with - postgresql jsonb example

¿Acoplar pares clave/valor agregados de un campo JSONB?

Question

Estoy trabajando en Postgres 9.4 con la siguiente tabla:

     Column      │         Type         │ Modifiers
─────────────────┼──────────────────────┼──────────────────────
 id              │ integer              │ not null default
 practice_id     │ character varying(6) │ not null
 date            │ date                 │ not null
 pct_id          │ character varying(3) │
 total_list_size │ double precision     │ not null
 star_pu         │ jsonb                │

Tengo la siguiente consulta:

SELECT date,
       AVG(total_list_size) AS total_list_size,
       json_object_agg(key, val) AS star_pu
FROM (SELECT date,
             SUM(total_list_size) AS total_list_size,
             key, SUM(value::numeric) val FROM frontend_practicelist p,
             jsonb_each_text(star_pu)
       GROUP BY date, key ) p
GROUP BY date
ORDER BY date;

Me da resultados con un objeto JSON adjunto a star_pu :

date            │ 2013-04-01
total_list_size │ 56025123.000000000000
star_pu         │ { "antidepressants_cost" : 180102416.8036909901975399, "antiepileptic_drugs_cost" : 296228344.171576079922216... }

En cambio, me gustaría aplanar el resultado JSON a una serie de claves de espacios de nombres, por lo que el resultado se ve así:

date            │ 2013-04-01
total_list_size │ 56025123.000000000000
star_pu.antidepressants_cost        │ 180102416.8036909901975399
star_pu.antiepileptic_drugs_cost    │ 296228344.171576079922216 
...

es posible?

Answer 1

Este caso particular

La siguiente función crea dinámicamente una vista basada en una tabla:

create or replace function create_totals_view(table_name text)
returns void language plpgsql as $$
declare
    s text;
begin
    execute format ($fmt$
        select string_agg(format(''star_pu->>''''%s'''' "%s"'', key, key), '','')
        from (
            select distinct key
            from %s, json_each(star_pu)
            order by 1
            ) s;
        $fmt$, ''%s'', ''%s'', table_name)
    into s;
    execute format(''
        drop view if exists %s_view;
        create view %s_view as 
        select date, total_list_size, %s from %s'', 
        table_name, table_name, s, table_name);
end $$;

Primero, cree una tabla a partir de su consulta.

create table totals as

    SELECT date,
           AVG(total_list_size) AS total_list_size,
           json_object_agg(key, val) AS star_pu
    FROM (SELECT date,
                 SUM(total_list_size) AS total_list_size,
                 key, SUM(value::numeric) val FROM frontend_practicelist p,
                 jsonb_each_text(star_pu)
           GROUP BY date, key ) p
    GROUP BY date
    ORDER BY date;

Luego, use la función, que creará una vista con el nombre de la tabla con _view postfix:

select create_totals_view(''totals'');

Finalmente, consulta la vista:

select * from totals_view;

Solución generalizada (para jsonb)

create or replace function create_jsonb_flat_view
    (table_name text, regular_columns text, json_column text)
    returns text language plpgsql as $$
declare
    cols text;
begin
    execute format ($ex$
        select string_agg(format(''%2$s->>%%1$L "%%1$s"'', key), '', '')
        from (
            select distinct key
            from %1$s, jsonb_each(%2$s)
            order by 1
            ) s;
        $ex$, table_name, json_column)
    into cols;
    execute format($ex$
        drop view if exists %1$s_view;
        create view %1$s_view as 
        select %2$s, %3$s from %1$s
        $ex$, table_name, regular_columns, cols);
    return cols;
end $$;

Uso:

create table example (id int, name text, params jsonb);
insert into example values
(1, ''Anna'', ''{"height": 175, "weight": 55}''),
(2, ''Bob'', ''{"age": 22, "height": 188}''),
(3, ''Cindy'', ''{"age": 25, "weight": 48, "pretty": true}'');

select create_jsonb_flat_view(''example'', ''id, name'', ''params'');

select * from example_view;

 id | name  | age | height | pretty | weight 
----+-------+-----+--------+--------+--------
  1 | Anna  |     | 175    |        | 55
  2 | Bob   | 22  | 188    |        | 
  3 | Cindy | 25  |        | true   | 48
(3 rows)