Twine.h

Go to the documentation of this file.

//===-- Twine.h - Fast Temporary String Concatenation -----------*- C++ -*-===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//

#ifndef LLVM_ADT_TWINE_H
#define LLVM_ADT_TWINE_H

#include "llvm/ADT/StringRef.h"
#include "llvm/Support/DataTypes.h"
#include <cassert>
#include <string>
#include <iostream>

namespace llvm {
  template <typename T>
  class SmallVectorImpl;
  class StringRef;

  class Twine {
    enum NodeKind {
      NullKind,

      EmptyKind,

      TwineKind,

      CStringKind,

      StdStringKind,

      StringRefKind,

      CharKind,

      DecUIKind,

      DecIKind,

      DecULKind,

      DecLKind,

      DecULLKind,

      DecLLKind,

      UHexKind
    };

    union Child
    {
      const Twine *twine;
      const char *cString;
      const std::string *stdString;
      const StringRef *stringRef;
      char character;
      unsigned int decUI;
      int decI;
      const unsigned long *decUL;
      const long *decL;
      const unsigned long long *decULL;
      const long long *decLL;
      const uint64_t *uHex;
    };

  private:
    Child LHS;
    Child RHS;
    // enums stored as unsigned chars to save on space while some compilers
    // don't support specifying the backing type for an enum
    unsigned char LHSKind;
    unsigned char RHSKind;

  private:
    explicit Twine(NodeKind Kind)
      : LHSKind(Kind), RHSKind(EmptyKind) {
      assert(isNullary() && "Invalid kind!");
    }

    explicit Twine(const Twine &_LHS, const Twine &_RHS)
      : LHSKind(TwineKind), RHSKind(TwineKind) {
      LHS.twine = &_LHS;
      RHS.twine = &_RHS;
      assert(isValid() && "Invalid twine!");
    }

    explicit Twine(Child _LHS, NodeKind _LHSKind,
                   Child _RHS, NodeKind _RHSKind)
      : LHS(_LHS), RHS(_RHS), LHSKind(_LHSKind), RHSKind(_RHSKind) {
      assert(isValid() && "Invalid twine!");
    }

    bool isNull() const {
      return getLHSKind() == NullKind;
    }

    bool isEmpty() const {
      return getLHSKind() == EmptyKind;
    }

    bool isNullary() const {
      return isNull() || isEmpty();
    }

    bool isUnary() const {
      return getRHSKind() == EmptyKind && !isNullary();
    }

    bool isBinary() const {
      return getLHSKind() != NullKind && getRHSKind() != EmptyKind;
    }

    bool isValid() const {
      // Nullary twines always have Empty on the RHS.
      if (isNullary() && getRHSKind() != EmptyKind)
        return false;

      // Null should never appear on the RHS.
      if (getRHSKind() == NullKind)
        return false;

      // The RHS cannot be non-empty if the LHS is empty.
      if (getRHSKind() != EmptyKind && getLHSKind() == EmptyKind)
        return false;

      // A twine child should always be binary.
      if (getLHSKind() == TwineKind &&
          !LHS.twine->isBinary())
        return false;
      if (getRHSKind() == TwineKind &&
          !RHS.twine->isBinary())
        return false;

      return true;
    }

    NodeKind getLHSKind() const { return (NodeKind) LHSKind; }

    NodeKind getRHSKind() const { return (NodeKind) RHSKind; }

    void printOneChild(std::ostream &OS, Child Ptr, NodeKind Kind) const;

    void printOneChildRepr(std::ostream &OS, Child Ptr,
                           NodeKind Kind) const;

  public:

    /*implicit*/ Twine() : LHSKind(EmptyKind), RHSKind(EmptyKind) {
      assert(isValid() && "Invalid twine!");
    }

    /*implicit*/ Twine(const char *Str)
      : RHSKind(EmptyKind) {
      if (Str[0] != '\0') {
        LHS.cString = Str;
        LHSKind = CStringKind;
      } else
        LHSKind = EmptyKind;

      assert(isValid() && "Invalid twine!");
    }

    /*implicit*/ Twine(const std::string &Str)
      : LHSKind(StdStringKind), RHSKind(EmptyKind) {
      LHS.stdString = &Str;
      assert(isValid() && "Invalid twine!");
    }

    /*implicit*/ Twine(const StringRef &Str)
      : LHSKind(StringRefKind), RHSKind(EmptyKind) {
      LHS.stringRef = &Str;
      assert(isValid() && "Invalid twine!");
    }

    explicit Twine(char Val)
      : LHSKind(CharKind), RHSKind(EmptyKind) {
      LHS.character = Val;
    }

    explicit Twine(signed char Val)
      : LHSKind(CharKind), RHSKind(EmptyKind) {
      LHS.character = static_cast<char>(Val);
    }

    explicit Twine(unsigned char Val)
      : LHSKind(CharKind), RHSKind(EmptyKind) {
      LHS.character = static_cast<char>(Val);
    }

    explicit Twine(unsigned Val)
      : LHSKind(DecUIKind), RHSKind(EmptyKind) {
      LHS.decUI = Val;
    }

    explicit Twine(int Val)
      : LHSKind(DecIKind), RHSKind(EmptyKind) {
      LHS.decI = Val;
    }

    explicit Twine(const unsigned long &Val)
      : LHSKind(DecULKind), RHSKind(EmptyKind) {
      LHS.decUL = &Val;
    }

    explicit Twine(const long &Val)
      : LHSKind(DecLKind), RHSKind(EmptyKind) {
      LHS.decL = &Val;
    }

    explicit Twine(const unsigned long long &Val)
      : LHSKind(DecULLKind), RHSKind(EmptyKind) {
      LHS.decULL = &Val;
    }

    explicit Twine(const long long &Val)
      : LHSKind(DecLLKind), RHSKind(EmptyKind) {
      LHS.decLL = &Val;
    }

    // FIXME: Unfortunately, to make sure this is as efficient as possible we
    // need extra binary constructors from particular types. We can't rely on
    // the compiler to be smart enough to fold operator+()/concat() down to the
    // right thing. Yet.

    /*implicit*/ Twine(const char *_LHS, const StringRef &_RHS)
      : LHSKind(CStringKind), RHSKind(StringRefKind) {
      LHS.cString = _LHS;
      RHS.stringRef = &_RHS;
      assert(isValid() && "Invalid twine!");
    }

    /*implicit*/ Twine(const StringRef &_LHS, const char *_RHS)
      : LHSKind(StringRefKind), RHSKind(CStringKind) {
      LHS.stringRef = &_LHS;
      RHS.cString = _RHS;
      assert(isValid() && "Invalid twine!");
    }

    static Twine createNull() {
      return Twine(NullKind);
    }


    // Construct a twine to print \arg Val as an unsigned hexadecimal integer.
    static Twine utohexstr(const uint64_t &Val) {
      Child LHS, RHS;
      LHS.uHex = &Val;
      RHS.twine = 0;
      return Twine(LHS, UHexKind, RHS, EmptyKind);
    }


    bool isTriviallyEmpty() const {
      return isNullary();
    }

    bool isSingleStringRef() const {
      if (getRHSKind() != EmptyKind) return false;

      switch (getLHSKind()) {
      case EmptyKind:
      case CStringKind:
      case StdStringKind:
      case StringRefKind:
        return true;
      default:
        return false;
      }
    }


    Twine concat(const Twine &Suffix) const;


    std::string str() const;

    void toVector(SmallVectorImpl<char> &Out) const;

    StringRef getSingleStringRef() const {
      assert(isSingleStringRef() &&"This cannot be had as a single stringref!");
      switch (getLHSKind()) {
      default: assert(0 && "Out of sync with isSingleStringRef");
      case EmptyKind:      return StringRef();
      case CStringKind:    return StringRef(LHS.cString);
      case StdStringKind:  return StringRef(*LHS.stdString);
      case StringRefKind:  return *LHS.stringRef;
      }
    }

    StringRef toStringRef(SmallVectorImpl<char> &Out) const;

    StringRef toNullTerminatedStringRef(SmallVectorImpl<char> &Out) const;

    void print(std::ostream &OS) const;

    void dump() const;

    void printRepr(std::ostream &OS) const;

    void dumpRepr() const;

  };


  inline Twine Twine::concat(const Twine &Suffix) const {
    // Concatenation with null is null.
    if (isNull() || Suffix.isNull())
      return Twine(NullKind);

    // Concatenation with empty yields the other side.
    if (isEmpty())
      return Suffix;
    if (Suffix.isEmpty())
      return *this;

    // Otherwise we need to create a new node, taking care to fold in unary
    // twines.
    Child NewLHS, NewRHS;
    NewLHS.twine = this;
    NewRHS.twine = &Suffix;
    NodeKind NewLHSKind = TwineKind, NewRHSKind = TwineKind;
    if (isUnary()) {
      NewLHS = LHS;
      NewLHSKind = getLHSKind();
    }
    if (Suffix.isUnary()) {
      NewRHS = Suffix.LHS;
      NewRHSKind = Suffix.getLHSKind();
    }

    return Twine(NewLHS, NewLHSKind, NewRHS, NewRHSKind);
  }

  inline Twine operator+(const Twine &LHS, const Twine &RHS) {
    return LHS.concat(RHS);
  }


  inline Twine operator+(const char *LHS, const StringRef &RHS) {
    return Twine(LHS, RHS);
  }


  inline Twine operator+(const StringRef &LHS, const char *RHS) {
    return Twine(LHS, RHS);
  }

inline std::ostream &operator<<(std::ostream &OS, const Twine &RHS) {
    RHS.print(OS);
    return OS;
  }

}

#endif

---