add functionality to annotation to allow a spel expressions

[scottmf]

It would be very useful if the @retryable annotation allowed an expression using spel to have an easy way to further inspect an exception in order to determine if a retry should occur.

For example:

@Retryable(include=HttpClientErrorException.class, retryOn="#result.getHttpStatus().value() == 401")
public void runExchange() { ... }

[garyrussell]

You could do that relatively easily today with

@Retryable(interceptor="myInterceptor")

Where myInterceptor is a RetryOperationsInterceptor bean with a custom retry policy that looks at the exception to decide in canRetry().

[scottmf]

thanks Gary. I know the interceptor is available but in the spirit of conciseness it would be nice to have a spel expression

[garyrussell]

It's probably not too hard for a single exception, but the include attribute is an array, and no include means retry for all exceptions. The expression might get unruly very quickly...

"#this instanceof T(com.Foo) ? #this.somCondition() : #this instanceof T(com.Bar) ? #this.someOtherCondition() : false"

I am happy to discuss further, though.

[scottmf]

Understood. I would think that for more complex situations a developer would create an interceptor and for simple situations like mine it would be nice not to have to add all the extra boiler plate. I see a lot of annotations using spel like this and situations can occur like you suggested, but for ~ 80% of the scenarios it is very clean.

I have need for the interceptor right now so that's what I am going to use. If you strongly feel that the expression is the wrong thing to do then that's fine. As a consumer of the api, who uses it a lot, it would be a nice feature to add.

[garyrussell]

I just hacked together a quick PoC garyrussell@f2c98b7

Let's see what @dsyer thinks about whether it's worth pursuing.

[shollander]

Hi,
This feature would be very useful for me as well. Any word on getting the fix by @garyrussell incorporated into a release?

What I am really looking for is a way to configure the maxAttempts and Backoff parameters so that they are configurable and not hard-coded at compile time.

[dsyer]

I didn't see this until now. Maybe a PR would help?

[richiep33]

I am very interested in this feature as well. I have considered writing my own due to this limitation. Otherwise, spring-retry appears very useful.

[garyrussell]

The PoC just evaluates an expression against the exception to decide whether it was retryable or not (which is the original request in this issue).

It does not address evaluating the maxAttempts and backOff attributes as expressions; that's really a separate issue and would take a bit more work.

I am offsite all next week so I don't think I could look at this before Dec 12 which might be too late for spring-retry 1.2. I'll see if I can spring free some time tomorrow, but no promises.

[garyrussell]

I went ahead and issued a PR; not sure if we feel it's too big to add to 1.2 after the RC.

I went with discrete attributes to avoid breaking changes.

[shollander]

Thanks for the quick action on this. I tried out the snapshot build and it works perfectly - just what we needed!

[garyrussell]

@dsyer Please note I submitted another PR.