JDBC Translator | Queries on batch commit and Update Count

[shiveeta.mattoo]

Hi,

Two questions related to functionality provided by JDBC Translator

1. Segregation of batch execution and commit mode in JDBC Translator

Currently we see that batch operations (Insert/Update) are processed as expected. However the batch execution and commit are coupled together. Hence, we only have the ability to commit after a batch completes. We are also seeking an ability to support the commit after a configurable 'x' no. of batches has completed, instead of an automatic commit after every batch.

We have a similar batch transaction handling framework for an internal JDBC layer. In that we provide the user with an option to specify the no of batches after which a commit is desired.We are able to achieve the same with the base driver, because it does not automatically commit when we hit batch size. We are seeking a similar support from Teiid JDBC translator. Please advise.

2. Update Count Reported on Batched Update Execution.

Typically, the executeBatch() returns an array of integers, and each element of the array represents the update count for the respective update statement.

However in case of JDBC Translator, over and above this, we are processing the array to return a sum of all the elements as in seen in JDBCUpdateExecution -> executeTranslatedCommand method

int[] results = pstatement.executeBatch();

for (int i=0; i<results.length; i++) {

updateCount += results[i];

}

Is there any specific reason to perform this? Can we change this to stick to the expected standard behaviour?

Reason for requesting this, is that we have a generic DB access layer built on top of our virtualization layer, which expects the updatecount in a standard format.

Instead of embedding some specific checks in our layer, I was keen to understand any definite need for the non-standard behaviour.

Thanks for your help in advance

Regards,

Shiveeta

[shawkins]

> 1. Segregation of batch execution and commit mode in JDBC Translator

Without an enhancement or a significant translator override, you'd have to proxy your driver to control the setAutoCommit and batch logic.

> Is there any specific reason to perform this? Can we change this to stick to the expected standard behaviour?

I think the code there is simply out of date.  It should handle the batch case differently, can you log an issue?

[shiveeta.mattoo]

Thank you Steven for prompt response.

>1. We are integrating with multiple databases, so proxying each and every driver may not work out very well for our use-case.

>2. I will log the issue for the Update count behaviour. Would it also be addressing the segregation of batch execution and commit as part of the batch code refactor ?

[shiveeta.mattoo]

Please confirm if I should log issue only for update Count behaviour or also on the batch execution and commit segregation.

Thanks a lot for your help.

[shawkins]

You can a bug for the update count and an enhancement request for the commit behavior.

[shiveeta.mattoo]

Thank you Steven. Will log the same.