I am trying to get select from a table and return row based on values of a column. Below is data and desired output. If column EmpRecord has multiple values not null to be returned, if it has only null then it should be returned.
Data Table
EmployeeNo EmpRecord
1 A
1 NULL
2 a
3 NULL
4 NULL
4 A
4 aa
Output
EmployeeNo EmpRecord
1 A
2 a
3 NULL
4 A
4 aa
Any advice on how to go ahead with it would be great?
Regards,
Sid
The first half of the UNION query below simply strips off records for which the EmpRecord be NULL. This almost gets the job done, except that for employees who have only one more NULL records, this would remove them from the result set as well. So the second part of the UNION adds these employees back as a single record with their employee number and NULL placeholder for the record.
SELECT t1.EmployeeNo,
t1.EmpRecord
FROM yourTable t1
WHERE t1.EmpRecord IS NOT NULL
UNION ALL
SELECT t2.EmployeeNo,
NULL AS EmpRecord
FROM yourTable t2.
GROUP BY t2.EmployeeNo
HAVING SUM(CASE WHEN t2.EmpRecord IS NULL THEN 1 ELSE 0 END) = COUNT(*)
Related
So this is my table structure
learning_paths
id
name
version
created_at
updated_at
learning_path_levels
id
name
learning_path_id
order
created_at
updated_at
learning_path_level_nodes
id
name
description
documentation_links
evaluation_methodology
learning_path_level_id
created_at
updated_at
learning_path_node_users
id
learning_path_level_node_id
user_id
evaluated_by
evaluated_at
is_successful
created_at
updated_at
I'm writing a query to retrieve the learning_path_name, count of the amount of levels each learning path has, the pending and completed nodes per level for the user, and the total amount of nodes per level.
I have the following query
select learning_paths."name",
sum(case when learning_path_node_users.is_successful and learning_path_node_users.user_id is not null then 1 else 0 end) as completed_nodes,
sum(case when learning_path_node_users.is_successful = false or learning_path_node_users.user_id is null then 1 else 0 end) as pending_nodes,
count(learning_path_levels.id) as total_levels,
count(*) as total_nodes
from learning_path_level_nodes
inner join learning_path_levels on learning_path_levels.id = learning_path_level_nodes.learning_path_level_id
inner join learning_paths on learning_paths.id = learning_path_levels.learning_path_id
left join learning_path_node_users on learning_path_node_users.learning_path_level_node_id = learning_path_level_nodes.id
group by learning_paths."name"
which returns:
name
completed_nodes
pending_nodes
total_levels
total_nodes
Devops
5
3
8
8
QA
0
1
1
1
Project manager
3
3
6
6
AI
0
5
5
5
Everything is correct, except for the levels count,
for example, for Devops,it should be 2, and it is returning 8
for Project Manager it should be 2, and it is returning 6
a pattern I see is that it returns the amount of nodes as the amount of levels,
How can I fix this?
I'd really appreciate any help or suggestions, as I've been struggling with this.
Thanks in advance
EDIT: As per your suggestion, I'm attaching a fiddle with the tables and data.
https://dbfiddle.uk/?rdbms=postgres_14&fiddle=f29676ff7051686a28de96928db1e3a6
While I don't get the exact results you want, I think you want to add a distinct to your count for the total levels:
select
lp.name,
sum(case when u.is_successful and u.user_id is not null then 1 else 0 end) as completed_nodes,
sum(case when u.is_successful = false or u.user_id is null then 1 else 0 end) as pending_nodes,
count(distinct lpl.id) as total_levels, -- added "distinct"
array_agg (lpl.id) as level_detail, -- debugging aid
count(*) as total_nodes
from
learning_path_level_nodes n
join learning_path_levels lpl on lpl.id = n.learning_path_level_id
join learning_paths lp on lp.id = lpl.learning_path_id
left join learning_path_node_users u on u.learning_path_level_node_id = n.id
group by
lp.name
To help expose the rationale, I added the field level_detail, which you can delete, to show why the results are what they are. You can obviously remove that once the results are what you want.
If it's not what you expect, perhaps you can explain or give by example what I might be missing.
I'd like to check if a value exists in an array, and if it does not, I'd like to return a NULL row for it, instead of no row.
SELECT
users.id
FROM
users
WHERE
users.name = ANY('{ John, avocado, Carl }'::text[]);
Currently returns
id
1
2
I'd like it to return
id
1
NULL
2
Since avocado is not present in our users table.
SELECT users.id
FROM unnest('{John, avocado, Carl}'::text[]) WITH ORDINALITY AS a(name, num)
LEFT JOIN users USING (name)
ORDER BY a.num;
id
----
2
1
(3 rows)
There are many similar questions which I've learned from, but my result set isn't returning the expected results.
My Objective:
Build a query that will return a result set containing all rows in table demo1 with user_id = "admin", and the only row of table demo2 with user_id = "admin". Each row in demo2 has a unique user_id so there's always only one row with "admin" as user_id.
However, I don't want demo2 data to wastefully repeat on every subsequent row of demo1. I only want the first row of the result set to contain demo2 data as non-null values. Null values for demo2 columns should only be returned for rows 2+ in the result set.
Current Status:
Right now my query is returning the appropriate columns (all demo1 and all demo2) but
all the data returned from demo2 is null.
Demo1:
id user_id product quantity warehouse
1 admin phone 3 A
2 admin desk 1 D
3 k45 chair 5 B
Demo2:
id user_id employee job country
1 admin james tech usa
2 c39 cindy tech spain
Query:
SELECT *
from demo1
left join (SELECT * FROM demo2 WHERE demo2.user_id = 'X' LIMIT 1) X
on (demo1.user_id = x.user_id)
WHERE demo1.user_id = 'admin'
Rationale:
The subquery's LIMIT 1 was my attempt to retrieve demo2 values for row 1 only, thinking the rest would be null. Instead, all values are null.
Current Result:
id user_id product quantity warehouse id employee job country
1 admin phone 3 A null null null null
2 admin desk 1 D null null null null
Desired Result:
id user_id product quantity warehouse id employee job country
1 admin phone 3 A 1 james tech usa
2 admin desk 1 D null null null null
I've tried substituting left join for left inner join, right join, full join, but nothing returns the desired result.
Your join is going to bring through ANY records that satisfies the join condition for your two tables. There is no changing that.
But you could suppress subsequent records in your result set from displaying the matching demo2 record that satisfied the join condition AFTER it's joined:
SELECT demo1.id ,
demo1.user_id,
demo1.product,
demo1.quantity,
demo1.warehouse
CASE WHEN ROW_NUMBER() OVER (PARTITION BY demo1.user_id ORDER BY demo1.id) = 1 THEN demo2.id END as demo2_id,
CASE WHEN ROW_NUMBER() OVER (PARTITION BY demo1.user_id ORDER BY demo1.id) = 1 THEN demo2.employee END AS demo2_employee,
CASE WHEN ROW_NUMBER() OVER (PARTITION BY demo1.user_id ORDER BY demo1.id) = 1 THEN demo2.job END as demo2_job,
CASE WHEN ROW_NUMBER() OVER (PARTITION BY demo1.user_id ORDER BY demo1.id) = 1 THEN demo2.country END as demo2_country
from demo1
left join demo2
on demo1.user_id = demo2.user_id
AND demo2.user_id = 'X'
WHERE demo1.user_id = 'admin'
That's just a quick rewrite of your original sql with the addition CASE expressions included.
That being said, this sql will produce no results for demo2 since the demo2.user_id can't satisfy both conditions in this query:
The join condition demo1.user_id = demo2.user_id with the where predicate of demo1.user_id = 'admin'
Also hold the value X.
It's either admin and satisfies your first join condition, but fails your second. Or it's X and satisfies your second condition, but nor your first.
Here is another nice approach:
sqlfiddle
I have two table
TABLE 1 : Stage_product
PRODUCT_ID SYS_ROWDATETIMEUTC
1 2015-03-13 06:09:30.040
2 ....
3
TABLE 2 : DIM_Product
PRODUCT_ID SYS_ROWSTARTDATETIMEUTC SYS_ROWISCURRENT
1 2014-03-13 06:09:30.040 0
2 2015-03-13 06:09:30.040 1
I want to do an update statement that if the value SYS_ROWDATETIMEUTC in the first table is more recent than the value SYS_ROWSTARTDATETIMEUTC in the table, then the value SYS_ROWISCURRENT in the second table is set to 0, else 1.
You can use the following query:
UPDATE t2
SET t2.SYS_ROWISCURRENT = CASE
WHEN t1.SYS_ROWDATETIMEUTC > t2.SYS_ROWSTARTDATETIMEUTC THEN 0
ELSE 1
END
FROM Table2 t2
INNER JOIN Table1 t1 ON t2.PRODUCT_ID = t1.PRODUCT_ID
I assume you want to compare dates between the two tables for the same product.
After a complex operation (some database merge-ing) I have a table that needs to be updated based on timestamp.
JobsTable
Id Time_stamp Resource RunNumber
121 1 A 1
122 2 A 1
123 3 B 1
124 4 B 1
125 5 A 2
The point is to Update the RunNumber column incrementally for each resource based on timestamp. So in the end the expected result is:
Id Time_stamp Resource RunNumber
121 1 A 1
122 2 A 2 //changed
123 3 B 1
124 4 B 2 //changed
125 5 A 3 //changed
I tried doing this in multiple ways. Since DB2 update does not support Join or With statements I tried something like:
update JOBSTABLE JT
SET RunNumber =
(SELECT RunNumber
FROM (Select ID, ROW_NUMBER() OVER (ORDER BY TIME_STAMP ) RunNumber from JobsTable, ORDER BY TIME_STAMP) AS AAA
WHERE AAA.ID = JT.ID)
WHERE ID = ?
Error:
Assignment of a NULL value to a NOT NULL column "TBSPACEID=6, TABLEID=16, COLNO=2" is not allowed.. SQLCODE=-407, SQLSTATE=23502, DRIVER=3.64.82 SQL Code: -407, SQL State: 23502
Is this even possible? (I am aiming at doing this operation in a single query rather than using Cursors, etc..)
Thank you
Firstly, your subselect has a syntax error, which tells me it's not the exact statement that you are trying to run. The error message is pretty clear -- in your actual statement the subselect sometimes returns NULL.
Secondly, you should probably be numbering rows within a partition by resource.
Thirdly, you could probably do with a single subselect anyway -- this is based on the statement you published:
update JOBSTABLE JT
SET RunNumber =
(SELECT ROW_NUMBER() OVER (partition by resource ORDER BY TIME_STAMP )
from JobsTable where id = JT.ID)