What is the most efficient way of joining tables of different dimensions? - sql

Question

I have the following schema:
CREATE TABLE products (
id BIGSERIAL NOT NULL,
created_at_timestamp TIMESTAMP NOT NULL DEFAULT NOW(),
last_update_timestamp TIMESTAMP NOT NULL DEFAULT NOW(),
PRIMARY KEY (id)
);
CREATE TABLE product_names (
product_id BIGINT NOT NULL,
language TEXT NOT NULL,
name TEXT NOT NULL,
PRIMARY KEY (product_id, language),
FOREIGN KEY (product_id) REFERENCES products (id)
);
CREATE TABLE product_summaries (
product_id BIGINT NOT NULL,
language TEXT NOT NULL,
summary TEXT NOT NULL,
PRIMARY KEY (product_id, language),
FOREIGN KEY (product_id) REFERENCES products (id)
);
And I want to select all Products.
However as you can see a Product contains a list of names and summaries (per language).
I can retrieve all Products
SELECT * FROM products
And then iterate all the rows (in this case in Kotlin), and then request the names and summaries:
SELECT * FROM product_names WHERE product_id = $id
And
SELECT * FROM product_summaries WHERE product_id = $id
However, this seems inefficient, since I am making 3 separate queries to the database.
I though of using JOINs to get all of this with one query, but then I get multiple repeated rows for each product_names and product_summaries entry.
So in the end, is there a better way of requesting all this data in one query?

Answer 1

You definitely don't want to do multiple queries and then iterate over them in the code. That's horribly inefficient. When you do the second JOIN, you need to include language in the JOIN. That should keep you from getting duplicate rows. This should give you one row for each unique combination of [products.id, product_names.language]
SELECT
products.id
,products.created_at_timestamp
,products.last_update_timestamp
,product_names.name
,product_summaries.summary
,product_names.language
FROM
products
INNER JOIN
product_names ON product_names.product_id = products.id
INNER JOIN
product_summaries ON product_summaries.product_id = products.id
AND product_summaries.language = product_names.language

Answer 2

I've found a way of doing it:
SELECT * FROM products as p INNER JOIN
(SELECT json_agg(product_names) as names, product_id FROM product_names GROUP BY product_id) as tb_names ON tb_names.product_id = p.id
INNER JOIN
(SELECT json_agg(product_summaries) as summaries, product_id FROM product_summaries GROUP BY product_id) as tb_summaries ON tb_summaries.product_id = p.id
returns:
1 | 2018-07-20 09:36:21.56904 | 2018-07-20 09:36:21.56904 | [{"product_id":1,"language":"EN","name":"lol"},
{"product_id":1,"language":"DE","name":"lel"}] | 1 [{"product_id":1,"language":"EN","summary":"deded"},
{"product_id":1,"language":"DE","summary":"rererere"},
{"product_id":1,"language":"FR","summary":"jejejeje"}] | 1
Basically I'm converting the multi-dimensional tables to JSON :)
Postgres is amazing!