I am trying to figure out how to work with slick streaming. I use slick 3.0.0 with postgres driver
The situation is following: server have to give client sequences of data split into chunks limited by size(in bytes). So, I wrote following slick query:
val sequences = TableQuery[Sequences]
def find(userId: Long, timestamp: Long) = sequences.filter(s ⇒ s.userId === userId && s.timestamp > timestamp).sortBy(_.timestamp.asc).result
val seq = db.stream(find(0L, 0L))
I combined seq with akka-streams Source
, wrote custom PushPullStage
, that limits size of data(in bytes) and finishes upstream when it reaches size limit. It works just fine. The problem is - when I look into postgres logs, I see query like that
select * from sequences where user_id = 0 and timestamp > 0 order by timestamp;
So, at first glance it appears to be much (and unnecessary) database querying going on, only to use a few bytes in each query. What is the right way to do streaming with Slick so as to minimize database querying and to make best use of the data transferred in each query?
The correct way to stream in Slick is as provided in documentation is
The "right way" to do streaming with Slick and Postgres includes three things:
Must use db.stream()
Must disable
autoCommit
in JDBC-driver. One way is to make the query run in a transaction by suffixing.transactionally
.Must set
fetchSize
to be something else than 0 or else postgres will push the whole resultSet to the client in one go.Ex:
Useful links:
https://github.com/slick/slick/issues/1038
https://github.com/slick/slick/issues/809