I've switched to using scalameta as a library, now I will have to pass it multiple paths
@nafg something like having a flag if a method is an override would be super useful here I guess, for now in Metals for example we analyze all files and build an index for inheritance. Then if needed we compare all methods between a super class and subclass to see if it's and overriden method
unfortunately it's pretty compelx currently: https://github.com/scalameta/metals/blob/42b7322352ae56a1668fbedf90274043645e7cc9/metals/src/main/scala/scala/meta/internal/implementation/ImplementationProvider.scala
Some heuristics to clean up the noise would be good
The docs are very unclear what dependency to add, in the end the regular
scalameta one seemed right, however metals (vscode) does not find all the dependencies when doing it as an ammonite script
This is what I'm doing right now:
import java.nio.file._
import scala.collection.immutable.SortedSet
import scala.jdk.CollectionConverters._
import $ivy.`org.scalameta::scalameta:4.4.0`
import scala.meta.internal.semanticdb.Locator
import scala.meta.prettyprinters.Structure
import scala.meta.internal.semanticdb.SymbolOccurrence.Role.REFERENCE
import scala.meta.internal.semanticdb.SymbolOccurrence.Role.DEFINITION
import scala.meta.internal.semanticdb.SymbolOccurrence.Role.Unrecognized
import scala.meta.internal.semanticdb.SymbolOccurrence.Role.UNKNOWN_ROLE
import scala.meta.internal.semanticdb.SymbolInformation
import scala.meta.internal.semanticdb.Scala.ScalaSymbolOps
var rs, ds = Set.empty[String]
var syms = Map.empty[String, SymbolInformation]
val targetDirs =
os.walk.attrs(os.pwd, skip = (_, i) => i.isFile, maxDepth = 3)
.map(_._1.toNIO)
.filter(_.endsWith("target"))
Locator(targetDirs.toList) { (path, docs) =>
docs.documents.foreach { doc =>
// println(doc.uri)
doc.symbols.foreach { info =>
syms += info.symbol -> info
}
doc.occurrences.foreach { o =>
o.role match {
case REFERENCE => rs += o.symbol
case DEFINITION => ds += o.symbol
}
}
}
}
(ds -- rs)
.filterNot(_.isLocal)
// .filterNot(_.startsWith("db/migration/"))
// .filterNot(_.startsWith("chesednow/views/html/"))
// .filterNot(_.startsWith("views/html/"))
.filter(_.startsWith("chesednow/requests/sjs/"))
.to(SortedSet)
.foreach { sym =>
println(sym)
}
It can't go do definition of most things, like TextDocument
are they autogenerated from protobuf and not added to the source jar?
@tgodzik which part is relevant? Just computeInheritance https://github.com/scalameta/metals/blob/42b7322352ae56a1668fbedf90274043645e7cc9/metals/src/main/scala/scala/meta/internal/implementation/ImplementationProvider.scala#L64-L84 ?
(Another metals issue is that it's acting like it's interpreting ammonite scripts on scala 2.12)
compute inheritance refreshes the inheritance indexes
we keep the whole structure of classes in memory
then we check an exact method if needed.
Hey guys, somehow the parser does not like the Unit Symbol () as a type:
package uml.strategies.rewriting.packagerep
import org.bitbucket.inkytonik.kiama.rewriting.Rewriter.rulef
import org.bitbucket.inkytonik.kiama.rewriting.Strategy
import uml.UMLUnit
import uml.strategies.rewriting.RewriteStrategy
object DeleteEmptyPackages extends RewriteStrategy[()] {
override def apply(v1: ()): Strategy = {
rulef {
case u@UMLUnit(_,toplevelElements) =>
val filtered = toplevelElements.filter( tp => tp match {
case uml.Package(Nil,_,_) => false
case _ => true
}
)
u.copy(toplevelElements = filtered)
case u@_ => u
}
}
}This is not parsed with an Unreachable Code Sequence Error thrown. If i exchange v1:() to v1:Unit and [()] to [Unit] it does parse. Do I need to enable some dialect for this ?
scala> def apply(v1: ()) = ""
<console>:1: error: '=>' expected but ')' found.
import scala.meta.dialects.Scala213 I think
But thanks! Seems Scala 2.12 does not support this.
() is not a type in itself, but it's interpreted as a singleton type literal ().type when a type is expected. That works out as equivalent to Unit because it has only the single inhabitant
might it be possible to use semanticdb to find unused symbols within the scope of a project? has someone perhaps done something like that already?
i mean within the larger scope of a project, like detect the situation where i have a public class
Foo somewhere but it’s not used by anything anywhere in the project
or a symbol within such a class
and check is a reference is just used once
so if a symbol has a single occurence
though it might not work perfectly if there is multiple classes with the same name and package
and there is no such functionality yet implemented really
@liff:matrix.org it's possible to implement this analysis but it's not easy. One challenge is to detect the usage of overridden methods. SemanticDB occurrences tell you which symbol a method call resolved to at compile-time but you need custom analysis to find all of the implementations of that method (or supermethods)
Runtime reflection and compile-time macros add another layer of complexity, which might be relevant or not depending on your use-case
Hey, i have extremely simple question: I read my scala file, parse it to Tree, make small modifications and want to save it back but i realised after i parsed my code to Tree i lost all the formatting. I found in documentation following statement:
Transformed trees do not preserve comments and formatting details when pretty-printed. Look into Scalafix if you need to implement fine-grained refactorings that preserve comments and formatting details. Does it mean i can use scalafix in programmatic way from my code to retain formatting before i will save the code back to the file? If yes could you point me to example? After reading scalafix docs i understood i need to recompile my project to apply formatting but that doesnt sound like a process i want to follow and i would prefer to simply apply formatting at stage of writing tree back to file. Thx in advance