Spring transaction support

[maxsap]

It would be most helpful to be able to wrap or even bind a dynamodb with a normal Spring @transactional method.

[derjust]

https://aws.amazon.com/blogs/aws/dynamodb-transaction-library/
https://github.com/awslabs/dynamodb-transactions/issues

[derjust]

Finally some native support: https://aws.amazon.com/blogs/aws/new-amazon-dynamodb-transactions/

[boostchicken]

@derjust It's interesting, it allows you to tranact saves of multiple objects in a single dynamo call, which kind of flys in the face of how spring data does things. We are probably going to have a make a new type of repo that is "crud" for this since it could take pretty much any object.

Either that or use the Transactional annotations and do some translation on the backend to wrap all the calls together some how. Any thoughts?

[boostchicken]

To expand on that the existing Spring interfaces won't really work for this. There is no persistent connection to the database, beyond that there is no rollback, Amazons API just throws an exception if something is using that item in a transaction at the moment.

We are going to have to come up with a new pattern.........

[boostchicken]

Also it's only supported in the DynamoDB Async client, (makes sense) and does not have support for the mapper, so if we want to do this we will really have to hit the drawing board

From the docs:
Transactional operations are not supported by DynamoDB Accelerator (DAX), DynamoDB local, or DynamoDB Mapper at this time.

[291277058]

To expand on that the existing Spring interfaces won't really work for this. There is no persistent connection to the database, beyond that there is no rollback, Amazons API just throws an exception if something is using that item in a transaction at the moment.

We are going to have to come up with a new pattern.........

I am thinking about this.I perferm use a threadlocal cache to cache the entity,and the id should generate at local server like snowflowId.My question is how to hit the cache when the query is complicated