sqlite3 generate running total with repeated field value - sqlite

Question

This has been driving me nuts because I feel like it should be possible. But I'm admittedly not a huge SQL whiz.
I have an sqlite3 table that looks essentially like this:
id date amount
-- ---- ------
51 2018.10.01 10.0
52 2018.11.15 100.0
53 2018.11.15 20.0
54 2018.09.10 -30.0
(At least, these are the pertinent fields; the others have been left out).
What I want to do is generate a running total of the amount column, but with the data sorted by date.
I'm aware of the 'join the table with itself' trick for calculating a running total. So if I wanted a new running total value for each id (which is a unique field), I can do this:
select T2.id, T2.date, T2.amount, sum(T2.amount)
from Transactions T1
inner join Transactions T2
on T1.id >= T2.id
group by T1.id
And I get this:
"51" "2018.10.01" "10.0" "10.0"
"52" "2018.11.15" "100.0" "110.0"
"53" "2018.11.15" "20.0" "130.0"
"54" "2018.09.10" "-30.0" "100.0"
Running total correct.
But if I want a running total on this data in date order, it breaks down. This is close:
select T1.id, T2.date, T2.amount, sum(T2.amount)
from Transactions T1
inner join Transactions T2
on T1.date >= T2.date
group by T1.date
Except that it over-counts (and combines) the amount values in the two rows where date is 2018.11.15. Presumably because the on T1.date >= T2.date clause applies to both rows twice each.
"54" "2018.09.10" "-30.0" "-30.0"
"51" "2018.09.10" "-30.0" "-20.0"
"53" "2018.09.10" "-30.0" "200.0"
As I see it, this technique will only work if the join is performed on a field whose values are both unique and sorted. Once I sort the table by date, the unique id values are out of order and no longer usable.
So then I thought -- maybe sort the table by date first, then add a temporary column of unique sorted numbers. Simply the row number would do.
Unfortunately, this appears to be a version of sqlite that does not support any of row_number(), rownum or the over clause.
I'm aware of this technique for generating row numbers:
select id, date,
(select count(*) from Transactions T1 where T1.id <= T2.id)
from Transactions T2
"51" "2018.10.01" "1"
"52" "2018.11.15" "2"
"53" "2018.11.15" "3"
"54" "2018.09.10" "4"
But in no amount of fiddling around have I been able to figure out a way to:
First sort the table by date
Then use the count(*) technique to generate the unique row numbers
Then join the table with itself to create the running total
in a single SQL statement.
Hope this makes sense. Thanks for any thoughts anyone might have.

Answer 1

If you're using Sqlite 3.25 or better, window functions make this easy. Example:
First, populate a table with your sample data:
CREATE TABLE example(id INTEGER PRIMARY KEY, date TEXT, amount REAL);
INSERT INTO example VALUES(51,'2018-10-01',10.0);
INSERT INTO example VALUES(52,'2018-11-15',100.0);
INSERT INTO example VALUES(53,'2018-11-15',20.0);
INSERT INTO example VALUES(54,'2018-09-10',-30.0);
(Note that I changed the date format to one that the sqlite date and time functions understand, as that makes life easier as soon as you want to do something more complicated than sorting them).
The query
SELECT *, sum(amount) OVER (ORDER BY date, id) AS running_total
FROM example
ORDER BY date, id;
produces:
id date amount running_total
---------- ---------- ---------- -------------
54 2018-09-10 -30.0 -30.0
51 2018-10-01 10.0 -20.0
52 2018-11-15 100.0 80.0
53 2018-11-15 20.0 100.0
If you're using an older version, you really should consider upgrading for more reasons than just having window functions.

Answer 2

Thank you Shawn -- you put me on the track to an answer.
It looks like the most recent beta version of DB Browser for SQLite does support window functions (I presume because the latest version of SQLite itself does).
Problem solved!