PostgreSQL – How to Show All Variations of 2 Characters

If you want to happen it automatically (I mean, without the application doing that), you have to create a sequence:

CREATE SEQUENCE some_table_pk_seq START 50;

Then change the column to use this as a default:

ALTER TABLE some_table
ALTER COLUMN id SET DEFAULT nextval('some_table_pk_seq'::regclass);

And that's it. Remember that the default will be used only if you don't pass a value to the given column in your INSERT statements.

WITH parametros AS (
   SELECT '2015-03-06'::timestamp AS fecha_desde  -- provide parameters here
        , '2015-03-12'::timestamp AS fecha_hasta
   )
SELECT tropa, max(fecha) AS fecha, sum(cantidad) AS sum_cantidad
     , sum(sum(cantidad)) OVER (PARTITION BY tropa
                                ORDER BY max(fecha)) AS saldo
FROM   movimiento_stock, parametros p
WHERE  fecha <= p.fecha_hasta
GROUP  BY tropa, CASE WHEN fecha >= p.fecha_desde THEN id END
ORDER  BY 1, 2;

The core feature is bold. Read the manual about CASE.
Older rows get NULL for id (the default in a CASE expression); id is NOT NULL in the underlying table, so collisions are not possible.

This groups all rows older than fecha_desde into one group per tropa and leaves newer rows ungrouped (form individual "groups"). It's unclear how to aggregate other columns, so I only included fecha.

The same, wrapped into an SQL function:

CREATE OR REPLACE FUNCTION informe_tldr(fecha_desde timestamp
                                      , fecha_hasta timestamp)
  RETURNS TABLE(tropa int, fecha timestamp, sum_cantidad int, saldo int) AS
$func$
SELECT m.tropa, max(m.fecha), sum(m.cantidad)::int
     , sum(sum(m.cantidad)) OVER (PARTITION BY m.tropa
                                  ORDER BY max(m.fecha))::int AS saldo
FROM   movimiento_stock m
WHERE  m.fecha <= fecha_hasta
GROUP  BY m.tropa, CASE WHEN fecha >= fecha_desde THEN m.id END
ORDER  BY 1, 2;
$func$ LANGUAGE sql;

SQL Fiddle.