我在使用下表的Postgres 9.4中工作:
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 │
我有以下查询:
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;
它为我提供了带有JSON对象的结果star_pu:
star_pu
date │ 2013-04-01 total_list_size │ 56025123.000000000000 star_pu │ { "antidepressants_cost" : 180102416.8036909901975399, "antiepileptic_drugs_cost" : 296228344.171576079922216... }
相反,我想将JSON结果展平为一系列命名空间的键,因此结果如下所示:
date │ 2013-04-01 total_list_size │ 56025123.000000000000 star_pu.antidepressants_cost │ 180102416.8036909901975399 star_pu.antiepileptic_drugs_cost │ 296228344.171576079922216 ...
这可能吗?
下面的函数基于表动态创建视图:
create or replace function create_totals_view(table_name text) returns void language plpgsql as $$ declare s text; begin execute format ($fmt$ select string_agg(format('star_pu->>''%s'' "%s"', key, key), ',') from ( select distinct key from %s, json_each(star_pu) order by 1 ) s; $fmt$, '%s', '%s', table_name) into s; execute format(' drop view if exists %s_view; create view %s_view as select date, total_list_size, %s from %s', table_name, table_name, s, table_name); end $$;
首先,从查询中创建一个表。
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;
接下来,使用函数,该函数将创建一个以_viewpostfix 表命名的视图:
_view
select create_totals_view('totals');
最后,查询视图:
select * from totals_view;
create or replace function create_jsonb_flat_view (table_name text, regular_columns text, json_column text) returns text language plpgsql as $$ declare cols text; begin execute format ($ex$ select string_agg(format('%2$s->>%%1$L "%%1$s"', key), ', ') from ( select distinct key from %1$s, jsonb_each(%2$s) order by 1 ) s; $ex$, table_name, json_column) into cols; execute format($ex$ drop view if exists %1$s_view; create view %1$s_view as select %2$s, %3$s from %1$s $ex$, table_name, regular_columns, cols); return cols; end $$;
用法:
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)