0Pricing
PostgreSQL Performance & Query Optimization · Aula

GIN versus índices de expressão em JSONB

Escolha entre índices GIN jsonb_path_ops e índices de expressão direcionados para os formatos das suas consultas.

GIN versus índices de expressão em JSONB é uma aula grátis de PostgreSQL Performance & Query Optimization no CoddyKit. Esta é a aula 2 de 4. Você pode ler a aula completa abaixo gratuitamente — depois pratica ao vivo no navegador com um editor de código integrado e um tutor de IA 24/7. Faz parte do caminho de aprendizado de PostgreSQL Performance & Query Optimization, e seu progresso é sincronizado entre a web e o app CoddyKit. O curso de PostgreSQL Performance & Query Optimization inclui 4 aulas no total.

Partes desta aula ainda não foram traduzidas e aparecem em inglês.

Two Ways to Index JSONB

When you store data in a jsonb column, an unindexed query forces PostgreSQL to read and parse every row. There are two very different tools to fix this:

  • GIN index — a general inverted index over the whole document, great for flexible containment and key/value lookups.
  • Expression (B-tree) index — a targeted index on one extracted scalar, great for a specific known query shape.

This lesson is about choosing the right one for your query patterns.

The Sample Table

Imagine an events table where each row carries a flexible JSON payload. We will index its data column.

Notice the payload mixes a few common keys (type, user_id) with arbitrary extras.

CREATE TABLE events (
  id     bigserial PRIMARY KEY,
  data   jsonb NOT NULL
);

INSERT INTO events (data) VALUES
  ('{"type": "login",  "user_id": 42, "ip": "10.0.0.1"}'),
  ('{"type": "logout", "user_id": 42}'),
  ('{"type": "login",  "user_id": 99, "mfa": true}');

The Default GIN: jsonb_ops

A plain GIN index uses the default jsonb_ops operator class. It indexes every key AND every value as separate entries.

This supports the widest set of operators: containment @>, key existence ?, ?|, and ?&.

The cost: it is larger on disk and slower to build/update because it stores far more entries per row.

CREATE INDEX idx_events_data_gin
  ON events USING gin (data);

-- Supports key existence AND containment:
-- WHERE data ? 'mfa'
-- WHERE data @> '{"type":"login"}'

The Leaner GIN: jsonb_path_ops

If you only ever use the containment operator @> (and the JSONPath operators @? / @@), prefer the jsonb_path_ops operator class.

  • It hashes whole key→value paths into single entries.
  • Result: noticeably smaller index and faster containment lookups.
  • Trade-off: it does NOT support the key-existence operators ?, ?|, ?&.
CREATE INDEX idx_events_data_pathops
  ON events USING gin (data jsonb_path_ops);

-- Great for:
SELECT id FROM events
WHERE data @> '{"type": "login"}';

How Containment Uses the GIN Index

The @> operator asks "does the left document contain the right one?" Both GIN operator classes accelerate it.

Containment is structural: it matches nested keys and values, not just top-level ones. This is why a single GIN index can serve many different filter combinations.

-- Match by one key:
SELECT * FROM events WHERE data @> '{"user_id": 42}';

-- Match by two keys at once (same index):
SELECT * FROM events
WHERE data @> '{"type": "login", "user_id": 42}';

-- Match a nested shape:
SELECT * FROM events WHERE data @> '{"flags": {"beta": true}}';

When GIN Falls Short: Range & Sort

GIN is built for equality-style containment. It canNOT help with:

  • Range comparisons on an extracted value (>, <, BETWEEN).
  • Ordering by a JSON field (ORDER BY ... LIMIT).
  • Prefix / pattern matching on a text value.

For these shapes you want a B-tree, and on JSONB that means an expression index.

-- GIN can't accelerate this range filter on an inner number:
SELECT * FROM events
WHERE (data ->> 'user_id')::int > 50
ORDER BY (data ->> 'user_id')::int
LIMIT 10;

Building an Expression Index

An expression index stores the result of an expression, not the raw column. You extract one scalar from the JSON and index that as a normal B-tree.

Two operators matter here:

  • -> returns jsonb.
  • ->> returns text — usually what you cast and index.
-- B-tree on user_id extracted as an integer:
CREATE INDEX idx_events_user_id
  ON events (((data ->> 'user_id')::int));

-- Now ranges, sorts and equality all use it:
SELECT * FROM events
WHERE (data ->> 'user_id')::int BETWEEN 40 AND 99
ORDER BY (data ->> 'user_id')::int;

Match the Index Expression Exactly

The planner only uses an expression index when the query expression matches the indexed expression token for token, including the cast.

If you index (data ->> 'user_id')::int but query (data ->> 'user_id') as plain text, the index is ignored.

Keep the extraction + cast identical everywhere.

-- Indexed expression:
--   ((data ->> 'user_id')::int)

-- USES the index:
WHERE (data ->> 'user_id')::int = 42

-- IGNORES the index (text vs int mismatch):
WHERE (data ->> 'user_id') = '42'

Reading EXPLAIN to Confirm

Never guess which index wins — ask the planner. Use EXPLAIN (ANALYZE, BUFFERS) and look at the node type:

  • Bitmap Heap Scan + Bitmap Index Scan on ...gin → your GIN index is serving containment.
  • Index Scan / Index Only Scan on the expression index → your B-tree is serving the range/sort.
  • Seq Scan → nothing matched; revisit the expression or operator.
EXPLAIN (ANALYZE, BUFFERS)
SELECT * FROM events
WHERE data @> '{"type": "login"}';

EXPLAIN (ANALYZE, BUFFERS)
SELECT * FROM events
WHERE (data ->> 'user_id')::int = 42;

Partial Expression Indexes

If queries only ever target a subset of rows, add a WHERE clause to the index. A partial expression index is smaller and cheaper to maintain because it only stores the rows you actually search.

Here we index user_id only for login events — perfect when that is the only query shape that needs it.

CREATE INDEX idx_events_login_user
  ON events (((data ->> 'user_id')::int))
  WHERE data @> '{"type": "login"}';

Choosing: A Quick Decision Guide

Pick by the shape of your queries, not by habit:

  • Flexible filters on many different keys, or key-existence (?) → GIN jsonb_ops.
  • Only containment @> / JSONPath, want it lean and fast → GIN jsonb_path_ops.
  • One known field with ranges, sorting, or equality on a scalar → expression B-tree index.
  • That field queried on a narrow slice of rows → partial expression index.

It is common and correct to keep BOTH a GIN and one or two expression indexes on the same column.

Quick Check

Test your understanding of the GIN vs expression decision.

Recap

You learned to choose JSONB indexes by query shape:

  • GIN jsonb_ops — widest operator support including key existence ?; largest.
  • GIN jsonb_path_ops — leaner and faster, containment @> and JSONPath only.
  • Expression B-tree — one extracted, casted scalar for ranges, sorts, and equality; the query expression must match the index expression exactly.
  • Partial expression index — same idea, scoped to a row subset for a smaller footprint.

Always confirm with EXPLAIN (ANALYZE, BUFFERS), and don't hesitate to keep a GIN and one or two expression indexes side by side.

Perguntas Frequentes

A aula “GIN versus índices de expressão em JSONB” é grátis?

Sim — o texto completo de “GIN versus índices de expressão em JSONB” é grátis para ler aqui na web. Para praticá-la interativamente (um editor de código integrado e um tutor de IA 24/7) e desbloquear o restante do curso de PostgreSQL Performance & Query Optimization, atualize para CoddyKit PRO. O curso de PostgreSQL Performance & Query Optimization inclui 4 aulas no total.

O que vou aprender em “GIN versus índices de expressão em JSONB”?

Escolha entre índices GIN jsonb_path_ops e índices de expressão direcionados para os formatos das suas consultas. Você pratica PostgreSQL Performance & Query Optimization com código prático que executa diretamente no navegador, e um tutor de IA 24/7 responde suas dúvidas enquanto trabalha na aula.

Preciso ter experiência prévia para começar PostgreSQL Performance & Query Optimization?

Nenhuma experiência prévia é necessária. PostgreSQL Performance & Query Optimization no CoddyKit é estruturado para alunos iniciantes até avançados, então você pode começar aqui ou desde o início e aprender no seu ritmo. Esta é a aula 2 de 4.

Quanto tempo leva a aula “GIN versus índices de expressão em JSONB”?

A maioria das aulas CoddyKit leva cerca de 5–10 minutos. Cada uma é compacta e interativa, então você faz progresso constante e retoma exatamente de onde parou entre web e app.

Posso escrever e executar código nesta aula de PostgreSQL Performance & Query Optimization?

Sim. Cada aula de PostgreSQL Performance & Query Optimization inclui um editor de código integrado, então você escreve e executa código real direto no navegador e recebe feedback de IA instantaneamente — nenhuma configuração local necessária.

Todas as aulas deste curso

  1. Operadores JSONB e consultas de contenção
  2. GIN versus índices de expressão em JSONB
  3. Consulta de JSONB com JSONPath
  4. Quando normalizar dados fora do JSONB
← Voltar para PostgreSQL Performance & Query Optimization