Re: Batches of single-insert statements vs batches ofmulti-insert statements

Hi Jeremy,

 

The resulting SQL is not hand-crafted, and is built through all sorts of layers depending on various rules. The end result that I exposed does not tell how it was built.

Note that from a theoretical point of view, if I were to add parameters for fixed values, then I would send through the wire as many times this parameter, while if it is part of the statement the driver can decide to only send the dynamic parameters. Also, we do have external multi-insert logic for that specific part of the code, and it also improves speed with jTDS; but SQL Server is limited to 2000 dynamic parameters, so if we waste those parameters with static values, less batches can go through at once.

In any case, it must be supported by the batch rewrite feature, whether people like the style or not :)

 

In fact, the latest version of my patch hooks into the existing query parsing, and just has to record the start and end of the values block. That combined with the positions of the parameters made the latest version reliable.

That being said, there is currently no safety in case we put too many batches overflowing Postgres limits, which is what I will try to work on when I get some free time.

Cheers,

-Christopher