Optimize matching to match on scalar values when possible (swiftlang#525)

- Adds new instructions for matching characters and scalars case insensitively
- Compiles ascii character matches into the faster scalar match instructions even in grapheme semantic mode
- Optimizes out unnecessary runtime grapheme boundary checks for all ascii strings
- Also includes fixes to scalar matching in grapheme semantic mode (swiftlang#565)

Author: rctcwyvrn

Sources/_StringProcessing/ByteCodeGen.swift

@@ -65,10 +65,14 @@ fileprivate extension Compiler.ByteCodeGen {
       emitDot()
 
     case let .char(c):
-      try emitCharacter(c)
+      emitCharacter(c)
 
     case let .scalar(s):
-      try emitScalar(s)
+      if options.semanticLevel == .graphemeCluster {
+        emitCharacter(Character(s))
+      } else {
+        emitMatchScalar(s)
+      }
 
     case let .assertion(kind):
       try emitAssertion(kind)
@@ -94,6 +98,34 @@ fileprivate extension Compiler.ByteCodeGen {
     }
   }
 
+  mutating func emitQuotedLiteral(_ s: String) {
+    guard options.semanticLevel == .graphemeCluster else {
+      for char in s {
+        for scalar in char.unicodeScalars {
+          emitMatchScalar(scalar)
+        }
+      }
+      return
+    }
+
+    // Fast path for eliding boundary checks for an all ascii quoted literal
+    if optimizationsEnabled && s.allSatisfy(\.isASCII) {
+      let lastIdx = s.unicodeScalars.indices.last!
+      for idx in s.unicodeScalars.indices {
+        let boundaryCheck = idx == lastIdx
+        let scalar = s.unicodeScalars[idx]
+        if options.isCaseInsensitive && scalar.properties.isCased {
+          builder.buildMatchScalarCaseInsensitive(scalar, boundaryCheck: boundaryCheck)
+        } else {
+          builder.buildMatchScalar(scalar, boundaryCheck: boundaryCheck)
+        }
+      }
+      return
+    }
+
+    for c in s { emitCharacter(c) }
+  }
+
   mutating func emitBackreference(
     _ ref: AST.Reference
   ) throws {
@@ -257,41 +289,47 @@ fileprivate extension Compiler.ByteCodeGen {
     }
   }
 
-  mutating func emitScalar(_ s: UnicodeScalar) throws {
-    // TODO: Native instruction buildMatchScalar(s)
-    if options.isCaseInsensitive {
-      // TODO: e.g. buildCaseInsensitiveMatchScalar(s)
-      builder.buildConsume(by: consumeScalar {
-        $0.properties.lowercaseMapping == s.properties.lowercaseMapping
-      })
+  mutating func emitMatchScalar(_ s: UnicodeScalar) {
+    assert(options.semanticLevel == .unicodeScalar)
+    if options.isCaseInsensitive && s.properties.isCased {
+      builder.buildMatchScalarCaseInsensitive(s, boundaryCheck: false)
     } else {
-      builder.buildConsume(by: consumeScalar {
-        $0 == s
-      })
+      builder.buildMatchScalar(s, boundaryCheck: false)
     }
   }
 
-  mutating func emitCharacter(_ c: Character) throws {
-    // Unicode scalar matches the specific scalars that comprise a character
+  mutating func emitCharacter(_ c: Character) {
+    // Unicode scalar mode matches the specific scalars that comprise a character
     if options.semanticLevel == .unicodeScalar {
       for scalar in c.unicodeScalars {
-        try emitScalar(scalar)
+        emitMatchScalar(scalar)
       }
       return
     }
 
     if options.isCaseInsensitive && c.isCased {
-      // TODO: buildCaseInsensitiveMatch(c) or buildMatch(c, caseInsensitive: true)
-      builder.buildConsume { input, bounds in
-        let inputChar = input[bounds.lowerBound].lowercased()
-        let matchChar = c.lowercased()
-        return inputChar == matchChar
-          ? input.index(after: bounds.lowerBound)
-          : nil
+      if optimizationsEnabled && c.isASCII {
+        // c.isCased ensures that c is not CR-LF,
+        // so we know that c is a single scalar
+        assert(c.unicodeScalars.count == 1)
+        builder.buildMatchScalarCaseInsensitive(
+          c.unicodeScalars.last!,
+          boundaryCheck: true)
+      } else {
+        builder.buildMatch(c, isCaseInsensitive: true)
       }
-    } else {
-      builder.buildMatch(c)
+      return
     }
+    
+    if optimizationsEnabled && c.isASCII {
+      let lastIdx = c.unicodeScalars.indices.last!
+      for idx in c.unicodeScalars.indices {
+        builder.buildMatchScalar(c.unicodeScalars[idx], boundaryCheck: idx == lastIdx)
+      }
+      return
+    }
+      
+    builder.buildMatch(c, isCaseInsensitive: false)
   }
 
   mutating func emitAny() {
@@ -741,11 +779,12 @@ fileprivate extension Compiler.ByteCodeGen {
     _ ccc: DSLTree.CustomCharacterClass
   ) throws {
     if let asciiBitset = ccc.asAsciiBitset(options),
-        options.semanticLevel == .graphemeCluster,
         optimizationsEnabled {
-      // future work: add a bit to .matchBitset to consume either a character
-      // or a scalar so we can have this optimization in scalar mode
-      builder.buildMatchAsciiBitset(asciiBitset)
+      if options.semanticLevel == .unicodeScalar {
+        builder.buildScalarMatchAsciiBitset(asciiBitset)
+      } else {
+        builder.buildMatchAsciiBitset(asciiBitset)
+      }
     } else {
       let consumer = try ccc.generateConsumer(options)
       builder.buildConsume(by: consumer)
@@ -822,45 +861,7 @@ fileprivate extension Compiler.ByteCodeGen {
       try emitAtom(a)
 
     case let .quotedLiteral(s):
-      if options.semanticLevel == .graphemeCluster {
-        if options.isCaseInsensitive {
-          // TODO: buildCaseInsensitiveMatchSequence(c) or alternative
-          builder.buildConsume { input, bounds in
-            var iterator = s.makeIterator()
-            var currentIndex = bounds.lowerBound
-            while let ch = iterator.next() {
-              guard currentIndex < bounds.upperBound,
-                    ch.lowercased() == input[currentIndex].lowercased()
-              else { return nil }
-              input.formIndex(after: &currentIndex)
-            }
-            return currentIndex
-          }
-        } else {
-          builder.buildMatchSequence(s)
-        }
-      } else {
-        builder.buildConsume {
-          [caseInsensitive = options.isCaseInsensitive] input, bounds in
-          // TODO: Case folding
-          var iterator = s.unicodeScalars.makeIterator()
-          var currentIndex = bounds.lowerBound
-          while let scalar = iterator.next() {
-            guard currentIndex < bounds.upperBound else { return nil }
-            if caseInsensitive {
-              if scalar.properties.lowercaseMapping != input.unicodeScalars[currentIndex].properties.lowercaseMapping {
-                return nil
-              }
-            } else {
-              if scalar != input.unicodeScalars[currentIndex] {
-                return nil
-              }
-            }
-            input.unicodeScalars.formIndex(after: &currentIndex)
-          }
-          return currentIndex
-        }
-      }
+      emitQuotedLiteral(s)
 
     case let .convertedRegexLiteral(n, _):
       return try emitNode(n)

Sources/_StringProcessing/ConsumerInterface.swift

@@ -11,6 +11,13 @@
 
 @_implementationOnly import _RegexParser
 
+extension Character {
+  var _singleScalarAsciiValue: UInt8? {
+    guard self != "\r\n" else { return nil }
+    return asciiValue
+  }
+}
+
 extension DSLTree.Node {
   /// Attempt to generate a consumer from this AST node
   ///
@@ -53,11 +60,50 @@ extension DSLTree._AST.Atom {
   }
 }
 
+extension Character {
+  func generateConsumer(
+    _ opts: MatchingOptions
+  ) throws -> MEProgram.ConsumeFunction? {
+    let isCaseInsensitive = opts.isCaseInsensitive
+    switch opts.semanticLevel {
+    case .graphemeCluster:
+      return { input, bounds in
+        let low = bounds.lowerBound
+        if isCaseInsensitive && isCased {
+          return input[low].lowercased() == lowercased()
+            ? input.index(after: low)
+            : nil
+        } else {
+          return input[low] == self
+            ? input.index(after: low)
+            : nil
+        }
+      }
+    case .unicodeScalar:
+      // TODO: This should only be reachable from character class emission, can
+      // we guarantee that? Otherwise we'd want a different matching behavior.
+      let consumers = unicodeScalars.map { s in consumeScalar {
+        isCaseInsensitive
+          ? $0.properties.lowercaseMapping == s.properties.lowercaseMapping
+          : $0 == s
+      }}
+      return { input, bounds in
+        for fn in consumers {
+          if let idx = fn(input, bounds) {
+            return idx
+          }
+        }
+        return nil
+      }
+    }
+  }
+}
+
 extension DSLTree.Atom {
   var singleScalarASCIIValue: UInt8? {
     switch self {
-    case let .char(c) where c != "\r\n":
-      return c.asciiValue
+    case let .char(c):
+      return c._singleScalarAsciiValue
     case let .scalar(s) where s.isASCII:
       return UInt8(ascii: s)
     case let .unconverted(atom):
@@ -72,44 +118,15 @@ extension DSLTree.Atom {
   func generateConsumer(
     _ opts: MatchingOptions
   ) throws -> MEProgram.ConsumeFunction? {
-    let isCaseInsensitive = opts.isCaseInsensitive
-    
     switch self {
     case let .char(c):
-      if opts.semanticLevel == .graphemeCluster {
-        return { input, bounds in
-          let low = bounds.lowerBound
-          if isCaseInsensitive && c.isCased {
-            return input[low].lowercased() == c.lowercased()
-              ? input.index(after: low)
-              : nil
-          } else {
-            return input[low] == c
-              ? input.index(after: low)
-              : nil
-          }
-        }
-      } else {
-        let consumers = c.unicodeScalars.map { s in consumeScalar {
-          isCaseInsensitive
-            ? $0.properties.lowercaseMapping == s.properties.lowercaseMapping
-            : $0 == s
-        }}
-        return { input, bounds in
-          for fn in consumers {
-            if let idx = fn(input, bounds) {
-              return idx
-            }
-          }
-          return nil
-        }
-      }
+      return try c.generateConsumer(opts)
+
     case let .scalar(s):
-      return consumeScalar {
-        isCaseInsensitive
-          ? $0.properties.lowercaseMapping == s.properties.lowercaseMapping
-          : $0 == s
-      }
+      // A scalar always matches the same as a single scalar character. This
+      // means it must match a whole grapheme in grapheme semantic mode, but
+      // can match a single scalar in scalar semantic mode.
+      return try Character(s).generateConsumer(opts)
 
     case .any:
       // FIXME: Should this be a total ordering?
@@ -230,16 +247,20 @@ extension AST.Atom {
   var singleScalar: UnicodeScalar? {
     switch kind {
     case .scalar(let s): return s.value
+    case .escaped(let e):
+      guard let s = e.scalarValue else { return nil }
+      return s
     default: return nil
     }
   }
 
   var singleScalarASCIIValue: UInt8? {
+    if let s = singleScalar, s.isASCII {
+       return UInt8(ascii: s)
+     }
     switch kind {
-    case let .char(c) where c != "\r\n":
-      return c.asciiValue
-    case let .scalar(s) where s.value.isASCII:
-      return UInt8(ascii: s.value)
+    case let .char(c):
+      return c._singleScalarAsciiValue
     default:
       return nil
     }

Sources/_StringProcessing/Engine/InstPayload.swift

@@ -147,6 +147,26 @@ extension Instruction.Payload {
   var string: StringRegister {
     interpret()
   }
+  
+  init(scalar: Unicode.Scalar) {
+    self.init(UInt64(scalar.value))
+  }
+  var scalar: Unicode.Scalar {
+    return Unicode.Scalar(_value: UInt32(self.rawValue))
+  }
+
+  init(scalar: Unicode.Scalar, caseInsensitive: Bool, boundaryCheck: Bool) {
+    let raw = UInt64(scalar.value)
+      + (caseInsensitive ? 1 << 55: 0)
+      + (boundaryCheck ? 1 << 54 : 0)
+    self.init(raw)
+  }
+  var scalarPayload: (Unicode.Scalar, caseInsensitive: Bool, boundaryCheck: Bool) {
+    let caseInsensitive = (self.rawValue >> 55) & 1 == 1
+    let boundaryCheck = (self.rawValue >> 54) & 1 == 1
+    let scalar = Unicode.Scalar(_value: UInt32(self.rawValue & 0xFFFF_FFFF))
+    return (scalar, caseInsensitive: caseInsensitive, boundaryCheck: boundaryCheck)
+  }
 
   init(sequence: SequenceRegister) {
     self.init(sequence)
@@ -190,18 +210,20 @@ extension Instruction.Payload {
     interpret()
   }
 
-  init(element: ElementRegister) {
-    self.init(element)
+  init(element: ElementRegister, isCaseInsensitive: Bool) {
+    self.init(isCaseInsensitive ? 1 : 0, element)
   }
-  var element: ElementRegister {
-    interpret()
+  var elementPayload: (isCaseInsensitive: Bool, ElementRegister) {
+    let pair: (UInt64, ElementRegister) = interpretPair()
+    return (isCaseInsensitive: pair.0 == 1, pair.1)
   }
 
-  init(bitset: AsciiBitsetRegister) {
-    self.init(bitset)
+  init(bitset: AsciiBitsetRegister, isScalar: Bool) {
+    self.init(isScalar ? 1 : 0, bitset)
   }
-  var bitset: AsciiBitsetRegister {
-    interpret()
+  var bitsetPayload: (isScalar: Bool, AsciiBitsetRegister) {
+    let pair: (UInt64, AsciiBitsetRegister) = interpretPair()
+    return (isScalar: pair.0 == 1, pair.1)
   }
 
   init(consumer: ConsumeFunctionRegister) {