1//===-- MsgPackDocument.h - MsgPack Document --------------------*- 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/// \file
9/// This file declares a class that exposes a simple in-memory representation
10/// of a document of MsgPack objects, that can be read from MsgPack, written to
11/// MsgPack, and inspected and modified in memory. This is intended to be a
12/// lighter-weight (in terms of memory allocations) replacement for
13/// MsgPackTypes.
14///
15//===----------------------------------------------------------------------===//
16
17#ifndef LLVM_BINARYFORMAT_MSGPACKDOCUMENT_H
18#define LLVM_BINARYFORMAT_MSGPACKDOCUMENT_H
19
20#include "llvm/BinaryFormat/MsgPackReader.h"
21#include <map>
22
23namespace llvm {
24namespace msgpack {
25
26class ArrayDocNode;
27class Document;
28class MapDocNode;
29
30/// The kind of a DocNode and its owning Document.
31struct KindAndDocument {
32 Document *Doc;
33 Type Kind;
34};
35
36/// A node in a MsgPack Document. This is a simple copyable and
37/// passable-by-value type that does not own any memory.
38class DocNode {
39 friend Document;
40
41public:
42 typedef std::map<DocNode, DocNode> MapTy;
43 typedef std::vector<DocNode> ArrayTy;
44
45private:
46 // Using KindAndDocument allows us to squeeze Kind and a pointer to the
47 // owning Document into the same word. Having a pointer to the owning
48 // Document makes the API of DocNode more convenient, and allows its use in
49 // YAMLIO.
50 const KindAndDocument *KindAndDoc;
51
52protected:
53 // The union of different values.
54 union {
55 int64_t Int;
56 uint64_t UInt;
57 bool Bool;
58 double Float;
59 StringRef Raw;
60 ArrayTy *Array;
61 MapTy *Map;
62 };
63
64public:
65 // Default constructor gives an empty node with no associated Document. All
66 // you can do with it is "isEmpty()".
67 DocNode() : KindAndDoc(nullptr) {}
68
69 // Type methods
70 bool isMap() const { return getKind() == Type::Map; }
71 bool isArray() const { return getKind() == Type::Array; }
72 bool isScalar() const { return !isMap() && !isArray(); }
73 bool isString() const { return getKind() == Type::String; }
74
75 // Accessors. isEmpty() returns true for both a default-constructed DocNode
76 // that has no associated Document, and the result of getEmptyNode(), which
77 // does have an associated document.
78 bool isEmpty() const { return !KindAndDoc || getKind() == Type::Empty; }
79 Type getKind() const { return KindAndDoc->Kind; }
80 Document *getDocument() const { return KindAndDoc->Doc; }
81
82 int64_t &getInt() {
83 assert(getKind() == Type::Int);
84 return Int;
85 }
86
87 uint64_t &getUInt() {
88 assert(getKind() == Type::UInt);
89 return UInt;
90 }
91
92 bool &getBool() {
93 assert(getKind() == Type::Boolean);
94 return Bool;
95 }
96
97 double &getFloat() {
98 assert(getKind() == Type::Float);
99 return Float;
100 }
101
102 int64_t getInt() const {
103 assert(getKind() == Type::Int);
104 return Int;
105 }
106
107 uint64_t getUInt() const {
108 assert(getKind() == Type::UInt);
109 return UInt;
110 }
111
112 bool getBool() const {
113 assert(getKind() == Type::Boolean);
114 return Bool;
115 }
116
117 double getFloat() const {
118 assert(getKind() == Type::Float);
119 return Float;
120 }
121
122 StringRef getString() const {
123 assert(getKind() == Type::String);
124 return Raw;
125 }
126
127 /// Get an ArrayDocNode for an array node. If Convert, convert the node to an
128 /// array node if necessary.
129 ArrayDocNode &getArray(bool Convert = false) {
130 if (getKind() != Type::Array) {
131 assert(Convert);
132 convertToArray();
133 }
134 // This could be a static_cast, except ArrayDocNode is a forward reference.
135 return *reinterpret_cast<ArrayDocNode *>(this);
136 }
137
138 /// Get a MapDocNode for a map node. If Convert, convert the node to a map
139 /// node if necessary.
140 MapDocNode &getMap(bool Convert = false) {
141 if (getKind() != Type::Map) {
142 assert(Convert);
143 convertToMap();
144 }
145 // This could be a static_cast, except MapDocNode is a forward reference.
146 return *reinterpret_cast<MapDocNode *>(this);
147 }
148
149 /// Comparison operator, used for map keys.
150 friend bool operator<(const DocNode &Lhs, const DocNode &Rhs) {
151 // This has to cope with one or both of the nodes being default-constructed,
152 // such that KindAndDoc is not set.
153 if (Rhs.isEmpty())
154 return false;
155 if (Lhs.KindAndDoc != Rhs.KindAndDoc) {
156 if (Lhs.isEmpty())
157 return true;
158 return (unsigned)Lhs.getKind() < (unsigned)Rhs.getKind();
159 }
160 switch (Lhs.getKind()) {
161 case Type::Int:
162 return Lhs.Int < Rhs.Int;
163 case Type::UInt:
164 return Lhs.UInt < Rhs.UInt;
165 case Type::Nil:
166 return false;
167 case Type::Boolean:
168 return Lhs.Bool < Rhs.Bool;
169 case Type::Float:
170 return Lhs.Float < Rhs.Float;
171 case Type::String:
172 case Type::Binary:
173 return Lhs.Raw < Rhs.Raw;
174 default:
175 llvm_unreachable("bad map key type");
176 }
177 }
178
179 /// Equality operator
180 friend bool operator==(const DocNode &Lhs, const DocNode &Rhs) {
181 return !(Lhs < Rhs) && !(Rhs < Lhs);
182 }
183
184 /// Inequality operator
185 friend bool operator!=(const DocNode &Lhs, const DocNode &Rhs) {
186 return !(Lhs == Rhs);
187 }
188
189 /// Convert this node to a string, assuming it is scalar.
190 std::string toString() const;
191
192 /// Convert the StringRef and use it to set this DocNode (assuming scalar). If
193 /// it is a string, copy the string into the Document's strings list so we do
194 /// not rely on S having a lifetime beyond this call. Tag is "" or a YAML tag.
195 StringRef fromString(StringRef S, StringRef Tag = "");
196
197 /// Convenience assignment operators. This only works if the destination
198 /// DocNode has an associated Document, i.e. it was not constructed using the
199 /// default constructor. The string one does not copy, so the string must
200 /// remain valid for the lifetime of the Document. Use fromString to avoid
201 /// that restriction.
202 DocNode &operator=(const char *Val) { return *this = StringRef(Val); }
203 DocNode &operator=(StringRef Val);
204 DocNode &operator=(bool Val);
205 DocNode &operator=(int Val);
206 DocNode &operator=(unsigned Val);
207 DocNode &operator=(int64_t Val);
208 DocNode &operator=(uint64_t Val);
209
210private:
211 // Private constructor setting KindAndDoc, used by methods in Document.
212 DocNode(const KindAndDocument *KindAndDoc) : KindAndDoc(KindAndDoc) {}
213
214 void convertToArray();
215 void convertToMap();
216};
217
218/// A DocNode that is a map.
219class MapDocNode : public DocNode {
220public:
221 MapDocNode() {}
222 MapDocNode(DocNode &N) : DocNode(N) { assert(getKind() == Type::Map); }
223
224 // Map access methods.
225 size_t size() const { return Map->size(); }
226 bool empty() const { return !size(); }
227 MapTy::iterator begin() { return Map->begin(); }
228 MapTy::iterator end() { return Map->end(); }
229 MapTy::iterator find(DocNode Key) { return Map->find(Key); }
230 MapTy::iterator find(StringRef Key);
231 MapTy::iterator erase(MapTy::const_iterator I) { return Map->erase(I); }
232 size_t erase(DocNode Key) { return Map->erase(Key); }
233 MapTy::iterator erase(MapTy::const_iterator First,
234 MapTy::const_iterator Second) {
235 return Map->erase(First, Second);
236 }
237 /// Member access. The string data must remain valid for the lifetime of the
238 /// Document.
239 DocNode &operator[](StringRef S);
240 /// Member access, with convenience versions for an integer key.
241 DocNode &operator[](DocNode Key);
242 DocNode &operator[](int Key);
243 DocNode &operator[](unsigned Key);
244 DocNode &operator[](int64_t Key);
245 DocNode &operator[](uint64_t Key);
246};
247
248/// A DocNode that is an array.
249class ArrayDocNode : public DocNode {
250public:
251 ArrayDocNode() {}
252 ArrayDocNode(DocNode &N) : DocNode(N) { assert(getKind() == Type::Array); }
253
254 // Array access methods.
255 size_t size() const { return Array->size(); }
256 bool empty() const { return !size(); }
257 DocNode &back() const { return Array->back(); }
258 ArrayTy::iterator begin() { return Array->begin(); }
259 ArrayTy::iterator end() { return Array->end(); }
260 void push_back(DocNode N) {
261 assert(N.isEmpty() || N.getDocument() == getDocument());
262 Array->push_back(N);
263 }
264
265 /// Element access. This extends the array if necessary, with empty nodes.
266 DocNode &operator[](size_t Index);
267};
268
269/// Simple in-memory representation of a document of msgpack objects with
270/// ability to find and create array and map elements. Does not currently cope
271/// with any extension types.
272class Document {
273 // Maps, arrays and strings used by nodes in the document. No attempt is made
274 // to free unused ones.
275 std::vector<std::unique_ptr<DocNode::MapTy>> Maps;
276 std::vector<std::unique_ptr<DocNode::ArrayTy>> Arrays;
277 std::vector<std::unique_ptr<char[]>> Strings;
278
279 // The root node of the document.
280 DocNode Root;
281
282 // The KindAndDocument structs pointed to by nodes in the document.
283 KindAndDocument KindAndDocs[size_t(Type::Empty) + 1];
284
285 // Whether YAML output uses hex for UInt.
286 bool HexMode = false;
287
288public:
289 Document() {
290 clear();
291 for (unsigned T = 0; T != unsigned(Type::Empty) + 1; ++T)
292 KindAndDocs[T] = {this, Type(T)};
293 }
294
295 /// Get ref to the document's root element.
296 DocNode &getRoot() { return Root; }
297
298 /// Restore the Document to an empty state.
299 void clear() { getRoot() = getEmptyNode(); }
300
301 /// Create an empty node associated with this Document.
302 DocNode getEmptyNode() {
303 auto N = DocNode(&KindAndDocs[size_t(Type::Empty)]);
304 return N;
305 }
306
307 /// Create a nil node associated with this Document.
308 DocNode getNode() {
309 auto N = DocNode(&KindAndDocs[size_t(Type::Nil)]);
310 return N;
311 }
312
313 /// Create an Int node associated with this Document.
314 DocNode getNode(int64_t V) {
315 auto N = DocNode(&KindAndDocs[size_t(Type::Int)]);
316 N.Int = V;
317 return N;
318 }
319
320 /// Create an Int node associated with this Document.
321 DocNode getNode(int V) {
322 auto N = DocNode(&KindAndDocs[size_t(Type::Int)]);
323 N.Int = V;
324 return N;
325 }
326
327 /// Create a UInt node associated with this Document.
328 DocNode getNode(uint64_t V) {
329 auto N = DocNode(&KindAndDocs[size_t(Type::UInt)]);
330 N.UInt = V;
331 return N;
332 }
333
334 /// Create a UInt node associated with this Document.
335 DocNode getNode(unsigned V) {
336 auto N = DocNode(&KindAndDocs[size_t(Type::UInt)]);
337 N.UInt = V;
338 return N;
339 }
340
341 /// Create a Boolean node associated with this Document.
342 DocNode getNode(bool V) {
343 auto N = DocNode(&KindAndDocs[size_t(Type::Boolean)]);
344 N.Bool = V;
345 return N;
346 }
347
348 /// Create a Float node associated with this Document.
349 DocNode getNode(double V) {
350 auto N = DocNode(&KindAndDocs[size_t(Type::Float)]);
351 N.Float = V;
352 return N;
353 }
354
355 /// Create a String node associated with this Document. If !Copy, the passed
356 /// string must remain valid for the lifetime of the Document.
357 DocNode getNode(StringRef V, bool Copy = false) {
358 if (Copy)
359 V = addString(V);
360 auto N = DocNode(&KindAndDocs[size_t(Type::String)]);
361 N.Raw = V;
362 return N;
363 }
364
365 /// Create a String node associated with this Document. If !Copy, the passed
366 /// string must remain valid for the lifetime of the Document.
367 DocNode getNode(const char *V, bool Copy = false) {
368 return getNode(StringRef(V), Copy);
369 }
370
371 /// Create an empty Map node associated with this Document.
372 MapDocNode getMapNode() {
373 auto N = DocNode(&KindAndDocs[size_t(Type::Map)]);
374 Maps.push_back(std::unique_ptr<DocNode::MapTy>(new DocNode::MapTy));
375 N.Map = Maps.back().get();
376 return N.getMap();
377 }
378
379 /// Create an empty Array node associated with this Document.
380 ArrayDocNode getArrayNode() {
381 auto N = DocNode(&KindAndDocs[size_t(Type::Array)]);
382 Arrays.push_back(std::unique_ptr<DocNode::ArrayTy>(new DocNode::ArrayTy));
383 N.Array = Arrays.back().get();
384 return N.getArray();
385 }
386
387 /// Read a document from a binary msgpack blob, merging into anything already
388 /// in the Document. The blob data must remain valid for the lifetime of this
389 /// Document (because a string object in the document contains a StringRef
390 /// into the original blob). If Multi, then this sets root to an array and
391 /// adds top-level objects to it. If !Multi, then it only reads a single
392 /// top-level object, even if there are more, and sets root to that. Returns
393 /// false if failed due to illegal format or merge error.
394 ///
395 /// The Merger arg is a callback function that is called when the merge has a
396 /// conflict, that is, it is trying to set an item that is already set. If the
397 /// conflict cannot be resolved, the callback function returns -1. If the
398 /// conflict can be resolved, the callback returns a non-negative number and
399 /// sets *DestNode to the resolved node. The returned non-negative number is
400 /// significant only for an array node; it is then the array index to start
401 /// populating at. That allows Merger to choose whether to merge array
402 /// elements (returns 0) or append new elements (returns existing size).
403 ///
404 /// If SrcNode is an array or map, the resolution must be that *DestNode is an
405 /// array or map respectively, although it could be the array or map
406 /// (respectively) that was already there. MapKey is the key if *DestNode is a
407 /// map entry, a nil node otherwise.
408 ///
409 /// The default for Merger is to disallow any conflict.
410 bool readFromBlob(
411 StringRef Blob, bool Multi,
412 function_ref<int(DocNode *DestNode, DocNode SrcNode, DocNode MapKey)>
413 Merger = [](DocNode *DestNode, DocNode SrcNode, DocNode MapKey) {
414 return -1;
415 });
416
417 /// Write a MsgPack document to a binary MsgPack blob.
418 void writeToBlob(std::string &Blob);
419
420 /// Copy a string into the Document's strings list, and return the copy that
421 /// is owned by the Document.
422 StringRef addString(StringRef S) {
423 Strings.push_back(std::unique_ptr<char[]>(new char[S.size()]));
424 memcpy(&Strings.back()[0], S.data(), S.size());
425 return StringRef(&Strings.back()[0], S.size());
426 }
427
428 /// Set whether YAML output uses hex for UInt. Default off.
429 void setHexMode(bool Val = true) { HexMode = Val; }
430
431 /// Get Hexmode flag.
432 bool getHexMode() const { return HexMode; }
433
434 /// Convert MsgPack Document to YAML text.
435 void toYAML(raw_ostream &OS);
436
437 /// Read YAML text into the MsgPack document. Returns false on failure.
438 bool fromYAML(StringRef S);
439};
440
441} // namespace msgpack
442} // namespace llvm
443
444#endif // LLVM_BINARYFORMAT_MSGPACKDOCUMENT_H
445