First, the scalaVersion in build.sbt was changed to 2.13 so that the Scala3-migrate tool (details here) could be used.
Prior to using the migration tool, I realised that org.apache.commons.lang3.StringEscapeUtils was depreciated and replaced it with org.apache.commons.text.StringEscapeUtils in Lexer.scala. I also changed the sbtVersion to 1.8.2 from 1.3.12 as the migration tool to be used required the version to be at least 1.5.
There was a warning that stated that the flag -Ypartial-unification was not able to be parsed. This is because from Scala 2.13 onwards, the flag is enabled by default. Therefore, I removed the flag from build.sbt.
There was also another warning that said that at Line 37 of Parser.scala, the match may not be exhaustive. I found that removing the "if" clause in Line 39 resolved this as it generalised the failure condition.
I installed the migration tool afterwards by adding it as a plugin in project/plugins.sbt.
According to the migration guide, I first ran migrateDependencies scalangj and discovered that the following libraries were not compatible with Scala 3:
| Library | Old Version | New Version |
|---|---|---|
| scala-parser-combinators | 1.1.2 | 2.3.0 |
| scalactic | 3.0.8 | 3.2.9 |
| scalatest | 3.0.8 | 3.2.9 |
| paiges-core | 0.3.0 | 0.4.2 |
After changing the libraries to the new versions, I used migrateScalacOptions scalangj. The migration tool removed the -Yrangepos flag as it is not available in Scala 3.
I received an error in Lexer.scala after this, and it was due to this line
import scala.collection.convert.DecorateAsJavaAs I could not find any methods that used this import, I removed it.
When I tried to run migrateSyntax scalangj, I got multiple errors in the test files. These stemmed from the change in the scalatest library version. I changed the line
import org.scalatest.{FunSuite, Matchers}to
import org.scalatest.{funsuite, matchers}For all the test classes, I changed the FunSuite to funsuite.AnyFunSuite and Matchers to matchers.should.Matchers. After these changes, I was able to run migrateSyntax scalangj successfully.
However, I was not able to run migrateTypes scalangj as the tool stated that the project could not be compiled. After changing the scalaVersion in build.sbt to 3.2.2, I discovered that there were multiple instances in Parser.scala where the compiler was unable to interpret the types, which mostly occurred in lambda parameters. One of the errors (in Line 868) occurred due to the case classes TypeMethodCall and ClassMethodCall not having the method apply being called. Warnings were also raised over other instanced where apply was not being called. Instead of explicitly calling apply, I instead wrote lambda functions as advised by Prof Kenny.
There was also a warning in Line 35 where the case Success(_, _) would not be caught. This was fixed by removing the if !rest.atEnd portion of the second case.
Turns out it has failed already in the original version. By writing new test cases for the lexer, it seems that it fails for any token that is made up of alphabets. After discussing my observations with Prof Kenny, he pointed out that it could be a problem with not taking whitespaces into account.
He suggested splitting the tokens into two categories.
In Category 1, the tokens need to have a trailing whitespace if it is succeeded by another Category 1 token (e.g. for the New token and an Identifier token "var1", it should be "new var1" instead of "newvar1").
In Category 2, the tokens do not need to have a trailing whitespace. e.g. parenthesis, variable assignment operator
However, I chose to do a slightly different method, in which a greedy form of lexing was done. I modified the function that lexed Identifier tokens. This is such that it lexed any token that starts with a JavaLetter such as the New token ("new") or Boolean tokens ("true"/"false") in addition to Identifier tokens, which was done through pattern matching.
Note: This version is still in Scala 2.12