Correct SQLite syntax - UPDATE SELECT with WHERE EXISTS

Question

I am trying to update a selected values in a column in a SQLite table. I only want update of the cells in the maintable where the criteria are met, and the cells must be updated to individual values, taken from a subtable.

I have tried the following syntax, but I get only a single cell update. I have also tried alternatives where all cells are updated to the first selected value of the subtable.

UPDATE maintable
SET value=(SELECT subtable.value FROM maintable, subtable
WHERE  maintable.key1=subtable.key1 AND maintable.key2=subtable.key2)
WHERE EXISTS (SELECT subtable.value FROM maintable, subtable
WHERE  maintable.key1=subtable.key1 AND maintable.key2=subtable.key2)

What is the appropriate syntax?

Answer 1

By default update with joins does not exist in SQLite; But we can use the with-clause + column-name-list + select-stmt from https://www.sqlite.org/lang_update.html to make something like this:

CREATE TABLE aa (
_id INTEGER PRIMARY KEY,
a1 INTEGER,
a2 INTEGER);

INSERT INTO aa  VALUES (1,10,20);
INSERT INTO aa  VALUES (2,-10,-20);
INSERT INTO aa  VALUES (3,0,0);

--a bit unpleasant because we have to select manually each column and it's just a lot to write
WITH bb (_id,b1, b2)  
AS  (SELECT _id,a1+2, a2+1 FROM aa WHERE _id<=2) 
UPDATE aa  SET a1=(SELECT b1 FROM bb WHERE bb._id=aa._id),a2=(SELECT b2 FROM bb WHERE bb._id=aa._id)
WHERE _id in (SELECT _id from bb);

--soo now it should be (1,10,20)->(1,12,21) and (2,-10,-20)->(2,-8,-19), and it is
SELECT * FROM aa;


--even better with one select for each row!
WITH bb (_id,b1, b2)  
AS  (SELECT _id,a1+2, a2+1 from aa WHERE _id<=2)
UPDATE aa  SET (a1,a2)=(SELECT b1,b2 FROM bb WHERE bb._id=aa._id)
WHERE _id in (SELECT _id from bb);

--soo now it should be (1,12,21)->(1,14,22) and (2,-8,-19)->(2,-6,-18), and it is
SELECT * FROM aa;


--you can skip the WITH altogether
UPDATE aa SET (a1,a2)=(SELECT bb.a1+2, bb.a2+1 FROM aa AS bb WHERE aa._id=bb._id)
WHERE _id<=2;

--soo now it should be (1,14,22)->(1,16,23) and (2,-6,-18)->(2,-4,-17), and it is
SELECT * FROM aa;

Hopefully sqlite is smart enough to not query incrementally but according to the documentation it is. When setting multiple columns using one select (case 2 and 3) a not valid id (no where _id in line) will give an error that can not be ignored using ON IGNORE, case 1 will set columns to null (for all ids >2) which is also bad.