1//===--- MemoryBuffer.h - Memory Buffer Interface ---------------*- C++ -*-===//
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//  This file defines the MemoryBuffer interface.
10//
11//===----------------------------------------------------------------------===//
12
13#ifndef LLVM_SUPPORT_MEMORYBUFFER_H
14#define LLVM_SUPPORT_MEMORYBUFFER_H
15
16#include "llvm-c/Types.h"
17#include "llvm/ADT/ArrayRef.h"
18#include "llvm/ADT/StringRef.h"
19#include "llvm/ADT/Twine.h"
20#include "llvm/Support/CBindingWrapping.h"
21#include "llvm/Support/ErrorOr.h"
22#include "llvm/Support/FileSystem.h"
23#include <cstddef>
24#include <cstdint>
25#include <memory>
26
27namespace llvm {
28
29class MemoryBufferRef;
30
31/// This interface provides simple read-only access to a block of memory, and
32/// provides simple methods for reading files and standard input into a memory
33/// buffer.  In addition to basic access to the characters in the file, this
34/// interface guarantees you can read one character past the end of the file,
35/// and that this character will read as '\0'.
36///
37/// The '\0' guarantee is needed to support an optimization -- it's intended to
38/// be more efficient for clients which are reading all the data to stop
39/// reading when they encounter a '\0' than to continually check the file
40/// position to see if it has reached the end of the file.
41class MemoryBuffer {
42  const char *BufferStart; // Start of the buffer.
43  const char *BufferEnd;   // End of the buffer.
44
45protected:
46  MemoryBuffer() = default;
47
48  void init(const char *BufStart, const char *BufEnd,
49            bool RequiresNullTerminator);
50
51  static constexpr sys::fs::mapped_file_region::mapmode Mapmode =
52      sys::fs::mapped_file_region::readonly;
53
54public:
55  MemoryBuffer(const MemoryBuffer &) = delete;
56  MemoryBuffer &operator=(const MemoryBuffer &) = delete;
57  virtual ~MemoryBuffer();
58
59  const char *getBufferStart() const { return BufferStart; }
60  const char *getBufferEnd() const   { return BufferEnd; }
61  size_t getBufferSize() const { return BufferEnd-BufferStart; }
62
63  StringRef getBuffer() const {
64    return StringRef(BufferStart, getBufferSize());
65  }
66
67  /// Return an identifier for this buffer, typically the filename it was read
68  /// from.
69  virtual StringRef getBufferIdentifier() const { return "Unknown buffer"; }
70
71  /// Open the specified file as a MemoryBuffer, returning a new MemoryBuffer
72  /// if successful, otherwise returning null. If FileSize is specified, this
73  /// means that the client knows that the file exists and that it has the
74  /// specified size.
75  ///
76  /// \param IsVolatile Set to true to indicate that the contents of the file
77  /// can change outside the user's control, e.g. when libclang tries to parse
78  /// while the user is editing/updating the file or if the file is on an NFS.
79  static ErrorOr<std::unique_ptr<MemoryBuffer>>
80  getFile(const Twine &Filename, int64_t FileSize = -1,
81          bool RequiresNullTerminator = true, bool IsVolatile = false);
82
83  /// Read all of the specified file into a MemoryBuffer as a stream
84  /// (i.e. until EOF reached). This is useful for special files that
85  /// look like a regular file but have 0 size (e.g. /proc/cpuinfo on Linux).
86  static ErrorOr<std::unique_ptr<MemoryBuffer>>
87  getFileAsStream(const Twine &Filename);
88
89  /// Given an already-open file descriptor, map some slice of it into a
90  /// MemoryBuffer. The slice is specified by an \p Offset and \p MapSize.
91  /// Since this is in the middle of a file, the buffer is not null terminated.
92  static ErrorOr<std::unique_ptr<MemoryBuffer>>
93  getOpenFileSlice(sys::fs::file_t FD, const Twine &Filename, uint64_t MapSize,
94                   int64_t Offset, bool IsVolatile = false);
95
96  /// Given an already-open file descriptor, read the file and return a
97  /// MemoryBuffer.
98  ///
99  /// \param IsVolatile Set to true to indicate that the contents of the file
100  /// can change outside the user's control, e.g. when libclang tries to parse
101  /// while the user is editing/updating the file or if the file is on an NFS.
102  static ErrorOr<std::unique_ptr<MemoryBuffer>>
103  getOpenFile(sys::fs::file_t FD, const Twine &Filename, uint64_t FileSize,
104              bool RequiresNullTerminator = true, bool IsVolatile = false);
105
106  /// Open the specified memory range as a MemoryBuffer. Note that InputData
107  /// must be null terminated if RequiresNullTerminator is true.
108  static std::unique_ptr<MemoryBuffer>
109  getMemBuffer(StringRef InputData, StringRef BufferName = "",
110               bool RequiresNullTerminator = true);
111
112  static std::unique_ptr<MemoryBuffer>
113  getMemBuffer(MemoryBufferRef Ref, bool RequiresNullTerminator = true);
114
115  /// Open the specified memory range as a MemoryBuffer, copying the contents
116  /// and taking ownership of it. InputData does not have to be null terminated.
117  static std::unique_ptr<MemoryBuffer>
118  getMemBufferCopy(StringRef InputData, const Twine &BufferName = "");
119
120  /// Read all of stdin into a file buffer, and return it.
121  static ErrorOr<std::unique_ptr<MemoryBuffer>> getSTDIN();
122
123  /// Open the specified file as a MemoryBuffer, or open stdin if the Filename
124  /// is "-".
125  static ErrorOr<std::unique_ptr<MemoryBuffer>>
126  getFileOrSTDIN(const Twine &Filename, int64_t FileSize = -1,
127                 bool RequiresNullTerminator = true);
128
129  /// Map a subrange of the specified file as a MemoryBuffer.
130  static ErrorOr<std::unique_ptr<MemoryBuffer>>
131  getFileSlice(const Twine &Filename, uint64_t MapSize, uint64_t Offset,
132               bool IsVolatile = false);
133
134  //===--------------------------------------------------------------------===//
135  // Provided for performance analysis.
136  //===--------------------------------------------------------------------===//
137
138  /// The kind of memory backing used to support the MemoryBuffer.
139  enum BufferKind {
140    MemoryBuffer_Malloc,
141    MemoryBuffer_MMap
142  };
143
144  /// Return information on the memory mechanism used to support the
145  /// MemoryBuffer.
146  virtual BufferKind getBufferKind() const = 0;
147
148  MemoryBufferRef getMemBufferRef() const;
149};
150
151/// This class is an extension of MemoryBuffer, which allows copy-on-write
152/// access to the underlying contents.  It only supports creation methods that
153/// are guaranteed to produce a writable buffer.  For example, mapping a file
154/// read-only is not supported.
155class WritableMemoryBuffer : public MemoryBuffer {
156protected:
157  WritableMemoryBuffer() = default;
158
159  static constexpr sys::fs::mapped_file_region::mapmode Mapmode =
160      sys::fs::mapped_file_region::priv;
161
162public:
163  using MemoryBuffer::getBuffer;
164  using MemoryBuffer::getBufferEnd;
165  using MemoryBuffer::getBufferStart;
166
167  // const_cast is well-defined here, because the underlying buffer is
168  // guaranteed to have been initialized with a mutable buffer.
169  char *getBufferStart() {
170    return const_cast<char *>(MemoryBuffer::getBufferStart());
171  }
172  char *getBufferEnd() {
173    return const_cast<char *>(MemoryBuffer::getBufferEnd());
174  }
175  MutableArrayRef<char> getBuffer() {
176    return {getBufferStart(), getBufferEnd()};
177  }
178
179  static ErrorOr<std::unique_ptr<WritableMemoryBuffer>>
180  getFile(const Twine &Filename, int64_t FileSize = -1,
181          bool IsVolatile = false);
182
183  /// Map a subrange of the specified file as a WritableMemoryBuffer.
184  static ErrorOr<std::unique_ptr<WritableMemoryBuffer>>
185  getFileSlice(const Twine &Filename, uint64_t MapSize, uint64_t Offset,
186               bool IsVolatile = false);
187
188  /// Allocate a new MemoryBuffer of the specified size that is not initialized.
189  /// Note that the caller should initialize the memory allocated by this
190  /// method. The memory is owned by the MemoryBuffer object.
191  static std::unique_ptr<WritableMemoryBuffer>
192  getNewUninitMemBuffer(size_t Size, const Twine &BufferName = "");
193
194  /// Allocate a new zero-initialized MemoryBuffer of the specified size. Note
195  /// that the caller need not initialize the memory allocated by this method.
196  /// The memory is owned by the MemoryBuffer object.
197  static std::unique_ptr<WritableMemoryBuffer>
198  getNewMemBuffer(size_t Size, const Twine &BufferName = "");
199
200private:
201  // Hide these base class factory function so one can't write
202  //   WritableMemoryBuffer::getXXX()
203  // and be surprised that he got a read-only Buffer.
204  using MemoryBuffer::getFileAsStream;
205  using MemoryBuffer::getFileOrSTDIN;
206  using MemoryBuffer::getMemBuffer;
207  using MemoryBuffer::getMemBufferCopy;
208  using MemoryBuffer::getOpenFile;
209  using MemoryBuffer::getOpenFileSlice;
210  using MemoryBuffer::getSTDIN;
211};
212
213/// This class is an extension of MemoryBuffer, which allows write access to
214/// the underlying contents and committing those changes to the original source.
215/// It only supports creation methods that are guaranteed to produce a writable
216/// buffer.  For example, mapping a file read-only is not supported.
217class WriteThroughMemoryBuffer : public MemoryBuffer {
218protected:
219  WriteThroughMemoryBuffer() = default;
220
221  static constexpr sys::fs::mapped_file_region::mapmode Mapmode =
222      sys::fs::mapped_file_region::readwrite;
223
224public:
225  using MemoryBuffer::getBuffer;
226  using MemoryBuffer::getBufferEnd;
227  using MemoryBuffer::getBufferStart;
228
229  // const_cast is well-defined here, because the underlying buffer is
230  // guaranteed to have been initialized with a mutable buffer.
231  char *getBufferStart() {
232    return const_cast<char *>(MemoryBuffer::getBufferStart());
233  }
234  char *getBufferEnd() {
235    return const_cast<char *>(MemoryBuffer::getBufferEnd());
236  }
237  MutableArrayRef<char> getBuffer() {
238    return {getBufferStart(), getBufferEnd()};
239  }
240
241  static ErrorOr<std::unique_ptr<WriteThroughMemoryBuffer>>
242  getFile(const Twine &Filename, int64_t FileSize = -1);
243
244  /// Map a subrange of the specified file as a ReadWriteMemoryBuffer.
245  static ErrorOr<std::unique_ptr<WriteThroughMemoryBuffer>>
246  getFileSlice(const Twine &Filename, uint64_t MapSize, uint64_t Offset);
247
248private:
249  // Hide these base class factory function so one can't write
250  //   WritableMemoryBuffer::getXXX()
251  // and be surprised that he got a read-only Buffer.
252  using MemoryBuffer::getFileAsStream;
253  using MemoryBuffer::getFileOrSTDIN;
254  using MemoryBuffer::getMemBuffer;
255  using MemoryBuffer::getMemBufferCopy;
256  using MemoryBuffer::getOpenFile;
257  using MemoryBuffer::getOpenFileSlice;
258  using MemoryBuffer::getSTDIN;
259};
260
261class MemoryBufferRef {
262  StringRef Buffer;
263  StringRef Identifier;
264
265public:
266  MemoryBufferRef() = default;
267  MemoryBufferRef(const MemoryBuffer& Buffer)
268      : Buffer(Buffer.getBuffer()), Identifier(Buffer.getBufferIdentifier()) {}
269  MemoryBufferRef(StringRef Buffer, StringRef Identifier)
270      : Buffer(Buffer), Identifier(Identifier) {}
271
272  StringRef getBuffer() const { return Buffer; }
273
274  StringRef getBufferIdentifier() const { return Identifier; }
275
276  const char *getBufferStart() const { return Buffer.begin(); }
277  const char *getBufferEnd() const { return Buffer.end(); }
278  size_t getBufferSize() const { return Buffer.size(); }
279};
280
281// Create wrappers for C Binding types (see CBindingWrapping.h).
282DEFINE_SIMPLE_CONVERSION_FUNCTIONS(MemoryBuffer, LLVMMemoryBufferRef)
283
284} // end namespace llvm
285
286#endif // LLVM_SUPPORT_MEMORYBUFFER_H
287