Backport "Add UTF8 abstraction in the TASTy format" to LTS

compiler/src/dotty/tools/dotc/core/tasty/CommentPickler.scala

@@ -22,11 +22,8 @@ object CommentPickler:
 
     def pickleComment(addr: Addr, comment: Comment): Unit =
       if addr != NoAddr then
-        val bytes = comment.raw.getBytes(StandardCharsets.UTF_8).nn
-        val length = bytes.length
         buf.writeAddr(addr)
-        buf.writeNat(length)
-        buf.writeBytes(bytes, length)
+        buf.writeUtf8(comment.raw)
         buf.writeLongInt(comment.span.coords)
 
     def traverse(x: Any): Unit = x match

compiler/src/dotty/tools/dotc/core/tasty/CommentUnpickler.scala

@@ -19,13 +19,9 @@ class CommentUnpickler(reader: TastyReader) {
     val comments = new HashMap[Addr, Comment]
     while (!isAtEnd) {
       val addr = readAddr()
-      val length = readNat()
-      if (length > 0) {
-        val bytes = readBytes(length)
-        val position = new Span(readLongInt())
-        val rawComment = new String(bytes, StandardCharsets.UTF_8)
-        comments(addr) = Comment(position, rawComment)
-      }
+      val rawComment = readUtf8()
+      val position = new Span(readLongInt())
+      comments(addr) = Comment(position, rawComment)
     }
     comments
   }

compiler/src/dotty/tools/dotc/core/tasty/TastyPickler.scala

@@ -13,14 +13,8 @@ import collection.mutable
 import core.Symbols.ClassSymbol
 import Decorators._
 
-object TastyPickler {
-
-  private val versionStringBytes = {
-    val compilerString = s"Scala ${config.Properties.simpleVersionString}"
-    compilerString.getBytes(java.nio.charset.StandardCharsets.UTF_8)
-  }
-
-}
+object TastyPickler:
+  private val versionString = s"Scala ${config.Properties.simpleVersionString}"
 
 class TastyPickler(val rootCls: ClassSymbol) {
 
@@ -48,13 +42,12 @@ class TastyPickler(val rootCls: ClassSymbol) {
     val uuidHi: Long = otherSectionHashes.fold(0L)(_ ^ _)
 
     val headerBuffer = {
-      val buf = new TastyBuffer(header.length + TastyPickler.versionStringBytes.length + 32)
+      val buf = new TastyBuffer(header.length + TastyPickler.versionString.length + 32)
       for (ch <- header) buf.writeByte(ch.toByte)
       buf.writeNat(MajorVersion)
       buf.writeNat(MinorVersion)
       buf.writeNat(ExperimentalVersion)
-      buf.writeNat(TastyPickler.versionStringBytes.length)
-      buf.writeBytes(TastyPickler.versionStringBytes, TastyPickler.versionStringBytes.length)
+      buf.writeUtf8(TastyPickler.versionString)
       buf.writeUncompressedLong(uuidLow)
       buf.writeUncompressedLong(uuidHi)
       buf

tasty/src/dotty/tools/tasty/TastyBuffer.scala

@@ -1,6 +1,7 @@
 package dotty.tools.tasty
 
 import util.Util.dble
+import java.nio.charset.StandardCharsets
 
 object TastyBuffer {
 
@@ -115,6 +116,16 @@ class TastyBuffer(initialSize: Int) {
     writeBytes(bytes, 8)
   }
 
+  /** Write a UTF8 string encoded as `Nat UTF8-CodePoint*`,
+   *  where the `Nat` is the length of the code-points bytes.
+   */
+  def writeUtf8(x: String): Unit = {
+    val bytes = x.getBytes(StandardCharsets.UTF_8)
+    val length = bytes.length
+    writeNat(length)
+    writeBytes(bytes, length)
+  }
+
   // -- Address handling --------------------------------------------
 
   /** Write natural number `x` right-adjusted in a field of `width` bytes

tasty/src/dotty/tools/tasty/TastyFormat.scala

@@ -16,6 +16,7 @@ Micro-syntax:
   Nat           = LongInt                 -- non-negative value, fits in an Int without overflow
   Digit         = 0 | ... | 127
   StopDigit     = 128 | ... | 255         -- value = digit - 128
+  Utf8          = Nat UTF8-CodePoint*
 ```
 
 Macro-format:
@@ -24,12 +25,12 @@ Macro-format:
                   nameTable_Length Name* Section*
   Header        = 0x5CA1AB1F
   UUID          = Byte*16                 -- random UUID
-  VersionString = Length UTF8-CodePoint*  -- string that represents the compiler that produced the TASTy
+  VersionString = Utf8                    -- string that represents the compiler that produced the TASTy
 
   Section       = NameRef Length Bytes
   Length        = Nat                     -- length of rest of entry in bytes
 
-  Name          = UTF8              Length UTF8-CodePoint*
+  Name          = UTF8              Utf8
                   QUALIFIED         Length qualified_NameRef selector_NameRef               -- A.B
                   EXPANDED          Length qualified_NameRef selector_NameRef               -- A$$B, semantically a NameKinds.ExpandedName
                   EXPANDPREFIX      Length qualified_NameRef selector_NameRef               -- A$B, prefix of expanded name, see NamedKinds.ExpandPrefixName
@@ -263,7 +264,7 @@ All elements of a position section are serialized as Ints
 
 Standard Section: "Comments" Comment*
 ```none
-  Comment       = Length Bytes LongInt      // Raw comment's bytes encoded as UTF-8, followed by the comment's coordinates.
+  Comment       = Utf8 LongInt              // Raw comment's bytes encoded as UTF-8, followed by the comment's coordinates.
 ```
 **************************************************************************************/
 

tasty/src/dotty/tools/tasty/TastyReader.scala

@@ -3,6 +3,7 @@ package dotty.tools.tasty
 import collection.mutable
 
 import TastyBuffer._
+import java.nio.charset.StandardCharsets
 
 /** A byte array buffer that can be filled with bytes or natural numbers in TASTY format,
  *  and that supports reading and patching addresses represented as natural numbers.
@@ -104,6 +105,15 @@ class TastyReader(val bytes: Array[Byte], start: Int, end: Int, val base: Int =
     x
   }
 
+  /** Read a UTF8 string encoded as `Nat UTF8-CodePoint*`,
+   *  where the `Nat` is the length of the code-points bytes.
+   */
+  def readUtf8(): String = {
+    val length = readNat()
+    if (length == 0) ""
+    else new String(readBytes(length), StandardCharsets.UTF_8)
+  }
+
   /** Read a natural number and return as a NameRef */
   def readNameRef(): NameRef = NameRef(readNat())