User_ID`, `Order`
Example using Order Field
ORDERING WITH GAPS
I have eventually found a way of maintaining the sort order even when there are gaps in the sequence.
SELECT ID, User_ID, `Order`
FROM Photos
WHERE `Order` IS NOT NULL
UNION ALL
SELECT Photos.ID,
Photos.user_ID,
Numbers.RowNum
FROM ( SELECT ID,
User_ID,
#r1:= IF(#u1 = User_ID,#r1 + 1,1) AS RowNum,
#u1:= User_ID
FROM Photos,
(SELECT #r1:= 0) AS r,
(SELECT #u1:= 0) AS u
WHERE `Order` IS NULL
ORDER BY User_ID, ID
) AS Photos
INNER JOIN
( SELECT User_ID,
RowNum,
#r2:= IF(#u2 = User_ID,#r2 + 1,1) AS RowNum2,
#u2:= User_ID
FROM ( SELECT DISTINCT p.User_ID, o.RowNum
FROM Photos AS p,
( SELECT #i:= #i + 1 AS RowNum
FROM INFORMATION_SCHEMA.COLLATION_CHARACTER_SET_APPLICABILITY,
( SELECT #i:= 0) AS i
) AS o
WHERE RowNum <= (SELECT COUNT(*) FROM Photos P1 WHERE p.User_ID = p1.User_ID)
AND NOT EXISTS
( SELECT 1
FROM Photos p2
WHERE p.User_ID = p2.User_ID
AND o.RowNum = p2.`Order`
)
AND p.`Order` IS NULL
ORDER BY User_ID, RowNum
) AS p,
(SELECT #r2:= 0) AS r,
(SELECT #u2:= 0) AS u
ORDER BY user_ID, RowNum
) AS numbers
ON Photos.User_ID = numbers.User_ID
AND photos.RowNum = numbers.RowNum2
ORDER BY User_ID, `Order`
However as you can see this is pretty complicated. This works by treating those with an order value separately to those without. The top query just ranks all photos with no order value in order of ID for each user. The bottom query uses a cross join to generates a sequential list from 1 to n for each user ID (up to the number of entries for each User_ID). So with a data set like this:
ID User_ID Order
1 1 NULL
2 2 NULL
3 1 NULL
4 1 1
5 1 3
6 2 2
7 2 3
It would generate
UserID RowNum
1 1
1 2
1 3
1 4
2 1
2 2
2 3
It then uses NOT EXISTS to elimiate all combinations already used by Photos with a non null order, and ranked in order of RowNum partitioned by User_ID giving
UserID RowNum Rownum2
1 2 1
1 4 2
2 1 1
The RowNum2 value can then be matched with the rownum value achieved in the from subquery, giving the correct order value. Long winded, but it works.
Example on SQL Fiddle
Worked for me. I needed to increment version grouping by 4 fields (host, folder, fileName, status) and sort by 1 (downloadedAtTicks).
This is is my SELECT
SET #status := NULL;
SET #version := NULL;
SELECT
id,
host,
folder,
fileName,
status,
downloadedAtTicks,
version,
IF(IF(status IS NULL, 0, status) = #status, #version := #version + 1, #version := 0) AS varVersion,
#status := IF(status IS NULL, 0, status) AS varStatus
FROM csvsource
ORDER BY host, folder, fileName, status, downloadedAtTicks;
And this is my UPDATE
SET #status := NULL;
SET #version := NULL;
UPDATE
csvsource csv,
(SELECT
id,
IF(IF(status IS NULL, 0, status) = #status, #version := #version + 1, #version := 0) AS varVersion,
#status := IF(status IS NULL, 0, status) AS varStatus
FROM csvsource
ORDER BY host, folder, fileName, status, downloadedAtTicks) AS sub
SET
csv.version = sub.varVersion
WHERE csv.id = sub.id;