Spring Rest应用程序中的Java静态/全局变量

Question

I'm working on a Spring Rest Service Web application where as soon as a HTTP request is received in controller I set a static String variable null . Further in application, this variable is modified multiple times by various classes and methods, and in the end, this is returned as rest service response.

I'm afraid that this static String variable might be updated by another concurrent HTTP request resulting in incorrect value in previous request.

I don't wish to use instance variable as that would need to be passed to multiple methods (>30) across the application.

Please suggest a way to declare a global variable which is not updated by any concurrent HTTP request and have a separate copy for each HTTP request.

Answer 1

All your validation methods already take a parameter, right? The data to be validated.

They also all need the ability to generate validation messages and store them somewhere, eg in a "result collector", right? So that should be a parameter too.

Side note: Giving both the data and the result collector to the validators as parameters allows you to unit test the validators.

I don't see the problem with passing the result collector around to all the validators, the same way you're passing the data to be validated around. It's how the code should have been written to begin with.

Never use "global" memory in a web application.

Viewed purely from a validators point of view, you pass in the data to be validated, but expect the result to be written to a static collector? Bad design , right there. The validator should not be responsible for figuring out how to find the result collector.

Answer 2

The controller has a singleton scope, it should not contain a data. If you need to store data to to the application use the corresponding scope of the bean containing the data string.

If you are using a bean which is shared among other users then the data should be protected from concurrent modifications. Which way you provide the protection to users is not important at the moment but affordable data structures simplify the implementation of such beans.