Re: [rdf4j-dev] Question about READ_UNCOMMITTED

[James Leigh <james.leigh@xxxxxxxxxxxx>]

Hi Håvard,

It looks like the test in wrong. I think it was originally written
before isolation levels and should be skipped for READ_UNCOMMITTED as
well.

Can you create a PR for this?

Thanks,
James

On Sat, 2017-07-22 at 16:28 +0200, Håvard Ottestad wrote:
> Hi,
> 
> There is a test in RepositoryConnectionTest called testEmptyCommit().
> This test seems to do the following:
> 
>  - Create two connections
>  - Add data to one of the connections
>  - Make sure the other connection doesn’t see said data
> 
> This test is being run at READ_UNCOMMITTED isolation level. I’m
> wondering then if the definition of READ_UNCOMMITTED is correct.
> 
> The current definition is: 
> /**
>  * Read Uncommitted: transactions can be rolled back, but not
> necessarily isolated: concurrent
>  * transactions might see each other's uncommitted data (so-called
> 'dirty reads')
>  */
> 
> From the test I wonder if READ_UNCOMMITTED is actually meant to be a
> snapshot isolation against other transactions adding data after a
> READ_UNCOMMITTED connection has opened? So you can do dirty reads,
> but only from writes that happened before the connection was opened.
> 
> Håvard
> _______________________________________________
> rdf4j-dev mailing list
> rdf4j-dev@xxxxxxxxxxx
> To change your delivery options, retrieve your password, or
> unsubscribe from this list, visit
> https://dev.eclipse.org/mailman/listinfo/rdf4j-dev