developmentalmadness / sqlstream Goto Github PK

The readme is currently too minimal. Include some sample code as well as why anyone would want to use TVPs. Include some stats.

Also, include the fact that the stream is buffered and the actual output to Sql Server is asyncronous and uses lock-free concurrency to share the data between threads.

I don't like the current defaut log4net output when running the test in NUnit. Find out why the configuration in the setup is being ignored.

The example in the test project isn't complete because it doesn't actually connect to Sql Server. Write a script that Defines a UDT and and Stored Procedure that uses it.

The current implementation using queues of buffers/chunks doesn't enforce any memory limits. The ring buffer a la LMAX would prevent running out of memory with large files sending data to a slow/overloaded Sql Server.

Sql Server doesn't start the stored procedure until the stream has completed (IEnumerator.MoveNext() == false). In cases where the full dataset may be too much for Sql Server to process at once. The client would prefer to call ExecuteXXX() each time the buffer flushes instead of buffering to the same stream. This should all work internally w/o any changes to the interface other than adding a new value to the SqlStreamBehavior enum.

NOTES:

• This option should be ignored by ExecuteReader and ExecuteScalar, or we can offer a callback for these so the client can handle what they want to do with each response.