SQL loop, condition queries - sql

Question

I have table 'db' which has these columns:
sname(student name)
cname (course name)
year ( the year when student take)
qtr ( the quarter when student take) : W , S, F (W>S>F)
grade
What I want to do is that list the students that gpa has increased every quarter (previous grade < current gpa). Students may have gaps between quarters.
SELECT a.sname
FROM db a
WHERE EXISTS (SELECT *
FROM db b
WHERE a.sname = b.sname
AND a.year = b.year
AND a.qtr < b.qtr
AND a.grade > b.grade)
UNION
After union, I will do the case when a.year > b.year.
This is what I am doing right now - is it correct way to do it this way?
I have no idea how I can loop in SQL queries...
Any advise will be appreciate. Thank you

Answer 1

If you're using the current version of sqlite (3.25.X), this is really easy to do using the lag() window function. In the below, I'm using numeric values for grades instead of letters because it's much easier to work with, and assuming the following table definition (Based on your description) and sample rows in it:
CREATE TABLE IF NOT EXISTS grades(
sname TEXT NOT NULL
, year INTEGER NOT NULL
, qtr INTEGER NOT NULL
, cname TEXT NOT NULL
, grade NUMERIC
, PRIMARY KEY (sname, year, qtr, cname)
) WITHOUT ROWID;
INSERT INTO grades(sname, cname, year, qtr, grade) VALUES
('Bob', 'Math', 2017, 3, 2.0), ('Bob', 'Math', 2017, 4, 2.5),
('Bob', 'Math', 2018, 1, 3.0), ('Amy', 'Math', 2017, 3, 4.0),
('Amy', 'Math', 2017, 4, 3.5), ('Amy', 'Math', 2018, 1, 4.0),
('Bob', 'History', 2017, 3, 3.5), ('Bob', 'History', 2017, 4, 3.0),
('Bob', 'History', 2018, 1, 3.5), ('Amy', 'History', 2017, 3, 2.5),
('Amy', 'History', 2017, 4, 3.5), ('Amy', 'History', 2018, 1, 4.0);
I assume that when you said previous grade < current gpa you meant previous gpa to match with gpa has increased every quartely[sic].
First, a query that calculates each student's GPA for each quarter and includes the previous GPA as well in each row (Thanks to lag()):
SELECT sname, year, qtr
, avg(grade) AS gpa
, lag(avg(grade), 1, 0.0)
OVER (PARTITION BY sname ORDER BY year, qtr) AS prev_gpa
FROM grades
GROUP BY sname, year, qtr
ORDER BY sname, year, qtr;
This produces:
sname year qtr gpa prev_gpa
---------- ---------- ---------- ---------- ----------
Amy 2017 3 3.25 0.0
Amy 2017 4 3.5 3.25
Amy 2018 1 4.0 3.5
Bob 2017 3 2.75 0.0
Bob 2017 4 2.75 2.75
Bob 2018 1 3.25 2.75
As you can see, with this sample data, Amy has an always-increasing GPA, while Bob doesn't. So, the issue is how to filter the results to just her? The answer to that lies in using HAVING with a GROUP BY. It's complicated a bit because values computed by window functions can only appear in the select column list and ORDER BY clauses, so I shove the above query into a CTE to work around that restriction:
WITH gpas AS (
SELECT sname, year, qtr
, avg(grade) AS gpa
, lag(avg(grade), 1, 0.0)
OVER (PARTITION BY sname ORDER BY year, qtr) AS prev_gpa
FROM grades
GROUP BY sname, year, qtr)
SELECT sname
FROM gpas
GROUP BY sname
HAVING sum(CASE WHEN gpa > prev_gpa THEN 1 ELSE 0 END) = count(gpa)
ORDER BY sname;
which produces
sname
----------
Amy
The GROUP BY sname HAVING ... part filters out the students where at least one of their rows doesn't have an increased GPA compared to their previous one. It's worth spending some time reading up on grouping, as it's probably the most difficult basic concept to grasp and also very useful and powerful.