1//===-- ConvertUTFWrapper.cpp - Wrap ConvertUTF.h with clang data types -----===
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8
9#include "llvm/ADT/ArrayRef.h"
10#include "llvm/ADT/StringRef.h"
11#include "llvm/Support/ConvertUTF.h"
12#include "llvm/Support/ErrorHandling.h"
13#include "llvm/Support/SwapByteOrder.h"
14#include <string>
15#include <vector>
16
17namespace llvm {
18
19bool ConvertUTF8toWide(unsigned WideCharWidth, llvm::StringRef Source,
20                       char *&ResultPtr, const UTF8 *&ErrorPtr) {
21  assert(WideCharWidth == 1 || WideCharWidth == 2 || WideCharWidth == 4);
22  ConversionResult result = conversionOK;
23  // Copy the character span over.
24  if (WideCharWidth == 1) {
25    const UTF8 *Pos = reinterpret_cast<const UTF8*>(Source.begin());
26    if (!isLegalUTF8String(&Pos, reinterpret_cast<const UTF8*>(Source.end()))) {
27      result = sourceIllegal;
28      ErrorPtr = Pos;
29    } else {
30      memcpy(ResultPtr, Source.data(), Source.size());
31      ResultPtr += Source.size();
32    }
33  } else if (WideCharWidth == 2) {
34    const UTF8 *sourceStart = (const UTF8*)Source.data();
35    // FIXME: Make the type of the result buffer correct instead of
36    // using reinterpret_cast.
37    UTF16 *targetStart = reinterpret_cast<UTF16*>(ResultPtr);
38    ConversionFlags flags = strictConversion;
39    result = ConvertUTF8toUTF16(
40        &sourceStart, sourceStart + Source.size(),
41        &targetStart, targetStart + Source.size(), flags);
42    if (result == conversionOK)
43      ResultPtr = reinterpret_cast<char*>(targetStart);
44    else
45      ErrorPtr = sourceStart;
46  } else if (WideCharWidth == 4) {
47    const UTF8 *sourceStart = (const UTF8*)Source.data();
48    // FIXME: Make the type of the result buffer correct instead of
49    // using reinterpret_cast.
50    UTF32 *targetStart = reinterpret_cast<UTF32*>(ResultPtr);
51    ConversionFlags flags = strictConversion;
52    result = ConvertUTF8toUTF32(
53        &sourceStart, sourceStart + Source.size(),
54        &targetStart, targetStart + Source.size(), flags);
55    if (result == conversionOK)
56      ResultPtr = reinterpret_cast<char*>(targetStart);
57    else
58      ErrorPtr = sourceStart;
59  }
60  assert((result != targetExhausted)
61         && "ConvertUTF8toUTFXX exhausted target buffer");
62  return result == conversionOK;
63}
64
65bool ConvertCodePointToUTF8(unsigned Source, char *&ResultPtr) {
66  const UTF32 *SourceStart = &Source;
67  const UTF32 *SourceEnd = SourceStart + 1;
68  UTF8 *TargetStart = reinterpret_cast<UTF8 *>(ResultPtr);
69  UTF8 *TargetEnd = TargetStart + 4;
70  ConversionResult CR = ConvertUTF32toUTF8(&SourceStart, SourceEnd,
71                                           &TargetStart, TargetEnd,
72                                           strictConversion);
73  if (CR != conversionOK)
74    return false;
75
76  ResultPtr = reinterpret_cast<char*>(TargetStart);
77  return true;
78}
79
80bool hasUTF16ByteOrderMark(ArrayRef<char> S) {
81  return (S.size() >= 2 &&
82          ((S[0] == '\xff' && S[1] == '\xfe') ||
83           (S[0] == '\xfe' && S[1] == '\xff')));
84}
85
86bool convertUTF16ToUTF8String(ArrayRef<char> SrcBytes, std::string &Out) {
87  assert(Out.empty());
88
89  // Error out on an uneven byte count.
90  if (SrcBytes.size() % 2)
91    return false;
92
93  // Avoid OOB by returning early on empty input.
94  if (SrcBytes.empty())
95    return true;
96
97  const UTF16 *Src = reinterpret_cast<const UTF16 *>(SrcBytes.begin());
98  const UTF16 *SrcEnd = reinterpret_cast<const UTF16 *>(SrcBytes.end());
99
100  // Byteswap if necessary.
101  std::vector<UTF16> ByteSwapped;
102  if (Src[0] == UNI_UTF16_BYTE_ORDER_MARK_SWAPPED) {
103    ByteSwapped.insert(ByteSwapped.end(), Src, SrcEnd);
104    for (unsigned I = 0, E = ByteSwapped.size(); I != E; ++I)
105      ByteSwapped[I] = llvm::sys::SwapByteOrder_16(ByteSwapped[I]);
106    Src = &ByteSwapped[0];
107    SrcEnd = &ByteSwapped[ByteSwapped.size() - 1] + 1;
108  }
109
110  // Skip the BOM for conversion.
111  if (Src[0] == UNI_UTF16_BYTE_ORDER_MARK_NATIVE)
112    Src++;
113
114  // Just allocate enough space up front.  We'll shrink it later.  Allocate
115  // enough that we can fit a null terminator without reallocating.
116  Out.resize(SrcBytes.size() * UNI_MAX_UTF8_BYTES_PER_CODE_POINT + 1);
117  UTF8 *Dst = reinterpret_cast<UTF8 *>(&Out[0]);
118  UTF8 *DstEnd = Dst + Out.size();
119
120  ConversionResult CR =
121      ConvertUTF16toUTF8(&Src, SrcEnd, &Dst, DstEnd, strictConversion);
122  assert(CR != targetExhausted);
123
124  if (CR != conversionOK) {
125    Out.clear();
126    return false;
127  }
128
129  Out.resize(reinterpret_cast<char *>(Dst) - &Out[0]);
130  Out.push_back(0);
131  Out.pop_back();
132  return true;
133}
134
135bool convertUTF16ToUTF8String(ArrayRef<UTF16> Src, std::string &Out)
136{
137  return convertUTF16ToUTF8String(
138      llvm::ArrayRef<char>(reinterpret_cast<const char *>(Src.data()),
139      Src.size() * sizeof(UTF16)), Out);
140}
141
142bool convertUTF8ToUTF16String(StringRef SrcUTF8,
143                              SmallVectorImpl<UTF16> &DstUTF16) {
144  assert(DstUTF16.empty());
145
146  // Avoid OOB by returning early on empty input.
147  if (SrcUTF8.empty()) {
148    DstUTF16.push_back(0);
149    DstUTF16.pop_back();
150    return true;
151  }
152
153  const UTF8 *Src = reinterpret_cast<const UTF8 *>(SrcUTF8.begin());
154  const UTF8 *SrcEnd = reinterpret_cast<const UTF8 *>(SrcUTF8.end());
155
156  // Allocate the same number of UTF-16 code units as UTF-8 code units. Encoding
157  // as UTF-16 should always require the same amount or less code units than the
158  // UTF-8 encoding.  Allocate one extra byte for the null terminator though,
159  // so that someone calling DstUTF16.data() gets a null terminated string.
160  // We resize down later so we don't have to worry that this over allocates.
161  DstUTF16.resize(SrcUTF8.size()+1);
162  UTF16 *Dst = &DstUTF16[0];
163  UTF16 *DstEnd = Dst + DstUTF16.size();
164
165  ConversionResult CR =
166      ConvertUTF8toUTF16(&Src, SrcEnd, &Dst, DstEnd, strictConversion);
167  assert(CR != targetExhausted);
168
169  if (CR != conversionOK) {
170    DstUTF16.clear();
171    return false;
172  }
173
174  DstUTF16.resize(Dst - &DstUTF16[0]);
175  DstUTF16.push_back(0);
176  DstUTF16.pop_back();
177  return true;
178}
179
180static_assert(sizeof(wchar_t) == 1 || sizeof(wchar_t) == 2 ||
181                  sizeof(wchar_t) == 4,
182              "Expected wchar_t to be 1, 2, or 4 bytes");
183
184template <typename TResult>
185static inline bool ConvertUTF8toWideInternal(llvm::StringRef Source,
186                                             TResult &Result) {
187  // Even in the case of UTF-16, the number of bytes in a UTF-8 string is
188  // at least as large as the number of elements in the resulting wide
189  // string, because surrogate pairs take at least 4 bytes in UTF-8.
190  Result.resize(Source.size() + 1);
191  char *ResultPtr = reinterpret_cast<char *>(&Result[0]);
192  const UTF8 *ErrorPtr;
193  if (!ConvertUTF8toWide(sizeof(wchar_t), Source, ResultPtr, ErrorPtr)) {
194    Result.clear();
195    return false;
196  }
197  Result.resize(reinterpret_cast<wchar_t *>(ResultPtr) - &Result[0]);
198  return true;
199}
200
201bool ConvertUTF8toWide(llvm::StringRef Source, std::wstring &Result) {
202  return ConvertUTF8toWideInternal(Source, Result);
203}
204
205bool ConvertUTF8toWide(const char *Source, std::wstring &Result) {
206  if (!Source) {
207    Result.clear();
208    return true;
209  }
210  return ConvertUTF8toWide(llvm::StringRef(Source), Result);
211}
212
213bool convertWideToUTF8(const std::wstring &Source, std::string &Result) {
214  if (sizeof(wchar_t) == 1) {
215    const UTF8 *Start = reinterpret_cast<const UTF8 *>(Source.data());
216    const UTF8 *End =
217        reinterpret_cast<const UTF8 *>(Source.data() + Source.size());
218    if (!isLegalUTF8String(&Start, End))
219      return false;
220    Result.resize(Source.size());
221    memcpy(&Result[0], Source.data(), Source.size());
222    return true;
223  } else if (sizeof(wchar_t) == 2) {
224    return convertUTF16ToUTF8String(
225        llvm::ArrayRef<UTF16>(reinterpret_cast<const UTF16 *>(Source.data()),
226                              Source.size()),
227        Result);
228  } else if (sizeof(wchar_t) == 4) {
229    const UTF32 *Start = reinterpret_cast<const UTF32 *>(Source.data());
230    const UTF32 *End =
231        reinterpret_cast<const UTF32 *>(Source.data() + Source.size());
232    Result.resize(UNI_MAX_UTF8_BYTES_PER_CODE_POINT * Source.size());
233    UTF8 *ResultPtr = reinterpret_cast<UTF8 *>(&Result[0]);
234    UTF8 *ResultEnd = reinterpret_cast<UTF8 *>(&Result[0] + Result.size());
235    if (ConvertUTF32toUTF8(&Start, End, &ResultPtr, ResultEnd,
236                           strictConversion) == conversionOK) {
237      Result.resize(reinterpret_cast<char *>(ResultPtr) - &Result[0]);
238      return true;
239    } else {
240      Result.clear();
241      return false;
242    }
243  } else {
244    llvm_unreachable(
245        "Control should never reach this point; see static_assert further up");
246  }
247}
248
249} // end namespace llvm
250
251