SELECT statement for all 3 tables in a many-to-many relationship - postgresql

Question

I am having trouble writing a SELECT query that includes all the 3 tables in a many-to-many relationship. I have the following tables:
Table "public.companies"
Column | Type | Modifiers | Storage | Stats target | Description
----------------+------------------------+--------------------------------------------------------+----------+--------------+-------------
id | integer | not null default nextval('companies_id_seq'::regclass) | plain | |
name | character varying(48) | not null | extended | |
description | character varying(512) | | extended | |
tagline | character varying(64) | | extended | |
featured_image | integer | | plain | |
Indexes:
"companies_pkey" PRIMARY KEY, btree (id)
Referenced by:
TABLE "company_category_associations" CONSTRAINT "company_category_associations_company_id_foreign" FOREIGN KEY (company_id) REFERENCES companies(id) ON DELETE CASCADE
Table "public.company_category_associations"
Column | Type | Modifiers
-------------+---------+----------------------------------------------------------------------------
id | integer | not null default nextval('company_category_associations_id_seq'::regclass)
company_id | integer | not null
category_id | integer | not null
Indexes:
"company_category_associations_pkey" PRIMARY KEY, btree (id)
Foreign-key constraints:
"company_category_associations_category_id_foreign" FOREIGN KEY (category_id) REFERENCES company_categories(id) ON DELETE RESTRICT
"company_category_associations_company_id_foreign" FOREIGN KEY (company_id) REFERENCES companies(id) ON DELETE CASCADE
Table "public.company_categories"
Column | Type | Modifiers
-------------+-----------------------+-----------------------------------------------------------------
id | integer | not null default nextval('company_categories_id_seq'::regclass)
name | character varying(32) | not null
description | character varying(96) |
Indexes:
"company_categories_pkey" PRIMARY KEY, btree (id)
Referenced by:
TABLE "company_category_associations" CONSTRAINT "company_category_associations_category_id_foreign" FOREIGN KEY (category_id) REFERENCES company_categories(id) ON DELETE RESTRICT
My companies table will have around 100k rows and a company can have up to 10 categories associated. Of course I won't be selecting more than 200 companies at a time.
I managed to get the results with the following query:
select
c.id as companyid,
c.name as companyname,
cat.id as categoryid,
cat.name as categoryname
from company_categories cat
left join company_category_associations catassoc on catassoc.category_id = cat.id
left join companies c on catassoc.company_id = c.id where c.id is not null;
This question comes from the fact that I need to present data in JSON format and I would like it to look like this:
{
"companies": [
{
"name": "...",
"description": "...",
"categories": [
{
"id": 12,
"name": "Technology"
},
{
"id": 14,
"name": "Computers"
},
]
},
/* ... */
]
}
And basically I want to take as much of that data in as few queries as possible.
How can I write that SELECT query to fit my needs?
Is there a problem with the database structure as it is my diagram?
Thank you!
P.S. I am using PostgreSQL 9.6.6

Answer 1

You can get json out of postgresql directly.
First define composite types (json objects) that you want to output. This is to get names for fields, otherwise they will be named f1,f2,...
create type cat as (id integer, name varchar);
create type comp as (id integer, name varchar, categories cat[]);
Then select array of comps with nested array of cats as json
select to_json(array(
select (
c.id,
c.name,
array(
select (cc.id, cc.name)::cat
from company_categories cc
join company_category_associations cca on (cca.category_id=cc.id and cca.company_id=c.id)
))::comp
from companies c
)) as companies
dbfiddle

Answer 2

You can nest two JSON aggregations to get there.
The first level creates a JSON value for each company with the categories stored as an array:
select to_jsonb(c) || jsonb_build_object('categories', jsonb_agg(cc)) comp_json
from companies c
join company_category_associations cca on cca.company_id = c.id
join company_categories cc on cc.id = cca.category_id
group by c.id;
This returns one row per company. Now we need to aggregate those rows into a single JSON value:
select jsonb_build_object('companies', jsonb_agg(comp_json))
from (
select to_jsonb(c) || jsonb_build_object('{categories}', jsonb_agg(cc)) comp_json
from companies c
join company_category_associations cca on cca.company_id = c.id
join company_categories cc on cc.id = cca.category_id
group by c.id
) t;
Online example: http://rextester.com/TRCR26633