Treatment table schema
id
doctor_id
location_id
patient_id
...
I need to be able to group patient treatments based on location and list all the doctors that were treating the patient for the specific location.
SELECT id,
(SELECT STRING_AGG(t.initials::text, ',') FROM (
SELECT DISTINCT treatments.doctor_id, doctors.initials FROM
treatments LEFT JOIN doctors ON doctors.id = treatments.doctor_id) t
) as treating_doctors_initials
GROUP BY treatments.patient_id, treatments.location_id
The problem is that this returns the same initials for all records -> it does perform the join I guess or maybe I am misunderstanding the aggregate function.
Related
I have a table of hospitals details, department details linked to it, types of workers (the staff) in different tables, and their salary information.
I want to extract the following for each hospital: the average and the sum of the salaries, the number of nurses, the number of research doctors and the number of beds in all the departments of a specific hospital.
I built this view of all the workers salary information:
CREATE VIEW workers AS
SELECT hospcod, docsal as sal, 'treatdoc' as typework
FROM doc NATURAL JOIN treatdoc NATURAL JOIN dept
UNION
SELECT hospcod, nursal, 'nurse'
FROM nurse NATURAL JOIN dept
UNION
SELECT hospcod, docsal, 'rsrchdoc'
FROM doc NATURAL JOIN rsrchdoc NATURAL JOIN lab;
the departments and the labs have the hospital code column to correlate a worker information to a specific hospital.
so I have one sceme for all the staff with their rules workers(hospital_code, salary, type_of_worker)
here is the query I'm trying to build:
SELECT hospname, sum(workers.sal), avg(workers.sal), count(dept.numbed),
(SELECT count(typework) from workers where typework = 'nurse') nurse_num,
(SELECT count(typework) from workers where typework = 'rsrchdoc') rsrchdoc_num
FROM hosp NATURAL JOIN dept NATURAL JOIN workers
GROUP BY hospname;
I want to count for each hospital, the number of nurses and the number of research doctors
but it should be correlated somehow to the different hospitals (in the above it gives me the same number of nurses / rsrchdocs for each hospital) , there should be columns that is grouped by hospnames and should get all the tuples like the salary info (avg, sum), as I got properly, but the workers information should be grouped HAVING typework = 'nurse' for the nurse_num, and for the column rsrchdoc_numit should be HAVING typework = 'rsrchdoc_num'
does someone have an idea how can I combine thouse columns in one query?
thank you!
There is an error in your query, I will try to explain.
When you do:
(SELECT count(typework) from workers where typework = 'nurse') nurse_num,
You are getting a constant, that is not affected by the "group by" you are doing after.
What you have to do is a JOIN (like you did in the view) and link the nurse and the rsrchdoc to an specific hospital.
I will give an example is pseudo code
SELECT hosp_name, sum(nurse.salary) , avg(nurse.salary)
FROM hosp
JOIN nurse ON nurse.hosp_name = hosp.hosp_name
GROUP BY hosp.hosp_name
This query will give you 1 row for each nurse in each hospital (assuming that a nurse may work in more than one hospital).
Then you have to do the same thing also for doctors, in a different operation.
SELECT hosp_name, sum(doctors.salary) , avg(doctors.salary)
FROM hosp
JOIN doctors ON doctors.hosp_name = hosp.hosp_name
GROUP BY hosp.hosp_name
And finally you will have to join both ( you may perform the sum first to make it more readable.
SELECT hosp_name, sum_sal_doc, avg_sal_doc, sum_nur_doc, avg_nur_doc
FROM hosp
LEFT JOIN ( SELECT doctors.hosp_name, sum(doctors.salary) as sum_sal_doc, avg(doctors.salary) as avg_sal_doc
FROM doctors
GROUP BY doctors.hosp_name
) t1 ON t1.hosp_name = hosp.hosp_name
LEFT JOIN ( SELECT nurses.hosp_name, sum(nurses.salary) as sum_nur_doc, avg(nurses.salary) as avg_nur_doc
FROM nurses
GROUP BY nurses.hosp_name
) t2 ON t2.hosp_name = hosp.hosp_name
There must be 1 to many relationship between hosp --> dept and hosp --> workers so if you join these 3 tables then you will definitely find the duplicates for dept and workers so you must have to create sub-query for one of the dept or workers to fetch single grouped record group by hospital as follows:
SELECT h.hospname,
sum(w.sal) total_all_worker_sal,
avg(w.sal) avg_all_workers_sal,
d.numbed,
count(case when w.typework = 'nurse' then 1 end) nurse_num,
count(case when w.typework = 'rsrchdoc' then 1 end) rsrchdoc_num
FROM hosp h
JOIN (select hospital_code , sum(numbed) numbed
-- used SUM as numbed must be number of bed in department
-- COUNT will give you only number of department if you use count(d.numbed)
from dept
group by hospital_code) d ON h.hospital_code = d.hospital_code
JOIN workers w ON h.hospital_code = d.hospital_code
GROUP BY h.hospital_code , h.hospname, d.numbed;
-- used h.hospital_code to separate the records if two hospitals have same name
I have three tables in my Oracle db:
Peoples:
IdPerson PK
Name
Surname
Earnings:
IdEarning PK
IdPerson
EarningValue
Awards:
IdAward PK
IdPerson FK
AwardDescription
So one person can have many earnings, one earning or can have no any earnings. Also one person can have many awards, one award, or no any award.
I have to select Surname and AwardDescription but only for people who have any earnings, because it is possible to have some award but, also don't have any earning!
My problem is to make a correct group by statement. I use query posted below and I am selecting surname of person with a description of award, but it is duplicating for each row in Earnings for this person.
SELECT AwardDescription, Surname
FROM Awards
INNER JOIN People ON People.IdPerson= Awards.IdPerson
INNER JOIN Earnings ON Earnings .IdPerson= People.IdPerson;
How to group it and avoid duplicating rows for each earning of person?
One person can be in many rows, but with different awards.
You could add DISTINCT to your query:
SELECT DISTINCT AwardDescription, Surname
FROM Awards
INNER JOIN People ON People.IdPerson= Awards.IdPerson
INNER JOIN Earnings ON Earnings .IdPerson= People.IdPerson;
Or another option is to use EXISTS:
SELECT AwardDescription, Surname
FROM Awards
INNER JOIN People P ON P.IdPerson= Awards.IdPerson
WHERE EXISTS (
SELECT 1
FROM Earnings E
WHERE P.IdPerson = E.IdPerson);
Do a left outer join among the tables like
SELECT p.Surname, a.AwardDescription, e.EarningValue
FROM People p
LEFT JOIN Awards a ON p.IdPerson= a.IdPerson
LEFT JOIN Earnings e ON e.IdPerson= p.IdPerson
WHERE a.AwardDescription IS NOT NULL
OR e.EarningValue IS NOT NULL;
I have a table of cars where each car belongs to a company. In another table I have a list of company locations by city.
I want to select all cars from the cars table whose company has locations on all cities passed into the stored procedure, otherwise exclude those cars all together even if it falls short of one city.
So, I've tried something like:
select id, cartype from cars where companyid in
(
select id from locations where cityid in
(
select id from cities
)
)
This doesn't work as it obviously satisfies the condition if ANY of the cities are in the list, not all of them.
It sounds like a group by count, but can't make it work with what I tried.
I"m using MS SQL 2005
One example:
select id, cartype from cars c
where ( select count(1) from cities where id in (...))
= ( select count(distinct cityid)
from locations
where c.companyid = locations.id and cityid in (...) )
Maybe try counting all the cities, and then select the car if the company has the same number of distinct location cities are there are total cities.
SELECT id, cartype FROM cars
WHERE
--Subquery to find the number of locations belonging to car's company
(SELECT count(distinct cities.id) FROM cities
INNER JOIN locations on locations.cityid = cities.id
WHERE locations.companyId = cars.companyId)
=
--Subquery to find the total number of locations
(SELECT count(distinct cities.id) FROM cities)
I haven't tested this, and it may not be the most efficient query, but I think this might work.
Try this
SELECT e.*
FROM cars e
WHERE NOT EXISTS (
SELECT 1
FROM Cities p
WHERE p.location = e.Location
)
I have this query:
select pa.id,pa.name
from patients pa
where (select count(distinct co.doctorID)
from consults co
where co.patientID=pa.id) = (select count(do.id)
from doctors do);
In which I select the patients who had consults with all the doctors.
I am counting the number of distinct doctors and see if is equal to the total of doctors.
But how can I do this with an exists and/or in without using count
You can use a nested NOT EXISTS:
select pa.id,pa.name
from patients pa
where NOT EXISTS
(
select 1 FROM doctors do
where NOT EXISTS
(
SELECT 1 FROM consults co
WHERE co.doctorID=do.id
AND pa.id=co.patientID
)
)
In other words, select all patients without a doctor that was not consulted. At least it should give you the idea.
I think you can try even better solution with NOT EXISTS that will not require counts at all:
SELCT pa.id,pa.name
FROM patients pa
WHERE NOT EXISTS (SELECT *
FROM doctors do
LEFT JOIN consults co
ON co.doctorID ON do.id
WHERE co.patientID=pa.id
AND co.doctorID IS NULL
)
In sub-query we get all the doctors and see whether all of them have consult for current patient. If co.doctorID IS NULL, that doctor was not visited by patient.
I'm trying to determine a correct way to isolate rows within a table that have the same values in 2 columns.
There are two tables, one (Name) with the person's names and IDs, and the other one (Nation) with people's IDs and their nations. I join the two tables with inner join, and now the new table columns consist of an ID, first name, last name, and nation. If I want to find pairs of people who have the same last name and are from the same nation, why isn't
select ID, FName, LName, Nation
from (Name inner join Nation on Name.ID = Nation.ID)
group by Name, Nation
having count(Name) > 1 and count(Nation) > 1
working?
I'm aiming for the result to be a table with columns:
ID -------First--------------- Last ---------Nation
where the last names and nations will be identical pairs while first names will be different.
I feel like the group by part isnt appropriate, but is there even an alternate way? Thanks for any help.
If you are using MS SQL Server:
select
*
from
(
select
Name.*,
Nation.Nation,
cnt = count(*) over(partition by LName, Nation)
from Name
join Nation on Nation.ID = Name.ID
) t
where cnt > 1
Try this:
SELECT * FROM (
SELECT Name.ID, Name.FName, Name.LName, Nation.Nation
FROM Name
INNER JOIN Nation ON (Name.ID = Nation.ID)
) a
INNER JOIN (
SELECT Name.ID, Name.FName, Name.LName, Nation.Nation
FROM Name
INNER JOIN Nation ON (Name.ID = Nation.ID)
) b ON (a.LName = b.LName AND a.Nation = b.Nation)
WHERE a.ID < b.ID
As Simon Righarts hinted, something's not right with the design.
Scenario 1)
If a name can have multiple nations, you would have 3 tables implementing an n:m relationship.
CREATE TABLE name (name_id int, name text, ...);
CREATE TABLE nation (nation_id int, nation text, ...);
CREATE TABLE nationality (name_id int references name(name_id)
,nation_id int references nation(nation_id)
... );
Query for the scenario:
SELECT a.name_id, a.fname, a.lname, n.nation
FROM name a
JOIN nationality na USING (name_id)
JOIN nation n USING (nation_id)
JOIN (
SELECT a.lname, na.nation_id
FROM name a
JOIN nationality na USING (name_id)
GROUP BY 1,2
HAVING count(*) > 1) x USING (lname, nation_id)
Scenario 2)
If a name can only have one nation, there would be a column nation_id in the table name:
CREATE TABLE name (name_id int
,name text
,nation_id int references nation(nation_id), ...);
CREATE TABLE nation (nation_id int, nation text, ...);
Query for this scenario:
SELECT a.name_id, a.fname, a.lname, n.nation
FROM name a
JOIN nation n USING (nation_id)
JOIN (
SELECT a.lname, a.nation_id
FROM name a
GROUP BY 1,2
HAVING count(*) > 1) x USING (lname, nation_id);
All multiple occurrences are included here, not just "pairs" - assuming you meant that.
Your actual description doesn't fit either scenario.