Also, it should be possible to create something like quill-generic in Scala 3. It's just a matter of time and labor.
@deusaquilus
Also, it should be possible to create something like quill-generic in Scala 3. It's just a matter of time and labor.
Proto Quill basically has a bus factor of 1 (you being the 1 :)), so the nice-to-haves might not come soon enough for those looking to use proto Quill in the near-term. Saying that, I noticed yesterday that there's a code generator for Quill that could be used to generate DAO boilerplate without relying on an external library.
Hopefully Scala 3 macros will be able to replicate full Scala 2 Quill feature set without too many workarounds/concessions required.
Impressive work, btw, you've clearly taken the deep dive into Scala 3 macros (just testing the waters here now)
ProtoQuill was a long time coming. I've worked on it with Nicolas Stucki, Aggelos Bibodis, and others from early 2020. I've had to take a break from it for a couple months because the Quill ZIO context needed to be reworked in order to be a really great experience. Hopefully in the coming months I'll be able to get back into it. Also, if there is desire, I can give a series of public deep-dives on ProtoQuill.
Any resources the community can access that reveal the inner workings of the Scala 3 compiler vis-a-vis macros would be greatly appreciated. Through much trial and error, tree printing, etc. I managed to convert a complex (for me) Scala 2 macro to the Scala 3 equivalent. Learned a lot in the process but it was more challenging than it needed to be. If we had thorough documentation and real world examples to work from, the Scala 2 to Scala 3 macro migration would go much more smoothly, not to mention faster.
At any rate Quill is the gold standard of Scala 2/3 macros.
Also, in regard to implicit conversion from case class companion object Foo to query[Foo], I suspect that until lampepfl/dotty#7000 is resolved, getting that functionality working in Scala 3 will be quite the challenge. Basically flying blind, all you get with Foo.join(x => ...) at the macro level is a measely Product, no companion object Mirror, no case class ClassDef, nada.
extension [T](inline ent: T.type)
inline def insert(inline f: (T => (Any, Any)), inline f2: (T => (Any, Any))*): Insert[T] = query[T].insert(f, f2: _*)
inline def update(inline f: (T => (Any, Any)), inline f2: (T => (Any, Any))*): Update[T] = query[T].update(f, f2: _*)
inline def insert(inline value: T): Insert[T] = query[T].insert(value)
inline def update(inline value: T): Update[T] = query[T].update(value)
inline def map[R](f: T => R): Query[R] = query[T].map(f)
inline def flatMap[R](f: T => Query[R]): Query[R] = query[T].flatMap(f)
inline def concatMap[R, U](f: T => U)(implicit ev: U => Iterable[R]): Query[R] = query[T].concatMap(f)
inline def withFilter(f: T => Boolean): Query[T] = query[T].withFilter(f)
inline def filter(f: T => Boolean): Query[T] = query[T].filter(f)
// etc... look into everything in Model.scala in regular-Quill@deusaquilus T.type isn't valid scala syntax, but it would be convenient if it was :)
Examining scala 3 case class compiler generated source code in more detail, it's worse than I thought -- all we have to work with in the companion object is java.lang.Object, but still that's a start, can at least trigger macro invocation with static Foo.filter(...) call. From there I think we should be able to add a second type param to represent companion class T <: Product and check if it's a case class within body of macro; if not, error and abort, or else proceed with construction of (x: U) => query[T] Conversion[U, T] macro return type.
i noticed a confusing (at least to me) behaviour when trying out quill with spark. i have this small example (scala=2.12.13,spark=3.1.2,quill=3.11.0)
import io.getquill.QuillSparkContext._
import io.getquill.Quoted
import org.apache.spark.sql.{Dataset, SQLContext}
import java.time.Instant
object TestQuill {
implicit val instantEncoder: MappedEncoding[Long, Instant] = MappedEncoding[Long, Instant](Instant.ofEpochMilli)
implicit val instantDecoder: MappedEncoding[Instant, Long] = MappedEncoding[Instant, Long](_.toEpochMilli)
implicit class InstantQuotes(left: Instant) {
def <(right: Instant): Quoted[Boolean] = quote(infix"$left < $right".as[Boolean])
}
implicit val sqlCtx: SQLContext = ???
import sqlCtx.implicits._
val ds: Dataset[Instant] = ???
val ts: Instant = ???
run {
liftQuery(ds).filter(_ < lift(ts))
}
}if i specify the output type for the extention method explicitly def <(right: Instant): Quoted[Boolean] = ... it doesn't compile:
The query definition must happen within a `quote` block.
[error] liftQuery(ds).filter(_ < lift(ts))
[error] ^if i leave out the type def <(right: Instant) = ...it compiles fine and generates the expected query.
i'm just curious if someone could explain this difference.
I need to accept a dynmica list of sort fields and ordering as input and translate that into a database query. I’ve seen a number of questions around how to accomplish dynamic sortBy blocks but still haven’t seen a working example that doesn’t use infix to dynamically generate the ORDER BY section of a query. Has anyone managed to accoplish this or have an example?
I’ve managed something like this but it generates a nested SQL query for each element in my sortFields list rather than a single query with a single ORDER BY block:
val sortFields: Seq[SortField] = ???
val qry = myTable.filter(…)…..
sortFields.foldLeft(qry) { (q, field) =>
field match {
case NameAsc => q.sortBy(_.name)(Ord.Asc)
case NameDesc => q.sortBy(_.name)(Ord.Desc)
case IdAsc => q.sortBy(_.id)(Ord.Asc)
case IdDesc => q.sortBy(_.id)(Ord.Desc)
...
}
}
Connection to support the old code. And to be able to get hold of it from within a transaction block, I should think ... Tips?
Here's the route in the webserver, which is based on Akka http
~ path(IntNumber) { id =>
concat(
get(complete(Future(DB.aTodo(id)))),
delete(complete(Future(DB.deleteTodo(id))))
)And here's the DB "implementation"
val ctx = new PostgresJdbcContext(SnakeCase, new HikariDataSource(config))
import ctx._
def aTodo(id: Int) : Seq[Todo] = ctx.run { query[Todo].filter(_.todoId == lift(id)) }
inline def updateTodo(todo: Todo) = ctx.run {
query[Todo].filter(_.todoId == lift(todo.todoId)).update(lift(todo))
}Issue 1
Looking at the docs, the "canonical compile time" solution for aTodo would be inline def. If I do that, I get a message about "could not summon a parser factory, and cannot find a "scala.Int" Encoder of Id. So figured I leave that without the inline.
Issue 2
"io.getquill" %% "quill-jdbc" % "3.10.0.Beta1.6",
Hi,
I would like to make use of the "insert on conflict update where is distinct from"-clause in Postgres in a generic fashion. Therefore I want to create a macro where I can write something like this:
case class Domains(id:Int, domain:String)
quote{
insert[Domains](Domains(1, "a"), _.id)
}Where the macro should be something like this:
def insert[A](row: A, idColumns: (A => Any)*): Insert[A] = macro InsertMacro.insertImpl[A]
def insertImpl[A: c.WeakTypeTag](c: Context)(row:c.Expr[A],idColumns:c.Expr[A => Any]*): c.universe.Tree= {
import c.universe._
val t=weakTypeOf[A]
val r=q"""
val q = query[$t].insert($row)
val tableName = naming.table(${t.typeSymbol.name.toString})
infix"$$q ON CONFLICT (id) DO UPDATE SET id = excluded.id WHERE (#$$tableName.*) IS DISTINCT FROM (excluded.*)".as[Insert[${t}]]
"""
r
}Unfortunately, the compile fails with Tree 'context.naming.table("Domains")' can't be parsed to 'Ast'.
Is there a way to translate the case class name to the table name and use it in the infix part within the quote{...} block?
@andyfr Is naming available at compile time? If so create the tableName variable in the macro & try something like this:
def insertImpl[A: c.WeakTypeTag](c: Context)(row:c.Expr[A],idColumns:c.Expr[A => Any]*): c.universe.Tree= {
import c.universe._
val t=weakTypeOf[A]
val tableName = naming.table(t.typeSymbol.name.toString)
val r=q"""
val q = query[$t].insert($row)
infix"$$q ON CONFLICT (id) DO UPDATE SET id = excluded.id WHERE (${io.getquill.ast.EntityQuery($tableName)}.*) IS DISTINCT FROM (excluded.*)".as[Insert[${t}]]
"""
r
}That way you can splice the tableName into the query as a static string and the whole thing should be able to produce a compile-time query
2 replies
Within the next couple of weeks I will look through the issues covered here and try to add relevant things to the Quill FAQ
(any help with this would be greatly appreciated, please let me know if you would like me to post on a specific question that has been asked in this thread).
Hello there!
I'm totally new to quill and playing a bit with proto-quill on top of a toy sqlite database, I can run simple queries fine but struggle with putting together joins,
here's the error I'm getting:
Main.scala:43:17: Exception occurred while executing macro expansion.
scala.MatchError: TypedOrTest(Unapply(TypeApply(Select(Ident("Some"), "unapply"), List(Inferred())), Nil, List(Bind("it", Wildcard()))), Inferred()) (of class java.lang.String)
Here are my dependencies:
ivy"org.xerial:sqlite-jdbc:3.36.0.3",
ivy"io.getquill::quill-jdbc:3.17.0.Beta3.0-RC2"and here is in substance the code
import io.getquill._
import io.getquill.QueryDsl.like
object Main {
case class Artist(name: String)
case class BuilderID(id: Int)
case class Builder(id: BuilderID, name: String)
case class InstrumentTypeID(id: Int)
case class InstrumentType(id: InstrumentTypeID, description: String)
case class InstrumentID(id: Int)
case class Instrument(
id: InstrumentID,
typeId: InstrumentTypeID,
builder: BuilderID,
name: String
)
val ctx = new SqliteJdbcContext(SnakeCase, "ctx")
import ctx._
def main(args: Array[String]): Unit = {
inline def someInstruments = quote {
query[Instrument]
.leftJoin(query[InstrumentType])
.on((i: Instrument, it: InstrumentType) => i.typeId == it.id)
.filter({ case (i, Some(it)) => it.description.startsWith("Ukulele") })
.map((i: Instrument, it: Option[InstrumentType]) => i)
}
println(run(someInstruments))interestingly, it's seems to be an idiosyncratic thing
inline def someInstruments = quote {
query[Instrument]
.leftJoin(query[InstrumentType])
.on((i: Instrument, it: InstrumentType) => i.typeId == it.id)
.filter(_._2.map(_.description.startsWith("Ukulele")).isDefined)
.map((i: Instrument, _) => i.name)
}would work, on the other hand. So I guess one has to find a syntax that pleases the compiler and stick to it.
val q = quote {
for {
user <- dynamicQuerySchema[User]("user")
_ <- dynamicQuerySchema[UserToContract]("userToContrat").join { utc =>
{
contractIdOpt match {
case Some(id) => quote(utc.contractId == lift(id))
case None => quote(true)
}
} && utc.userId == user.id
}
} yield user
}quote {
query[User]
.join(query[UserToContract])
.on((u: User, utc: UserToContract) => utc.contractId == lift(id) && utc.userId == u.id)
//.map((u: User, _) => u)
}
Hi, I am using the quill-cassandra-zio library.
I have a simple query like this:
private def itemItemsetQuery(itemId: Long, catalog: String) = quote(
querySchemaItemItemsets.filter(c => c.itemId == lift(itemId)).filter(.catalog == lift(catalog)).map(.itemsets)
)
run(itemItemsetQuery(itemId, catalog))This compiles fine and works fine too but it gives me the below warning every time it is run:
Re-preparing already prepared query is generally an anti-pattern and will likely affect performance. Consider preparing the statement only once. Query='SELECT itemsets FROM item_itemsets WHERE item_id = ? AND catalog = ?
Does someone know what is wrong with my query? this is how the documentation suggests it should be done too so I have not been able to fix this yet
