milgner
commented
2 years ago Found the following TODO comment in stream.rs of the mssql module:
pub(crate) fn write_packet<'en, T: Encode<'en>>(&mut self, ty: PacketType, payload: T) {
// TODO: Support packet chunking for large packet sizes
// We likely need to double-buffer the writes so we know to chunkAs the default packet size is 4096, this is in all likelihood the reason for the problem, as the encoded length of a ~2000 character-long query (assuming it's encoded as UTF-16) could result in that threshold being exceeded.
I found this in the specs:
We'll try to allocate some time in one of our next sprints to add support for packet chunking, in the meantime we have resorted to chunking our queries instead :grimacing:
If anyone has some pointers, requests and recommendations, please leave them here :slightly_smiling_face:
lovasoa
When trying to run a long query against an MS SQL server, I'm getting
Io(Kind(ConnectionAborted)). After some trial and error, I have narrowed down the issue to the query string being 1992 (works) or 1993 (breaks) characters in length.However, I can run my long queries without problems using other tools like the database explorer in IntelliJ IDEA (uses JDBC) so I must assume that the problem is not with the server itself but somewhere on the client side. I'll try to dig and debug deeper into the library and see whether I can find anything but haven't had much luck so far. Considering that a few bytes will be added protocol-wise and there's also one bound variable that needs to be transferred, I was thinking that maybe there's a buffer limited to 2kb somewhere along the line.