Parsing using the vector API

.github/workflows/ci.yml

@@ -19,11 +19,11 @@ jobs:
         with:
           fetch-depth: 0
 
-      - name: Install JDK 11
+      - name: Install JDK 21
         uses: actions/setup-java@v4
         with:
           distribution: zulu
-          java-version: '11'
+          java-version: '21'
           java-package: jdk
 
       - name: Install SBT

.github/workflows/release.yml

@@ -11,11 +11,11 @@ jobs:
       - name: Checkout
         uses: actions/checkout@v4
 
-      - name: Install JDK 11
+      - name: Install JDK 21
         uses: actions/setup-java@v4
         with:
           distribution: zulu
-          java-version: '11'
+          java-version: '21'
           java-package: jdk
 
       - name: Install SBT

.jvmopts

@@ -0,0 +1,2 @@
+--enable-preview
+--add-modules=jdk.incubator.vector

.scalafmt.conf

@@ -26,7 +26,7 @@ rewrite.rules = [ Imports ]
 rewrite.imports.expand = true
 rewrite.imports.sort = ascii
 rewrite.imports.groups = [
-  ["(?!javax?\\.|scala\\.).+"],
+  ["(?!javax?\\.|jdk\\.|scala\\.).+"],
 ]
 rewrite.trailingCommas.style = always
 spaces.neverAroundInfixTypes = [ "##" ]

LICENSE

@@ -1,6 +1,6 @@
 The MIT License (MIT)
 
-Copyright (c) 2022-2022 Sam Guymer
+Copyright (c) 2022-2025 Sam Guymer
 
 Permission is hereby granted, free of charge, to any person obtaining a copy of
 this software and associated documentation files (the "Software"), to deal in

README.md

@@ -14,9 +14,8 @@ case class Test(
   int: Int,
   bool: Boolean,
   optInt: Option[Int],
-)
+) derives CsvRecordDecoder
 object Test {
-  implicit val decoder: CsvRecordDecoder[Test] = CsvRecordDecoder.derive
   val header = ::("str", List("int", "bool", "opt_int"))
   val csvHeader = CsvHeader.create(header)(decoder)
 }

benchmark/README.md

@@ -1,26 +1,30 @@
 
 `benchmark/Jmh/run -i 10 -wi 5 -f 1 -t 2 ceesvee.benchmark.ParserBenchmark`
 
+AMD Ryzen 9 9950X
 ```
 # JMH version: 1.37
 # VM version: JDK 25.0.1, OpenJDK 64-Bit Server VM, 25.0.1
-Benchmark                  Mode  Cnt    Score   Error  Units
-ParserBenchmark.ceesvee    avgt   10  261.357 ± 1.787  us/op
-ParserBenchmark.scalaCsv   avgt   10  741.778 ± 6.433  us/op
-ParserBenchmark.univocity  avgt   10  200.482 ± 2.715  us/op
+Benchmark                      Mode  Cnt    Score   Error  Units
+ParserBenchmark.ceesvee        avgt   10  263.230 ± 0.679  us/op
+ParserBenchmark.ceesveeVector  avgt   10  134.205 ± 0.302  us/op
+ParserBenchmark.scalaCsv       avgt   10  748.232 ± 2.016  us/op
+ParserBenchmark.univocity      avgt   10  198.765 ± 0.982  us/op
 ```
 
 ```
 # JMH version: 1.37
 #  VM version: JDK 25, OpenJDK 64-Bit Server VM, 25+37-jvmci-b01
-Benchmark                  Mode  Cnt    Score   Error  Units
-ParserBenchmark.ceesvee    avgt   10  197.994 ± 2.344  us/op
-ParserBenchmark.scalaCsv   avgt   10  776.080 ± 1.457  us/op
-ParserBenchmark.univocity  avgt   10  208.226 ± 2.501  us/op
+Benchmark                      Mode  Cnt     Score    Error  Units
+ParserBenchmark.ceesvee        avgt   10   187.441 ±  1.345  us/op
+ParserBenchmark.ceesveeVector  avgt   10  1484.755 ± 14.298  us/op
+ParserBenchmark.scalaCsv       avgt   10   780.945 ±  2.340  us/op
+ParserBenchmark.univocity      avgt   10   204.178 ±  1.702  us/op
 ```
 
 `benchmark/Jmh/run -i 10 -wi 5 -f 1 -t 2 ceesvee.benchmark.DecoderBenchmark`
 
+AMD Ryzen 9 9950X
 ```
 # JMH version: 1.37
 # VM version: JDK 25.0.1, OpenJDK 64-Bit Server VM, 25.0.1

benchmark/src/main/scala/ceesvee/benchmark/ParserBenchmark.scala

@@ -11,16 +11,23 @@ import java.util.concurrent.TimeUnit
 @State(Scope.Thread)
 @BenchmarkMode(Array(Mode.AverageTime))
 @OutputTimeUnit(TimeUnit.MICROSECONDS)
+@Fork(
+  jvmArgs = Array(
+    "--enable-preview",
+    "--add-modules=jdk.incubator.vector",
+  ),
+)
 class ParserBenchmark {
 
   private def line(i: Int) = List("basic string", " \"quoted \nstring\" ", i.toString, "456.789", "true").mkString(",")
 
+  private val charset = StandardCharsets.UTF_8
   private val lines = (1 to 1000).map(line(_)).mkString("\n")
-  private def linesChunked = lines.grouped(8192)
+  private val linesBytes = lines.getBytes(charset)
   private def linesReader = {
     val streams = new java.util.ArrayList[ByteArrayInputStream]()
-    linesChunked.foreach { str =>
-      streams.add(new ByteArrayInputStream(str.getBytes(StandardCharsets.UTF_8)))
+    linesBytes.grouped(8192).foreach { bytes =>
+      streams.add(new ByteArrayInputStream(bytes))
     }
     val is = new SequenceInputStream(java.util.Collections.enumeration(streams))
     new InputStreamReader(is)
@@ -32,7 +39,12 @@ class ParserBenchmark {
 
   @Benchmark
   def ceesvee: List[List[String]] = {
-    _root_.ceesvee.CsvParser.parse[List](linesChunked, ceesveeOptions).toList
+    _root_.ceesvee.CsvParser.parse[List](lines.grouped(8192), ceesveeOptions).toList
+  }
+
+  @Benchmark
+  def ceesveeVector: List[List[String]] = {
+    _root_.ceesvee.CsvParserVector.parse[List](linesBytes.grouped(8192), charset, ceesveeOptions).toList
   }
 
   @Benchmark

build.sbt

@@ -26,7 +26,7 @@ lazy val commonSettings = Seq(
     "-deprecation",
     "-encoding", "UTF-8",
     "-feature",
-    "-release", "11",
+    "-release", "21",
     "-unchecked",
   ),
   scalacOptions ++= (CrossVersion.partialVersion(scalaVersion.value) match {
@@ -47,6 +47,7 @@ lazy val commonSettings = Seq(
     case Some((2, _)) => Seq(
       "-Vimplicits",
       "-Vtype-diffs",
+      "-Wconf:cat=scala3-migration:silent",
       "-Wdead-code",
       "-Wextra-implicit",
       "-Wnonunit-statement",
@@ -59,15 +60,12 @@ lazy val commonSettings = Seq(
       "-Xlint:_,-byname-implicit", // exclude byname-implicit https://github.com/scala/bug/issues/12072
     )
     case _ => Seq(
+      "-Wconf:name=PatternMatchExhaustivity:error",
       "-Wnonunit-statement",
       "-Wunused:all",
       "-Wvalue-discard",
     )
   }),
-  Test / scalacOptions ++= (CrossVersion.partialVersion(scalaVersion.value) match {
-    case Some((2, _)) => Seq("-Wconf:cat=scala3-migration:silent")
-    case _ => Seq.empty
-  }),
 
   Compile / console / scalacOptions ~= filterScalacConsoleOpts,
   Test / console / scalacOptions ~= filterScalacConsoleOpts,

docs/architecture.md

@@ -0,0 +1,9 @@
+## Architecture
+
+### Module Structure
+
+- **core** - Main CSV parsing, encoding/decoding logic. Only dependency is an optional one on [cats](https://github.com/typelevel/cats)
+- **fs2** - Integration with [fs2](https://github.com/typelevel/fs2) streams
+- **zio** - Integration with [ZIO](https://github.com/zio/zio) streams
+- **benchmark** - JMH performance benchmarks comparing against other CSV libraries
+- **tests** - Integration tests with real-world CSV files

modules/core/src/main/scala/ceesvee/CsvParser.scala

@@ -331,7 +331,7 @@ object CsvParser {
     fields.result()
   }
 
-  private def trimString(options: Options, str: String) = {
+  private[ceesvee] def trimString(options: Options, str: String) = {
     // always ignore whitespace around a quoted cell
     val trimmed = Options.Trim.True.strip(str)