1 //===- BinaryStreamArray.h - Array backed by an arbitrary stream *- C++ -*-===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 #ifndef LLVM_SUPPORT_BINARYSTREAMARRAY_H
11 #define LLVM_SUPPORT_BINARYSTREAMARRAY_H
13 #include "llvm/ADT/ArrayRef.h"
14 #include "llvm/ADT/iterator.h"
15 #include "llvm/Support/BinaryStreamRef.h"
16 #include "llvm/Support/Error.h"
20 /// Lightweight arrays that are backed by an arbitrary BinaryStream. This file
21 /// provides two different array implementations.
23 /// VarStreamArray - Arrays of variable length records. The user specifies
24 /// an Extractor type that can extract a record from a given offset and
25 /// return the number of bytes consumed by the record.
27 /// FixedStreamArray - Arrays of fixed length records. This is similar in
28 /// spirit to ArrayRef<T>, but since it is backed by a BinaryStream, the
29 /// elements of the array need not be laid out in contiguous memory.
32 /// VarStreamArrayExtractor is intended to be specialized to provide customized
33 /// extraction logic. On input it receives a BinaryStreamRef pointing to the
34 /// beginning of the next record, but where the length of the record is not yet
35 /// known. Upon completion, it should return an appropriate Error instance if
36 /// a record could not be extracted, or if one could be extracted it should
37 /// return success and set Len to the number of bytes this record occupied in
38 /// the underlying stream, and it should fill out the fields of the value type
39 /// Item appropriately to represent the current record.
41 /// You can specialize this template for your own custom value types to avoid
42 /// having to specify a second template argument to VarStreamArray (documented
44 template <typename T> struct VarStreamArrayExtractor {
45 // Method intentionally deleted. You must provide an explicit specialization
46 // with the following method implemented.
47 Error operator()(BinaryStreamRef Stream, uint32_t &Len,
48 T &Item) const = delete;
51 /// VarStreamArray represents an array of variable length records backed by a
52 /// stream. This could be a contiguous sequence of bytes in memory, it could
53 /// be a file on disk, or it could be a PDB stream where bytes are stored as
54 /// discontiguous blocks in a file. Usually it is desirable to treat arrays
55 /// as contiguous blocks of memory, but doing so with large PDB files, for
56 /// example, could mean allocating huge amounts of memory just to allow
57 /// re-ordering of stream data to be contiguous before iterating over it. By
58 /// abstracting this out, we need not duplicate this memory, and we can
59 /// iterate over arrays in arbitrarily formatted streams. Elements are parsed
60 /// lazily on iteration, so there is no upfront cost associated with building
61 /// or copying a VarStreamArray, no matter how large it may be.
63 /// You create a VarStreamArray by specifying a ValueType and an Extractor type.
64 /// If you do not specify an Extractor type, you are expected to specialize
65 /// VarStreamArrayExtractor<T> for your ValueType.
67 /// By default an Extractor is default constructed in the class, but in some
68 /// cases you might find it useful for an Extractor to maintain state across
69 /// extractions. In this case you can provide your own Extractor through a
70 /// secondary constructor. The following examples show various ways of
71 /// creating a VarStreamArray.
73 /// // Will use VarStreamArrayExtractor<MyType> as the extractor.
74 /// VarStreamArray<MyType> MyTypeArray;
76 /// // Will use a default-constructed MyExtractor as the extractor.
77 /// VarStreamArray<MyType, MyExtractor> MyTypeArray2;
79 /// // Will use the specific instance of MyExtractor provided.
80 /// // MyExtractor need not be default-constructible in this case.
81 /// MyExtractor E(SomeContext);
82 /// VarStreamArray<MyType, MyExtractor> MyTypeArray3(E);
85 template <typename ValueType, typename Extractor> class VarStreamArrayIterator;
87 template <typename ValueType,
88 typename Extractor = VarStreamArrayExtractor<ValueType>>
89 class VarStreamArray {
90 friend class VarStreamArrayIterator<ValueType, Extractor>;
93 typedef VarStreamArrayIterator<ValueType, Extractor> Iterator;
95 VarStreamArray() = default;
97 explicit VarStreamArray(const Extractor &E) : E(E) {}
99 explicit VarStreamArray(BinaryStreamRef Stream, uint32_t Skew = 0)
100 : Stream(Stream), Skew(Skew) {}
102 VarStreamArray(BinaryStreamRef Stream, const Extractor &E, uint32_t Skew = 0)
103 : Stream(Stream), E(E), Skew(Skew) {}
105 Iterator begin(bool *HadError = nullptr) const {
106 return Iterator(*this, E, Skew, nullptr);
109 bool valid() const { return Stream.valid(); }
111 uint32_t skew() const { return Skew; }
112 Iterator end() const { return Iterator(E); }
114 bool empty() const { return Stream.getLength() == 0; }
116 VarStreamArray<ValueType, Extractor> substream(uint32_t Begin,
117 uint32_t End) const {
118 assert(Begin >= Skew);
119 // We should never cut off the beginning of the stream since it might be
120 // skewed, meaning the initial bytes are important.
121 BinaryStreamRef NewStream = Stream.slice(0, End);
122 return {NewStream, E, Begin};
125 /// given an offset into the array's underlying stream, return an
126 /// iterator to the record at that offset. This is considered unsafe
127 /// since the behavior is undefined if \p Offset does not refer to the
128 /// beginning of a valid record.
129 Iterator at(uint32_t Offset) const {
130 return Iterator(*this, E, Offset, nullptr);
133 const Extractor &getExtractor() const { return E; }
134 Extractor &getExtractor() { return E; }
136 BinaryStreamRef getUnderlyingStream() const { return Stream; }
137 void setUnderlyingStream(BinaryStreamRef S, uint32_t Skew = 0) {
142 void drop_front() { Skew += begin()->length(); }
145 BinaryStreamRef Stream;
150 template <typename ValueType, typename Extractor>
151 class VarStreamArrayIterator
152 : public iterator_facade_base<VarStreamArrayIterator<ValueType, Extractor>,
153 std::forward_iterator_tag, ValueType> {
154 typedef VarStreamArrayIterator<ValueType, Extractor> IterType;
155 typedef VarStreamArray<ValueType, Extractor> ArrayType;
158 VarStreamArrayIterator(const ArrayType &Array, const Extractor &E,
159 uint32_t Offset, bool *HadError)
160 : IterRef(Array.Stream.drop_front(Offset)), Extract(E),
161 Array(&Array), AbsOffset(Offset), HadError(HadError) {
162 if (IterRef.getLength() == 0)
165 auto EC = Extract(IterRef, ThisLen, ThisValue);
167 consumeError(std::move(EC));
173 VarStreamArrayIterator() = default;
174 explicit VarStreamArrayIterator(const Extractor &E) : Extract(E) {}
175 ~VarStreamArrayIterator() = default;
177 bool operator==(const IterType &R) const {
178 if (Array && R.Array) {
179 // Both have a valid array, make sure they're same.
180 assert(Array == R.Array);
181 return IterRef == R.IterRef;
184 // Both iterators are at the end.
185 if (!Array && !R.Array)
188 // One is not at the end and one is.
192 const ValueType &operator*() const {
193 assert(Array && !HasError);
197 ValueType &operator*() {
198 assert(Array && !HasError);
202 IterType &operator+=(unsigned N) {
203 for (unsigned I = 0; I < N; ++I) {
204 // We are done with the current record, discard it so that we are
205 // positioned at the next record.
206 AbsOffset += ThisLen;
207 IterRef = IterRef.drop_front(ThisLen);
208 if (IterRef.getLength() == 0) {
209 // There is nothing after the current record, we must make this an end
213 // There is some data after the current record.
214 auto EC = Extract(IterRef, ThisLen, ThisValue);
216 consumeError(std::move(EC));
218 } else if (ThisLen == 0) {
219 // An empty record? Make this an end iterator.
227 uint32_t offset() const { return AbsOffset; }
228 uint32_t getRecordLength() const { return ThisLen; }
238 if (HadError != nullptr)
243 BinaryStreamRef IterRef;
245 const ArrayType *Array{nullptr};
247 uint32_t AbsOffset{0};
248 bool HasError{false};
249 bool *HadError{nullptr};
252 template <typename T> class FixedStreamArrayIterator;
254 /// FixedStreamArray is similar to VarStreamArray, except with each record
255 /// having a fixed-length. As with VarStreamArray, there is no upfront
256 /// cost associated with building or copying a FixedStreamArray, as the
257 /// memory for each element is not read from the backing stream until that
258 /// element is iterated.
259 template <typename T> class FixedStreamArray {
260 friend class FixedStreamArrayIterator<T>;
263 typedef FixedStreamArrayIterator<T> Iterator;
265 FixedStreamArray() = default;
266 explicit FixedStreamArray(BinaryStreamRef Stream) : Stream(Stream) {
267 assert(Stream.getLength() % sizeof(T) == 0);
270 bool operator==(const FixedStreamArray<T> &Other) const {
271 return Stream == Other.Stream;
274 bool operator!=(const FixedStreamArray<T> &Other) const {
275 return !(*this == Other);
278 FixedStreamArray &operator=(const FixedStreamArray &) = default;
280 const T &operator[](uint32_t Index) const {
281 assert(Index < size());
282 uint32_t Off = Index * sizeof(T);
283 ArrayRef<uint8_t> Data;
284 if (auto EC = Stream.readBytes(Off, sizeof(T), Data)) {
285 assert(false && "Unexpected failure reading from stream");
286 // This should never happen since we asserted that the stream length was
287 // an exact multiple of the element size.
288 consumeError(std::move(EC));
290 assert(llvm::alignmentAdjustment(Data.data(), alignof(T)) == 0);
291 return *reinterpret_cast<const T *>(Data.data());
294 uint32_t size() const { return Stream.getLength() / sizeof(T); }
296 bool empty() const { return size() == 0; }
298 FixedStreamArrayIterator<T> begin() const {
299 return FixedStreamArrayIterator<T>(*this, 0);
302 FixedStreamArrayIterator<T> end() const {
303 return FixedStreamArrayIterator<T>(*this, size());
306 const T &front() const { return *begin(); }
307 const T &back() const {
308 FixedStreamArrayIterator<T> I = end();
312 BinaryStreamRef getUnderlyingStream() const { return Stream; }
315 BinaryStreamRef Stream;
318 template <typename T>
319 class FixedStreamArrayIterator
320 : public iterator_facade_base<FixedStreamArrayIterator<T>,
321 std::random_access_iterator_tag, const T> {
324 FixedStreamArrayIterator(const FixedStreamArray<T> &Array, uint32_t Index)
325 : Array(Array), Index(Index) {}
327 FixedStreamArrayIterator<T> &
328 operator=(const FixedStreamArrayIterator<T> &Other) {
334 const T &operator*() const { return Array[Index]; }
335 const T &operator*() { return Array[Index]; }
337 bool operator==(const FixedStreamArrayIterator<T> &R) const {
338 assert(Array == R.Array);
339 return (Index == R.Index) && (Array == R.Array);
342 FixedStreamArrayIterator<T> &operator+=(std::ptrdiff_t N) {
347 FixedStreamArrayIterator<T> &operator-=(std::ptrdiff_t N) {
348 assert(std::ptrdiff_t(Index) >= N);
353 std::ptrdiff_t operator-(const FixedStreamArrayIterator<T> &R) const {
354 assert(Array == R.Array);
355 assert(Index >= R.Index);
356 return Index - R.Index;
359 bool operator<(const FixedStreamArrayIterator<T> &RHS) const {
360 assert(Array == RHS.Array);
361 return Index < RHS.Index;
365 FixedStreamArray<T> Array;
371 #endif // LLVM_SUPPORT_BINARYSTREAMARRAY_H