Its how its been, its the pattern I followed. I don't have a better explanation unfortunately, this is all pushing past where I have any sort of confidence in what i know and understand
What would an alternative to a pipe look like? One the pipe for parts is just an abstraction that wraps doobies
updateManyWithGeneratedKeys. What else might it be if not a pipe? Or, why does using a pipe here cause trouble? Someone else mentioned refactoring the pipes to Chunk[A] => IO[Chunk[A]] might help but I'm not quite sure what they mean by that or where the change would happen. I think they thought that perhaps having these all in ConnectionIO could be a source of issues?
(also, happy to continue chatting in fs2 if this fits better there)
@gnfisher, with chunks the above can be simplified to
input.evalMapChunk { chunk =>
processA(chunk.map(_.a)) >> processB(chunk.map(_.b))
}Here processA: Chunk[A] => ConnectionIO[Chunk[B]] and so on. It's a little easier to reason about than streams.
Upd: removed the .chunks call.
Also less opportunity to confuse
.eval and .eval_.
I will try that refactor to see if it makes a difference
Oh, I am on 2.0.1 of fs2 so i dont have evalMapChunk :(
but
processA does not return a nything that processB takes as input
I need to go, but text me later, and I can explain you
>> vs ++
Hi, I'm having an issue with a mssql query. I try to fetch all ~25k rows from a table, and get this:
Jun 04, 2021 9:05:24 AM com.microsoft.sqlserver.jdbc.TDSReader throwInvalidTDS
SEVERE: ConnectionID:3 ClientConnectionId: 3c8eb486-fe5f-4e72-b267-69e79e780b93 got unexpected value in TDS response at offset:7992
Jun 04, 2021 9:05:24 AM com.microsoft.sqlserver.jdbc.TDSParser throwUnexpectedTokenException
SEVERE: ConnectionID:3 ClientConnectionId: 3c8eb486-fe5f-4e72-b267-69e79e780b93: getNextResult: Encountered unexpected TDS_FEATURE_EXT_UTF8SUPPORT (0x0A)which to me looks like an encoding issue, but when I paginate the query; e.g. order by foo offset 0 rows fetch next 5000 rows only – I can query all the pages fine.
I'm using doobie, but this might be a pure jdbc issue?
@nor: yep looks like it. Give raw jdbc a go and see if you get the same issue
ok.... here's a question.... I have a bunch of ConnectionIO's strung together in a for comprehension.... half way through these I need to save something to the file system and base on where it ends up update a database. Because I'm using zio, this is somewhat what I have:
def save(saveMe: MyComplicatedObject, saveToFileSystem: MyComplicatedObject => ZIO[A,B,File]): ConnectionIO[MyComplicatedObject] = {
for {
res1 <- savePart1(saveMe) //returns ConnectionIO[MyComplicatedObject]
res2 <- savePart2(res1) //returns ConnectionIO[MyComplicatedObject]
res3 <- saveToFileSystem(res2) // <<<<< here... how to convert ZIO to ConnectionIO
res4 <- savePart3(res3) //returns ConnectionIO[MyComplicatedObject]
} yield res4
}As you can see, the basic question, is how to convert a ZIO to a ConnectionIO? I know I can use .pure to get a ConnectionIO out of a normal thing, but I doubt .pure on a ZIO IO would give me what I want.
But in any case if you use cats-effect IO you can say myIO.to[ConnectionIO]. It may or may not work with ZIO, I don’t know.
@rleibman I assume you already have
import zio.interop.catz._ in this module. So when you define savePart1, savePart2, savePart3, they should return zio.Task[?], instead of ConnectionIO. Same goes with this savemethod. The interop should perform the magic and convert ConnectionIO to zio.Task automatically.
Hi @rleibman, in that case, for the example you gave,
I think I understand what you're trying to do in the example. But in reality, having a non-DB operation as part of a DB transaction is a bit weird because I don't think a DB rollback can undo the change made in
saveToFileSystem itself should return ConnectionIO instead of ZIO. Perhaps in that method, you are using a lib or other custom code that returns a ZIO. In that case, I'm not aware of a lib that can convert from ZIO to ConnectionIO. You might have to evaluate the ZIO effect right there and then wrap the result into a ConnectionIO.I think I understand what you're trying to do in the example. But in reality, having a non-DB operation as part of a DB transaction is a bit weird because I don't think a DB rollback can undo the change made in
saveToFileSystem. In the distributed (or not) saga pattern, you would specify compensation actions in event of a rollback (for all participating systems). I don't think ConnectionIO has something like that because I assume it just relies on the underlying RDBMS engine's rollback mechanics. I'm interested in a solution though.
I skimmed through ZIO's
interop-cats source code and I didn't spot routines going from ZIO to Cats IO. I might be wrong though. You should post this question on ZIO discord. Like I mentioned, you can always call unsafeRun on a ZIO effect and lift the result into Cats IO if you choose to do so.
@rleibman: In our project, we are using https://github.com/gaelrenoux/tranzactio to lift the
ConnectionIO to ZIO. You might want to look into that.
Hello!
I'm new around. I have a question: Let's say I have a stream of Integers from a query, and I want to save all of them in a single transaction but without converting it to a List. Is it possible? :/
def (intStream: fs2.Stream[ConnectionIO,Int]): IO[Unit] = {
???
}I really miss more documentation and examples on fs2.Streams and doobie related :_(
Is something like this correct? Would it be memory-safe?
// for simplicity, insertValIntoOtherTable: Int => ConnectionIO[Unit]
def (intStream: fs2.Stream[ConnectionIO,Int]): IO[Unit] = {
intStream.compile.fold(connection.unit)(
(accConnectionIO, intVal) => {
accConnectionIO *> insertValIntoOtherTable(intVal)
}
).flatten.transact(transactor)
}That last flatten... :/
Hi,
Im trying to write an INSERT ON DUPLICATE UPDATE query in doobie. Usually when doing an INSERT only I use
Im trying to write an INSERT ON DUPLICATE UPDATE query in doobie. Usually when doing an INSERT only I use
withUniqueGeneratedKeys[Int]("id") to get the auto-generated ID of the inserted row. I tried doing the same for the INSERT ON DUPLICATE UPDATE query but it was returning the following error when a query updates instead of insertingExpected ResultSet exhaustion, but more rows were available.. I think this occurs because on UPDATE MYSQL returns 2 rows updated to indicate that there was an update instead of an insert. Does this error happen because the withUniqueGeneratedKeys is expecting 1 row only to be updated? I 'fixed' it by using update.run instead. Still I would like to get the original id in case of update. Do you have any recommendations on how to go around it please?
I don't know, sorry. If you can figure out how you would do it with JDBC we can figure out how it would look in doobie.
If it doesn't happen out of the box, it seems reasonable that the feature should be added - open a github ticket?
Hi! I'm using the notify/listen functionality of Postgres. The notify is executed in a DB trigger. It seems after some time my program stops receiving notifications but the stream does not stop with an error:
def notificationStream(
channelName: String,
pollingInterval: FiniteDuration
): Stream[F, PGNotification] = {
val inner: Pipe[ConnectionIO, FiniteDuration, PGNotification] = ticks =>
for {
_ <- Stream.resource(channel(channelName))
_ <- ticks
ns <- Stream.eval(PHC.pgGetNotifications <* HC.commit)
_ <- Stream.eval(logger.info(s"Received ${ns.size} notifications"))
n <- Stream.emits(ns)
} yield n
awakeEvery[F](pollingInterval).through(inner.transact(xa))
}Does somebody have a hint what could be wrong?
@tpolecat , I found impossibl/pgjdbc-ng#517 which sounds similar. They are blaming it on GC. Could this be the right direction?
A few more observations: In psql I can see a similar situation. After some time I get
The connection to the server was lost. Attempting reset: Succeeded. and need to execute listen <channel> again to continue to receive notifications. As a brute force attempt I added Stream.eval(PHC.pgListen(channelName)) to the inner pipe. This seems to work. But I will continue to look for a smarter way.
trying to use scala 3 opaque type with Put and Get. the code compiles but querying Country is non-terminating.
opaque type Code = String
object Code:
def apply(code: String): Code = code
extension (code: Code) def value: String = code
given Get[Code] = Get[String].tmap[Code](x => Code(x))
given Put[Code] = Put[String].tcontramap[Code](x => x.value)
case class Country(code: Code, name: String, pop: Int, gnp: Option[Double])