Merge pull request #190 from unisonweb/topic/varlong-arrays

use VarLong for lengths and backref Pos

parser-typechecker/src/Unison/Codecs.hs

@@ -5,6 +5,10 @@ module Unison.Codecs where
 import Data.Text (Text)
 import           Control.Arrow (second)
 import           Control.Monad.State
+import           Data.Bits (Bits)
+import qualified Data.Bytes.Serial as BS
+import           Data.Bytes.Signed (Unsigned)
+import           Data.Bytes.VarInt (VarInt(..))
 import qualified Data.ByteString as B
 import           Data.ByteString.Builder (doubleBE, int64BE, toLazyByteString)
 import qualified Data.ByteString.Lazy as BL
@@ -257,10 +261,12 @@ serializeCase1 (MatchCase p guard body) = do
   pure $ MatchCase p posg posb
 
 putBackref :: MonadPut m => Pos -> m ()
-putBackref = putWord64be
+putBackref = BS.serialize . VarInt
 
-putLength :: (MonadPut m, Integral n) => n -> m ()
-putLength = putWord64be . fromIntegral
+putLength :: (MonadPut m, Integral n, Integral (Unsigned n),
+                          Bits n,     Bits (Unsigned n))
+          => n -> m ()
+putLength = BS.serialize . VarInt
 
 serializeMaybe :: (MonadPut m) => (a -> m ()) -> Maybe a -> m ()
 serializeMaybe f b = case b of
@@ -270,7 +276,7 @@ serializeMaybe f b = case b of
 lengthEncode :: MonadPut m => Text -> m ()
 lengthEncode text = do
   let bs = encodeUtf8 text
-  putWord32be . fromIntegral $ B.length bs
+  putLength $ B.length bs
   putByteString bs
 
 serializeFoldable :: (MonadPut m, Foldable f) => (a -> m ()) -> f a -> m ()
@@ -286,7 +292,7 @@ serializeReference ref = case ref of
   Derived hash -> do
     putWord8 1
     let bs = Hash.toBytes hash
-    putWord32be . fromIntegral $ B.length bs
+    putLength $ B.length bs
     putByteString bs
 
 serializeConstructorArities :: MonadPut m => Reference -> [Int] -> m ()
@@ -302,4 +308,4 @@ serializeFile (UnisonFile dataDecls effectDecls body) = do
   serializeFoldable (uncurry serializeConstructorArities) effectDecls'
   pos <- serializeTerm body
   putWord8 0
-  putWord64be pos
+  putBackref pos

runtime-jvm/main/src/main/scala/util/Sink.scala

@@ -1,6 +1,7 @@
 package org.unisonweb.util
 
 import java.nio.{ByteBuffer,BufferOverflowException}
+import java.lang.Long.{compareUnsigned}
 import Text.Text
 
 /**
@@ -14,15 +15,34 @@ trait Sink {
   def putInt(n: Int): Unit
   def putLong(n: Long): Unit
 
-  // todo: the UTF-8 of Long encoding, use a single byte if possible
-  def putVarLong(n: Long): Unit =
-    putLong(n)
+  /**
+    * Uses the little-endian variable length encoding of unsigned integers:
+    * https://developers.google.com/protocol-buffers/docs/encoding#varints
+    */
+  def putVarLong(n: Long): Unit = {
+    val lsb = n.toShort & 0xff
+    if (compareUnsigned(n, 0x80) < 0) putByte(lsb.toByte)
+    else {
+      putByte((lsb | 0x80).toByte)
+      putVarLong(n >>> 7)
+    }
+  }
+
+  /**
+    * Uses the zigzag encoding for variable-length signed numbers, described at:
+    * https://developers.google.com/protocol-buffers/docs/encoding#signed-integers
+    * https://github.com/google/protobuf/blob/0400cca/java/core/src/main/java/com/google/protobuf/CodedOutputStream.java#L949-L952
+    */
+  def putVarSignedLong(n: Long): Unit = {
+    putVarLong((n << 1) ^ (n >> 63))
+  }
+
   def putDouble(n: Double): Unit
   def putString(s: String): Unit
   def putText(txt: Text): Unit
   def position: Long
   def putFramed(bs: Array[Byte]): Unit = {
-    putInt(bs.length)
+    putVarLong(bs.length)
     put(bs)
   }
   def putFramedSeq[A](seq: Seq[A])(f: (Sink,A) => Unit): Unit =

runtime-jvm/main/src/main/scala/util/Source.scala

@@ -12,17 +12,40 @@ import scala.reflect.ClassTag
  * The cursor position can be accessed via the `position` method.
  */
 trait Source { self =>
+  // todo: use a representation that supports 64-bit lengths, unlike Array
   def get(n: Int): Array[Byte]
   def getBoolean: Boolean = getByte != 0
   def getByte: Byte
   def getInt: Int
   def getLong: Long
-  // todo: The UTF-8 of Long encodings, uses a single byte where possible
-  def getVarLong: Long = getLong
+
+  /**
+    * Uses the little-endian variable length encoding of unsigned integers:
+    * https://developers.google.com/protocol-buffers/docs/encoding#varints
+    */
+  def getVarLong: Long = {
+    val b = getByte
+    if ((b & 0x80) == 0) b
+    else (getVarLong << 7) | (b & 0x7f)
+  }
+
+  /**
+    * Uses the zigzag encoding for variable-length signed numbers, described at:
+    * https://developers.google.com/protocol-buffers/docs/encoding#signed-integers
+    * https://github.com/google/protobuf/blob/0400cca/java/core/src/main/java/com/google/protobuf/CodedOutputStream.java#L949-L952
+    */
+  def getVarSignedLong: Long = {
+    val n = getVarLong
+    (n >>> 1) ^ -(n & 1)
+  }
+
   def getDouble: Double
   def position: Long
-  def getFramed: Array[Byte] = get(getInt)
 
+  // todo: use a representation that supports 64-bit lengths, unlike Array
+  def getFramed: Array[Byte] = get(getVarLong.toInt)
+
+  // todo: use a representation that supports 64-bit lengths, unlike String
   final def getString: String = {
     val bytes = getFramed
     new String(bytes, java.nio.charset.StandardCharsets.UTF_8)