Selector format

[flowersinthesand]

Currently id and tag are provided as selector for socket, but there is no universal selector concept. The universal selector format will make it neat and enjoyable to do 'select some sockets and do something with them' pattern. It's worth sacrificing performance within limits.

server.by(".donghwan");

In any format, we might need to modify and accept it in our own way but I don't want to reinvent the wheel and modify too much things.

URI

It deals with id and tag as path component of URI which doesn't start with '/'.

• As URI means literally, it is for identifying not selecting. It can provide a clean way to identify various tags but that's all.
• It's awkward to use it as selector format in my opinion.
• Though Ant have used the wildcard to URI with such that purpose. For example, a/* and a/** match with a/b and a/c for former and a/b, a/b/d, a/c, a/c/e for latter respectively.

  CSS

It deals with id and tag as similar one from CSS selector.

• CSS selector is powerful to use but complex to implement and use sometimes.
• A socket id is an element id and a socket tag is an element class.
• CSS selector is for tree, hierarchical data structure. In our case, there is no hierarchy between sockets. This is the root difference making the situation complex.
• It can cover everything we want but too much. Many selectors are not meaningful here.
• .a:not(#881f80cf-4f8b-4065-8286-b42f2fb62a51): Tag a excluding id 881f80cf-4f8b-4065-8286-b42f2fb62a51 that is a typical pattern meaning 'everybody in room, a, excpet specific one'.

  SQL

It thinks of given selector as where clause where the fixed SQL is select socket from server where ${selector} and regard socket's properties like uri, id, tags as column of table.

• Describable but monotonous and tedious
• It's not easy to implement in my opinion.
• tags in a and id <> 881f80cf-4f8b-4065-8286-b42f2fb62a51: the same as above.

[flowersinthesand]

Expression Language like JSP EL, OGNL, SPEL would be adopted as well. It's easy to adopt since there is already implementation for similar reason but it's only for Java. It might make the protocol specific language-dependent.

[flowersinthesand]

A plain SAM in Java 8 would be considered also. However it is too similar to a usage of SAM as filter in all method.

server.by(s-> s.tags().contains("a") && !s.id().equals("881f80cf-4f8b-4065-8286-b42f2fb62a51"));