Selective Consumer

Filtering on the queue side rather than in the consumer code — to avoid pulling 1000 messages just to drop 999.

Problem

A shared queue carries messages destined to several consumers per a criterion (partner, type, environment). A naive consumer pulls everything and filters in code — wasted bandwidth and CPU.

Forces

• A single queue is often imposed by a shared EDI broker; per-consumer queues are not always possible.
• Pulling then dropping consumes the message lock, blocks other consumers, and inflates the commit log.
• Queue-side filtering applies only to headers / properties, not to encrypted payload.
• An ill-fitted selector can leave a message stuck if no consumer matches.

Solution

Configure the consumer with a broker-side selector. Under JMS, the selector is a SQL-like expression on headers (`AS2_PARTNER = 'WALMART' AND DOC_TYPE = 'ORDERS'`). Under Kafka, rely on key-based partitioning plus a Streams filter. Under RabbitMQ, use a binding key on a topic exchange. The result: each consumer receives only its messages, no pull-and-drop.

EDI implementation

In EDI, Selective Consumer is used to split processing per partner (one consumer per major retailer Walmart / Target / Amazon) without duplicating the queue. Headers are populated by the AS2 Gateway at ingestion: `AS2_FROM`, `AS2_TO`, `DOC_TYPE`. Camel consumers subscribe with `selector='DOC_TYPE = ORDERS'`. Beware of relying on headers not guaranteed (the partner may omit `AS2_FROM` in some signed-only flows).

Anti-patterns

• Too-broad selector that matches everything — the pattern degenerates into disguised pull-and-drop.
• Selector that depends on payload (impossible broker-side) — confused with an application Message Filter.
• Overlapping selectors across consumers — same message pulled twice.
• Non-exhaustive selectors (no consumer matches) — message stuck until TTL or DLQ.

Related patterns

• Message Filter — application-side filter (post-pull).
• Competing Consumers — scaling out a single queue.
• Datatype Channel — alternative: one queue per type.
• Event-Driven Consumer — complementary push consumer.

Sources

• Hohpe G., Woolf B. — EIP, Selective Consumer (p. 515). www.enterpriseintegrationpatterns.com/patterns/messaging/MessageSelector.html
• Apache ActiveMQ — Message Selectors. activemq.apache.org/components/classic/documentation/selectors