How create a column that increase according to the value of another column - postgresql

Question

I have a table where I want to put all the information about articles use, and I need to create a column with autoincrement, where the ID can have the same value if the field (tipo) have another value, unique for this particular ID. For example:
ID / TIPO
1 / AJE -- Is Ok
1 / AJS -- Is Ok (because this Tipo is AJS, different from AJE)
1 / SI -- Is Ok
2 / AJE -- Is Ok (New ID)
2 / AJE -- Is Wrong, because ID=2, TIPO=AJE already exist.
I've done the unique sentence:
ALTER TABLE public.art_movimientos
ADD CONSTRAINT uk_mov UNIQUE (id,tipo) USING INDEX TABLESPACE sistema_index;
But how I can create the autoincrement covering two columns?
My table Code:
CREATE TABLE public.art_movimientos
(
id integer NOT NULL DEFAULT nextval('seq_art_movimientos'::regclass),
tipo character(3) NOT NULL, -- Tipos de Valores:...
documento integer,
fecha_doc date[] NOT NULL,
fecha_mov date[] NOT NULL,
calmacen integer NOT NULL,
status character(13) NOT NULL DEFAULT 'PENDIENTE'::bpchar, -- PENDIENTE...
mes integer NOT NULL,
"año" integer NOT NULL,
donado integer NOT NULL DEFAULT 0
)
WITH (
OIDS=FALSE
);

Answer 1

You can manage this situation by using an before insert trigger, mimicking the behaviour statedby #dhke:
CREATE TABLE art_movimientos
(
id integer NOT NULL DEFAULT NULL, -- You don't want a serial, nor a default
tipo character(3) NOT NULL, -- Tipos de Valores:...
documento integer,
fecha_doc date[] NOT NULL,
fecha_mov date[] NOT NULL,
calmacen integer NOT NULL,
status character(13) NOT NULL DEFAULT 'PENDIENTE'::bpchar, -- PENDIENTE...
mes integer NOT NULL,
"año" integer NOT NULL,
donado integer NOT NULL DEFAULT 0,
/* You have actually a 2-column Primary Key */
PRIMARY KEY (tipo, id)
);
-- Create a trigger function to generate 'id'
CREATE FUNCTION art_movimientos_insert_trigger()
RETURNS trigger
AS
$$
BEGIN
/* Compute "id", as the following id for a certain "tipo" */
new.id = coalesce(
(SELECT max(id) + 1
FROM art_movimientos a
WHERE a.tipo = new.tipo), 1);
return new;
END
$$
LANGUAGE 'plpgsql'
VOLATILE ;
-- This trigger will be called whenever a new row is inserted, and "id" is
-- not specified (i.e.: it defaults to null), or is specified as null
CREATE TRIGGER art_movimientos_ins_trg
BEFORE INSERT
ON art_movimientos
FOR EACH ROW
WHEN (new.id IS NULL)
EXECUTE PROCEDURE art_movimientos_insert_trigger();
You can then insert the following rows (without specifying the id column):
INSERT INTO art_movimientos
(tipo, documento, fecha_doc, fecha_mov, calmacen, mes, "año")
VALUES
('AJE', 1, array['20170128'::date], array['20170128'::date], 1, 1, 2017),
('AJS', 2, array['20170128'::date], array['20170128'::date], 1, 1, 2017),
('SI', 3, array['20170128'::date], array['20170128'::date], 1, 1, 2017),
('AJE', 4, array['20170128'::date], array['20170128'::date], 1, 1, 2017),
('AJE', 5, array['20170128'::date], array['20170128'::date], 1, 1, 2017) ;
... and see that you get what you intended:
SELECT
id, tipo
FROM
art_movimientos
ORDER BY
documento ;
| id | tipo |
|----|------|
| 1 | AJE |
| 1 | AJS |
| 1 | SI |
| 2 | AJE |
| 3 | AJE |
You can check everything a SQLFiddle (which is a bit picky about PL/pgSQL functions and semicolons).
Side note: There can be a few corner cases where this procedure might fail within a transaction because of deadlocks and/or race conditions, if other transactions are also inserting data into the same table at the same time. So, your overall code should be able to handle aborted transactions, and retry them or show an error to the user.