PostgreSQL as NoSQL

This post is continuation of the series. If you do not have Postgres on your machine or sample data, you can read Installing PostgreSQL and Loading Sample Data post first.

There are two JSON (JavaScript Object Notation) types for storing JSON data.

The json data type stores an exact copy of the input text, which processing functions must reparse on each execution; while jsonb data is stored in a decomposed binary format that makes it slightly slower to input due to added conversion overhead, but significantly faster to process, since no reparsing is needed.

Create a table with a column with jsonb data type.

create table film_docs(data jsonb);

Insert all the data in the row to that column. The original table comes from the sample data.

insert into film_docs(data)
select row_to_json(film)::jsonb
from film;

select (data ->> 'title') as Title,
(data -> 'length') as Length
from film_docs
where data -> 'title' ? 'Chamber Italian';

There are lots of JSON functions and operators. I think Entity Framework does not have these functions. I will search for this next time.

explain select (data ->> 'title') as Title,
(data -> 'length') as Length
from film_docs
where data @> '{"title": "Chamber Italian"}';

jsonb also supports indexing, which can be a significant advantage. Here we have an explanation of the above query without index.

Seq Scan on film_docs  (cost=0.00..96.50 rows=1 width=64)
  Filter: (data @> '{"title": "Chamber Italian"}'::jsonb)
Time: 0.009s

If we create a GIN index and run explain statement again, it would give us less cost. jsonb_path_ops means generated index will be specific for @> operator.

create index on film_docs using GIN(data jsonb_path_ops)

Bitmap Heap Scan on film_docs (cost=12.01..16.03 rows=1 width=64) Recheck Cond: (data @> ‘{“title”: “Chamber Italian”}’::jsonb) -> Bitmap Index Scan on film_docs_data_idx (cost=0.00..12.01 rows=1 width=0) Index Cond: (data @> ‘{“title”: “Chamber Italian”}’::jsonb) Time: 0.008s

If you want to continue with more advanced topics on PostgreSQL, search for table inheritance.