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 
17 namespace llvm {
18 
19 bool 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 
65 bool 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 
80 bool 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 
86 bool 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   assert((uintptr_t)Src % sizeof(UTF16) == 0);
101 
102   // Byteswap if necessary.
103   std::vector<UTF16> ByteSwapped;
104   if (Src[0] == UNI_UTF16_BYTE_ORDER_MARK_SWAPPED) {
105     ByteSwapped.insert(ByteSwapped.end(), Src, SrcEnd);
106     for (UTF16 &I : ByteSwapped)
107       I = llvm::ByteSwap_16(I);
108     Src = &ByteSwapped[0];
109     SrcEnd = &ByteSwapped[ByteSwapped.size() - 1] + 1;
110   }
111 
112   // Skip the BOM for conversion.
113   if (Src[0] == UNI_UTF16_BYTE_ORDER_MARK_NATIVE)
114     Src++;
115 
116   // Just allocate enough space up front.  We'll shrink it later.  Allocate
117   // enough that we can fit a null terminator without reallocating.
118   Out.resize(SrcBytes.size() * UNI_MAX_UTF8_BYTES_PER_CODE_POINT + 1);
119   UTF8 *Dst = reinterpret_cast<UTF8 *>(&Out[0]);
120   UTF8 *DstEnd = Dst + Out.size();
121 
122   ConversionResult CR =
123       ConvertUTF16toUTF8(&Src, SrcEnd, &Dst, DstEnd, strictConversion);
124   assert(CR != targetExhausted);
125 
126   if (CR != conversionOK) {
127     Out.clear();
128     return false;
129   }
130 
131   Out.resize(reinterpret_cast<char *>(Dst) - &Out[0]);
132   Out.push_back(0);
133   Out.pop_back();
134   return true;
135 }
136 
137 bool convertUTF16ToUTF8String(ArrayRef<UTF16> Src, std::string &Out)
138 {
139   return convertUTF16ToUTF8String(
140       llvm::ArrayRef<char>(reinterpret_cast<const char *>(Src.data()),
141       Src.size() * sizeof(UTF16)), Out);
142 }
143 
144 bool convertUTF8ToUTF16String(StringRef SrcUTF8,
145                               SmallVectorImpl<UTF16> &DstUTF16) {
146   assert(DstUTF16.empty());
147 
148   // Avoid OOB by returning early on empty input.
149   if (SrcUTF8.empty()) {
150     DstUTF16.push_back(0);
151     DstUTF16.pop_back();
152     return true;
153   }
154 
155   const UTF8 *Src = reinterpret_cast<const UTF8 *>(SrcUTF8.begin());
156   const UTF8 *SrcEnd = reinterpret_cast<const UTF8 *>(SrcUTF8.end());
157 
158   // Allocate the same number of UTF-16 code units as UTF-8 code units. Encoding
159   // as UTF-16 should always require the same amount or less code units than the
160   // UTF-8 encoding.  Allocate one extra byte for the null terminator though,
161   // so that someone calling DstUTF16.data() gets a null terminated string.
162   // We resize down later so we don't have to worry that this over allocates.
163   DstUTF16.resize(SrcUTF8.size()+1);
164   UTF16 *Dst = &DstUTF16[0];
165   UTF16 *DstEnd = Dst + DstUTF16.size();
166 
167   ConversionResult CR =
168       ConvertUTF8toUTF16(&Src, SrcEnd, &Dst, DstEnd, strictConversion);
169   assert(CR != targetExhausted);
170 
171   if (CR != conversionOK) {
172     DstUTF16.clear();
173     return false;
174   }
175 
176   DstUTF16.resize(Dst - &DstUTF16[0]);
177   DstUTF16.push_back(0);
178   DstUTF16.pop_back();
179   return true;
180 }
181 
182 static_assert(sizeof(wchar_t) == 1 || sizeof(wchar_t) == 2 ||
183                   sizeof(wchar_t) == 4,
184               "Expected wchar_t to be 1, 2, or 4 bytes");
185 
186 template <typename TResult>
187 static inline bool ConvertUTF8toWideInternal(llvm::StringRef Source,
188                                              TResult &Result) {
189   // Even in the case of UTF-16, the number of bytes in a UTF-8 string is
190   // at least as large as the number of elements in the resulting wide
191   // string, because surrogate pairs take at least 4 bytes in UTF-8.
192   Result.resize(Source.size() + 1);
193   char *ResultPtr = reinterpret_cast<char *>(&Result[0]);
194   const UTF8 *ErrorPtr;
195   if (!ConvertUTF8toWide(sizeof(wchar_t), Source, ResultPtr, ErrorPtr)) {
196     Result.clear();
197     return false;
198   }
199   Result.resize(reinterpret_cast<wchar_t *>(ResultPtr) - &Result[0]);
200   return true;
201 }
202 
203 bool ConvertUTF8toWide(llvm::StringRef Source, std::wstring &Result) {
204   return ConvertUTF8toWideInternal(Source, Result);
205 }
206 
207 bool ConvertUTF8toWide(const char *Source, std::wstring &Result) {
208   if (!Source) {
209     Result.clear();
210     return true;
211   }
212   return ConvertUTF8toWide(llvm::StringRef(Source), Result);
213 }
214 
215 bool convertWideToUTF8(const std::wstring &Source, std::string &Result) {
216   if (sizeof(wchar_t) == 1) {
217     const UTF8 *Start = reinterpret_cast<const UTF8 *>(Source.data());
218     const UTF8 *End =
219         reinterpret_cast<const UTF8 *>(Source.data() + Source.size());
220     if (!isLegalUTF8String(&Start, End))
221       return false;
222     Result.resize(Source.size());
223     memcpy(&Result[0], Source.data(), Source.size());
224     return true;
225   } else if (sizeof(wchar_t) == 2) {
226     return convertUTF16ToUTF8String(
227         llvm::ArrayRef<UTF16>(reinterpret_cast<const UTF16 *>(Source.data()),
228                               Source.size()),
229         Result);
230   } else if (sizeof(wchar_t) == 4) {
231     const UTF32 *Start = reinterpret_cast<const UTF32 *>(Source.data());
232     const UTF32 *End =
233         reinterpret_cast<const UTF32 *>(Source.data() + Source.size());
234     Result.resize(UNI_MAX_UTF8_BYTES_PER_CODE_POINT * Source.size());
235     UTF8 *ResultPtr = reinterpret_cast<UTF8 *>(&Result[0]);
236     UTF8 *ResultEnd = reinterpret_cast<UTF8 *>(&Result[0] + Result.size());
237     if (ConvertUTF32toUTF8(&Start, End, &ResultPtr, ResultEnd,
238                            strictConversion) == conversionOK) {
239       Result.resize(reinterpret_cast<char *>(ResultPtr) - &Result[0]);
240       return true;
241     } else {
242       Result.clear();
243       return false;
244     }
245   } else {
246     llvm_unreachable(
247         "Control should never reach this point; see static_assert further up");
248   }
249 }
250 
251 } // end namespace llvm
252 
253