Hi, As I know, RocketMQ only support filter message by tag which one message could own one. It's too simple and not enough to meet some customized scenarios.
So RockeMQ should have the ability to filter message through some reasonable languages, such as SQL92, AVIATOR(from google) and any others, no matter which one. I think the ActiveMQ's solution which define the SQL92 grammar with javaCC is simple and scalable, so could integrate all or part grammars into RocketMQ. Just as Von Gosling said, we could make a comment in the code, which is learn from ActiveMQ. Later, we also could integrate aviator or any other module into RocketMQ to support more features. Thanks, Eric On Fri, Mar 17, 2017 at 9:51 AM, Von Gosling <[email protected]> wrote: > Hi, > > > There is no doubt that the feature is a great polish for RocketMQ filter > implementation, which is using a proxy mechanism. So, IMO, we can use the > experience of ActiveMQ’s solution - SQL 92 grammar with JavaCC compiling. > But, we must make a comment in the code, which is learn from ActiveMQ - > just like other Apache project similar practice. In fact, some guys have > contact us privately, they have done this work. We encourage them to make > discussion in dev list. Believe, if they have seen this mail, he must be > greatly happy to make work together. Let’s polish together, making RocketMQ > filter become more practical :-） > > > > 在 2017年3月17日，06:31，Rick Zhang <[email protected]> 写道： > > > > Hi Bruce, RocketMQ Team, > > > > RocketMQ Team hopes to add the functionality that support message filter > > based on a subset of SQL92 expression syntax [1]. As everyone knows, JMS > > Spec already defined a set of selector syntax , which is widely known by > > people around java world. In particular, open source like ActiveMQ has > > already implemented it very well (ActiveMQ also extends it to support > > XPATH). > > > > My question: Is it appropriate that extracting ActiveMQ's JMS selector > > implementation into RocketMQ codebase(maybe we will add some extensions > > upon it)? > > > > I have no idea that if this behavior is correct. I just feel do the same > > thing again may be a waste of time, especially the performance of JavaCC > > used by ActiveMQ is good enough, and I have no sufficient reason to > rewrite > > a brand new one. > > > > Looking forward you to give us some suggestions. > > > > Thanks > > > > [1] https://issues.apache.org/jira/browse/ROCKETMQ-121 > >
