Change Log

Changes between 3.0.1 and 3.1.0

Improved Region configuration support

Whereas before the only acceptable string format for the #aws/region tag was the region enum format (i.e. "US_WEST_1"), the tag can now also take the more common name format (i.e. "us-west-1"). The start-consumer now can take either of the string formats or a com.amazonaws.regions.Region instance, whereas before it only took the instance type.

See the Region section of the README.

Changes between 3.0.0 and 3.0.1

Excluded logback.xml from library JAR file

Logging config files don't belong in library JARs, and having this one in there resulted in logback-using consumer code generating warnings about multiple logback configs on the classpath. Library consumers should get full control over their logging configs, warning-free.

Changes between 2.0.0 and 3.0.0

consume-messages now returns a consumer-id instead of a future

In 1.0.0 it returned a future that you could cancel when you wanted to stop consuming messages. But in 2.0.0 that stopped working because of the retry loop added in that version.

In 3.0.0 it will return a consumer-id instead. You can then call (squishy.core/stop-consumer consumer-id) when you want to stop processing.

This is a breaking change.

consume-messages will now handle increasing the visibility timeout of SQS messages

By default SQS queues have a visibility timeout of 30 seconds (though you can increase it in each queue's config). This means that if it takes your code longer than that timeout to process a message, SQS will make it available to other consumers. In 3.0.0, squishy will now tell SQS to increase the visibility timeout of a message that is still being processed so that you don't have redundant processing.

Changes between Squishy 1.0.0 and 2.0.0

consume-messages can recover from some SQS connection problems

consume-messages used to not be able to recover from, for example, a connection reset when trying to read a message from the queue. And since consume-messages does its work in a separate thread, the exception thrown would not be handled. Now, however, Squishy will attempt to retry, backing off, before finally giving up, logging a fatal error, and realizing the future.

This required the first argument for consume-messages to change. Instead of passing a client, you provide a map of configuration containing your :access-key, :access-secret and :region. Squishy will create clients as needed.

This is a breaking change.