| 1 | // Copyright (c) 2013-2014 Sandstorm Development Group, Inc. and contributors |
| 2 | // Licensed under the MIT License: |
| 3 | // |
| 4 | // Permission is hereby granted, free of charge, to any person obtaining a copy |
| 5 | // of this software and associated documentation files (the "Software"), to deal |
| 6 | // in the Software without restriction, including without limitation the rights |
| 7 | // to use, copy, modify, merge, publish, distribute, sublicense, and/or sell |
| 8 | // copies of the Software, and to permit persons to whom the Software is |
| 9 | // furnished to do so, subject to the following conditions: |
| 10 | // |
| 11 | // The above copyright notice and this permission notice shall be included in |
| 12 | // all copies or substantial portions of the Software. |
| 13 | // |
| 14 | // THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
| 15 | // IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
| 16 | // FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE |
| 17 | // AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
| 18 | // LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
| 19 | // OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN |
| 20 | // THE SOFTWARE. |
| 21 | |
| 22 | #include "schema-parser.h" |
| 23 | #include "message.h" |
| 24 | #include <capnp/compiler/compiler.h> |
| 25 | #include <capnp/compiler/lexer.capnp.h> |
| 26 | #include <capnp/compiler/lexer.h> |
| 27 | #include <capnp/compiler/grammar.capnp.h> |
| 28 | #include <capnp/compiler/parser.h> |
| 29 | #include <unordered_map> |
| 30 | #include <kj/mutex.h> |
| 31 | #include <kj/vector.h> |
| 32 | #include <kj/debug.h> |
| 33 | #include <kj/io.h> |
| 34 | #include <map> |
| 35 | |
| 36 | namespace capnp { |
| 37 | |
| 38 | namespace { |
| 39 | |
| 40 | template <typename T> |
| 41 | size_t findLargestElementBefore(const kj::Vector<T>& vec, const T& key) { |
| 42 | KJ_REQUIRE(vec.size() > 0 && vec[0] <= key); |
| 43 | |
| 44 | size_t lower = 0; |
| 45 | size_t upper = vec.size(); |
| 46 | |
| 47 | while (upper - lower > 1) { |
| 48 | size_t mid = (lower + upper) / 2; |
| 49 | if (vec[mid] > key) { |
| 50 | upper = mid; |
| 51 | } else { |
| 52 | lower = mid; |
| 53 | } |
| 54 | } |
| 55 | |
| 56 | return lower; |
| 57 | } |
| 58 | |
| 59 | } // namespace |
| 60 | |
| 61 | // ======================================================================================= |
| 62 | |
| 63 | class SchemaParser::ModuleImpl final: public compiler::Module { |
| 64 | public: |
| 65 | ModuleImpl(const SchemaParser& parser, kj::Own<const SchemaFile>&& file) |
| 66 | : parser(parser), file(kj::mv(file)) {} |
| 67 | |
| 68 | kj::StringPtr getSourceName() override { |
| 69 | return file->getDisplayName(); |
| 70 | } |
| 71 | |
| 72 | Orphan<compiler::ParsedFile> loadContent(Orphanage orphanage) override { |
| 73 | kj::Array<const char> content = file->readContent(); |
| 74 | |
| 75 | lineBreaks.get([&](kj::SpaceFor<kj::Vector<uint>>& space) { |
| 76 | auto vec = space.construct(content.size() / 40); |
| 77 | vec->add(0); |
| 78 | for (const char* pos = content.begin(); pos < content.end(); ++pos) { |
| 79 | if (*pos == '\n') { |
| 80 | vec->add(pos + 1 - content.begin()); |
| 81 | } |
| 82 | } |
| 83 | return vec; |
| 84 | }); |
| 85 | |
| 86 | MallocMessageBuilder lexedBuilder; |
| 87 | auto statements = lexedBuilder.initRoot<compiler::LexedStatements>(); |
| 88 | compiler::lex(content, statements, *this); |
| 89 | |
| 90 | auto parsed = orphanage.newOrphan<compiler::ParsedFile>(); |
| 91 | compiler::parseFile(statements.getStatements(), parsed.get(), *this); |
| 92 | return parsed; |
| 93 | } |
| 94 | |
| 95 | kj::Maybe<Module&> importRelative(kj::StringPtr importPath) override { |
| 96 | KJ_IF_MAYBE(importedFile, file->import(importPath)) { |
| 97 | return parser.getModuleImpl(kj::mv(*importedFile)); |
| 98 | } else { |
| 99 | return nullptr; |
| 100 | } |
| 101 | } |
| 102 | |
| 103 | kj::Maybe<kj::Array<const byte>> embedRelative(kj::StringPtr embedPath) override { |
| 104 | KJ_IF_MAYBE(importedFile, file->import(embedPath)) { |
| 105 | return importedFile->get()->readContent().releaseAsBytes(); |
| 106 | } else { |
| 107 | return nullptr; |
| 108 | } |
| 109 | } |
| 110 | |
| 111 | void addError(uint32_t startByte, uint32_t endByte, kj::StringPtr message) override { |
| 112 | auto& lines = lineBreaks.get( |
| 113 | [](kj::SpaceFor<kj::Vector<uint>>& space) { |
| 114 | KJ_FAIL_REQUIRE("Can't report errors until loadContent() is called." ); |
| 115 | return space.construct(); |
| 116 | }); |
| 117 | |
| 118 | // TODO(someday): This counts tabs as single characters. Do we care? |
| 119 | uint startLine = findLargestElementBefore(lines, startByte); |
| 120 | uint startCol = startByte - lines[startLine]; |
| 121 | uint endLine = findLargestElementBefore(lines, endByte); |
| 122 | uint endCol = endByte - lines[endLine]; |
| 123 | |
| 124 | file->reportError( |
| 125 | SchemaFile::SourcePos { startByte, startLine, startCol }, |
| 126 | SchemaFile::SourcePos { endByte, endLine, endCol }, |
| 127 | message); |
| 128 | |
| 129 | // We intentionally only set hadErrors true if reportError() didn't throw. |
| 130 | parser.hadErrors = true; |
| 131 | } |
| 132 | |
| 133 | bool hadErrors() override { |
| 134 | return parser.hadErrors; |
| 135 | } |
| 136 | |
| 137 | private: |
| 138 | const SchemaParser& parser; |
| 139 | kj::Own<const SchemaFile> file; |
| 140 | |
| 141 | kj::Lazy<kj::Vector<uint>> lineBreaks; |
| 142 | // Byte offsets of the first byte in each source line. The first element is always zero. |
| 143 | // Initialized the first time the module is loaded. |
| 144 | }; |
| 145 | |
| 146 | // ======================================================================================= |
| 147 | |
| 148 | namespace { |
| 149 | |
| 150 | struct SchemaFileHash { |
| 151 | inline bool operator()(const SchemaFile* f) const { |
| 152 | return f->hashCode(); |
| 153 | } |
| 154 | }; |
| 155 | |
| 156 | struct SchemaFileEq { |
| 157 | inline bool operator()(const SchemaFile* a, const SchemaFile* b) const { |
| 158 | return *a == *b; |
| 159 | } |
| 160 | }; |
| 161 | |
| 162 | } // namespace |
| 163 | |
| 164 | struct SchemaParser::DiskFileCompat { |
| 165 | // Stuff we only create if parseDiskFile() is ever called, in order to translate that call into |
| 166 | // KJ filesystem API calls. |
| 167 | |
| 168 | kj::Own<kj::Filesystem> ownFs; |
| 169 | kj::Filesystem& fs; |
| 170 | |
| 171 | struct ImportDir { |
| 172 | kj::String pathStr; |
| 173 | kj::Path path; |
| 174 | kj::Own<const kj::ReadableDirectory> dir; |
| 175 | }; |
| 176 | std::map<kj::StringPtr, ImportDir> cachedImportDirs; |
| 177 | |
| 178 | std::map<std::pair<const kj::StringPtr*, size_t>, kj::Array<const kj::ReadableDirectory*>> |
| 179 | cachedImportPaths; |
| 180 | |
| 181 | DiskFileCompat(): ownFs(kj::newDiskFilesystem()), fs(*ownFs) {} |
| 182 | DiskFileCompat(kj::Filesystem& fs): fs(fs) {} |
| 183 | }; |
| 184 | |
| 185 | struct SchemaParser::Impl { |
| 186 | typedef std::unordered_map< |
| 187 | const SchemaFile*, kj::Own<ModuleImpl>, SchemaFileHash, SchemaFileEq> FileMap; |
| 188 | kj::MutexGuarded<FileMap> fileMap; |
| 189 | compiler::Compiler compiler; |
| 190 | |
| 191 | kj::MutexGuarded<kj::Maybe<DiskFileCompat>> compat; |
| 192 | }; |
| 193 | |
| 194 | SchemaParser::SchemaParser(): impl(kj::heap<Impl>()) {} |
| 195 | SchemaParser::~SchemaParser() noexcept(false) {} |
| 196 | |
| 197 | ParsedSchema SchemaParser::parseFromDirectory( |
| 198 | const kj::ReadableDirectory& baseDir, kj::Path path, |
| 199 | kj::ArrayPtr<const kj::ReadableDirectory* const> importPath) const { |
| 200 | return parseFile(SchemaFile::newFromDirectory(baseDir, kj::mv(path), importPath)); |
| 201 | } |
| 202 | |
| 203 | ParsedSchema SchemaParser::parseDiskFile( |
| 204 | kj::StringPtr displayName, kj::StringPtr diskPath, |
| 205 | kj::ArrayPtr<const kj::StringPtr> importPath) const { |
| 206 | auto lock = impl->compat.lockExclusive(); |
| 207 | DiskFileCompat* compat; |
| 208 | KJ_IF_MAYBE(c, *lock) { |
| 209 | compat = c; |
| 210 | } else { |
| 211 | compat = &lock->emplace(); |
| 212 | } |
| 213 | |
| 214 | auto& root = compat->fs.getRoot(); |
| 215 | auto cwd = compat->fs.getCurrentPath(); |
| 216 | |
| 217 | const kj::ReadableDirectory* baseDir = &root; |
| 218 | kj::Path path = cwd.evalNative(diskPath); |
| 219 | |
| 220 | kj::ArrayPtr<const kj::ReadableDirectory* const> translatedImportPath = nullptr; |
| 221 | |
| 222 | if (importPath.size() > 0) { |
| 223 | auto importPathKey = std::make_pair(importPath.begin(), importPath.size()); |
| 224 | auto& slot = compat->cachedImportPaths[importPathKey]; |
| 225 | |
| 226 | if (slot == nullptr) { |
| 227 | slot = KJ_MAP(path, importPath) -> const kj::ReadableDirectory* { |
| 228 | auto iter = compat->cachedImportDirs.find(path); |
| 229 | if (iter != compat->cachedImportDirs.end()) { |
| 230 | return iter->second.dir; |
| 231 | } |
| 232 | |
| 233 | auto parsed = cwd.evalNative(path); |
| 234 | kj::Own<const kj::ReadableDirectory> dir; |
| 235 | KJ_IF_MAYBE(d, root.tryOpenSubdir(parsed)) { |
| 236 | dir = kj::mv(*d); |
| 237 | } else { |
| 238 | // Ignore paths that don't exist. |
| 239 | dir = kj::newInMemoryDirectory(kj::nullClock()); |
| 240 | } |
| 241 | |
| 242 | const kj::ReadableDirectory* result = dir; |
| 243 | |
| 244 | kj::StringPtr pathRef = path; |
| 245 | KJ_ASSERT(compat->cachedImportDirs.insert(std::make_pair(pathRef, |
| 246 | DiskFileCompat::ImportDir { kj::str(path), kj::mv(parsed), kj::mv(dir) })).second); |
| 247 | |
| 248 | return result; |
| 249 | }; |
| 250 | } |
| 251 | |
| 252 | translatedImportPath = slot; |
| 253 | |
| 254 | // Check if `path` appears to be inside any of the import path directories. If so, adjust |
| 255 | // to be relative to that directory rather than absolute. |
| 256 | kj::Maybe<DiskFileCompat::ImportDir&> matchedImportDir; |
| 257 | size_t bestMatchLength = 0; |
| 258 | for (auto importDir: importPath) { |
| 259 | auto iter = compat->cachedImportDirs.find(importDir); |
| 260 | KJ_ASSERT(iter != compat->cachedImportDirs.end()); |
| 261 | |
| 262 | if (path.startsWith(iter->second.path)) { |
| 263 | // Looks like we're trying to load a file from inside this import path. Treat the import |
| 264 | // path as the base directory. |
| 265 | if (iter->second.path.size() > bestMatchLength) { |
| 266 | bestMatchLength = iter->second.path.size(); |
| 267 | matchedImportDir = iter->second; |
| 268 | } |
| 269 | } |
| 270 | } |
| 271 | |
| 272 | KJ_IF_MAYBE(match, matchedImportDir) { |
| 273 | baseDir = match->dir; |
| 274 | path = path.slice(match->path.size(), path.size()).clone(); |
| 275 | } |
| 276 | } |
| 277 | |
| 278 | return parseFile(SchemaFile::newFromDirectory( |
| 279 | *baseDir, kj::mv(path), translatedImportPath, kj::str(displayName))); |
| 280 | } |
| 281 | |
| 282 | void SchemaParser::setDiskFilesystem(kj::Filesystem& fs) { |
| 283 | auto lock = impl->compat.lockExclusive(); |
| 284 | KJ_REQUIRE(*lock == nullptr, "already called parseDiskFile() or setDiskFilesystem()" ); |
| 285 | lock->emplace(fs); |
| 286 | } |
| 287 | |
| 288 | ParsedSchema SchemaParser::parseFile(kj::Own<SchemaFile>&& file) const { |
| 289 | KJ_DEFER(impl->compiler.clearWorkspace()); |
| 290 | uint64_t id = impl->compiler.add(getModuleImpl(kj::mv(file))); |
| 291 | impl->compiler.eagerlyCompile(id, |
| 292 | compiler::Compiler::NODE | compiler::Compiler::CHILDREN | |
| 293 | compiler::Compiler::DEPENDENCIES | compiler::Compiler::DEPENDENCY_DEPENDENCIES); |
| 294 | return ParsedSchema(impl->compiler.getLoader().get(id), *this); |
| 295 | } |
| 296 | |
| 297 | kj::Maybe<schema::Node::SourceInfo::Reader> SchemaParser::getSourceInfo(Schema schema) const { |
| 298 | return impl->compiler.getSourceInfo(schema.getProto().getId()); |
| 299 | } |
| 300 | |
| 301 | SchemaParser::ModuleImpl& SchemaParser::getModuleImpl(kj::Own<SchemaFile>&& file) const { |
| 302 | auto lock = impl->fileMap.lockExclusive(); |
| 303 | |
| 304 | auto insertResult = lock->insert(std::make_pair(file.get(), kj::Own<ModuleImpl>())); |
| 305 | if (insertResult.second) { |
| 306 | // This is a newly-inserted entry. Construct the ModuleImpl. |
| 307 | insertResult.first->second = kj::heap<ModuleImpl>(*this, kj::mv(file)); |
| 308 | } |
| 309 | return *insertResult.first->second; |
| 310 | } |
| 311 | |
| 312 | SchemaLoader& SchemaParser::getLoader() { |
| 313 | return impl->compiler.getLoader(); |
| 314 | } |
| 315 | |
| 316 | kj::Maybe<ParsedSchema> ParsedSchema::findNested(kj::StringPtr name) const { |
| 317 | return parser->impl->compiler.lookup(getProto().getId(), name).map( |
| 318 | [this](uint64_t childId) { |
| 319 | return ParsedSchema(parser->impl->compiler.getLoader().get(childId), *parser); |
| 320 | }); |
| 321 | } |
| 322 | |
| 323 | ParsedSchema ParsedSchema::getNested(kj::StringPtr nestedName) const { |
| 324 | KJ_IF_MAYBE(nested, findNested(nestedName)) { |
| 325 | return *nested; |
| 326 | } else { |
| 327 | KJ_FAIL_REQUIRE("no such nested declaration" , getProto().getDisplayName(), nestedName); |
| 328 | } |
| 329 | } |
| 330 | |
| 331 | schema::Node::SourceInfo::Reader ParsedSchema::getSourceInfo() const { |
| 332 | return KJ_ASSERT_NONNULL(parser->getSourceInfo(*this)); |
| 333 | } |
| 334 | |
| 335 | // ------------------------------------------------------------------- |
| 336 | |
| 337 | class SchemaFile::DiskSchemaFile final: public SchemaFile { |
| 338 | public: |
| 339 | DiskSchemaFile(const kj::ReadableDirectory& baseDir, kj::Path pathParam, |
| 340 | kj::ArrayPtr<const kj::ReadableDirectory* const> importPath, |
| 341 | kj::Own<const kj::ReadableFile> file, |
| 342 | kj::Maybe<kj::String> displayNameOverride) |
| 343 | : baseDir(baseDir), path(kj::mv(pathParam)), importPath(importPath), file(kj::mv(file)) { |
| 344 | KJ_IF_MAYBE(dn, displayNameOverride) { |
| 345 | displayName = kj::mv(*dn); |
| 346 | displayNameOverridden = true; |
| 347 | } else { |
| 348 | displayName = path.toString(); |
| 349 | displayNameOverridden = false; |
| 350 | } |
| 351 | } |
| 352 | |
| 353 | kj::StringPtr getDisplayName() const override { |
| 354 | return displayName; |
| 355 | } |
| 356 | |
| 357 | kj::Array<const char> readContent() const override { |
| 358 | return file->mmap(0, file->stat().size).releaseAsChars(); |
| 359 | } |
| 360 | |
| 361 | kj::Maybe<kj::Own<SchemaFile>> import(kj::StringPtr target) const override { |
| 362 | if (target.startsWith("/" )) { |
| 363 | auto parsed = kj::Path::parse(target.slice(1)); |
| 364 | for (auto candidate: importPath) { |
| 365 | KJ_IF_MAYBE(newFile, candidate->tryOpenFile(parsed)) { |
| 366 | return kj::implicitCast<kj::Own<SchemaFile>>(kj::heap<DiskSchemaFile>( |
| 367 | *candidate, kj::mv(parsed), importPath, kj::mv(*newFile), nullptr)); |
| 368 | } |
| 369 | } |
| 370 | return nullptr; |
| 371 | } else { |
| 372 | auto parsed = path.parent().eval(target); |
| 373 | |
| 374 | kj::Maybe<kj::String> displayNameOverride; |
| 375 | if (displayNameOverridden) { |
| 376 | // Try to create a consistent display name override for the imported file. This is for |
| 377 | // backwards-compatibility only -- display names are only overridden when using the |
| 378 | // deprecated parseDiskFile() interface. |
| 379 | kj::runCatchingExceptions([&]() { |
| 380 | displayNameOverride = kj::Path::parse(displayName).parent().eval(target).toString(); |
| 381 | }); |
| 382 | } |
| 383 | |
| 384 | KJ_IF_MAYBE(newFile, baseDir.tryOpenFile(parsed)) { |
| 385 | return kj::implicitCast<kj::Own<SchemaFile>>(kj::heap<DiskSchemaFile>( |
| 386 | baseDir, kj::mv(parsed), importPath, kj::mv(*newFile), kj::mv(displayNameOverride))); |
| 387 | } else { |
| 388 | return nullptr; |
| 389 | } |
| 390 | } |
| 391 | } |
| 392 | |
| 393 | bool operator==(const SchemaFile& other) const override { |
| 394 | auto& other2 = kj::downcast<const DiskSchemaFile>(other); |
| 395 | return &baseDir == &other2.baseDir && path == other2.path; |
| 396 | } |
| 397 | bool operator!=(const SchemaFile& other) const override { |
| 398 | return !operator==(other); |
| 399 | } |
| 400 | size_t hashCode() const override { |
| 401 | // djb hash with xor |
| 402 | // TODO(someday): Add hashing library to KJ. |
| 403 | size_t result = reinterpret_cast<uintptr_t>(&baseDir); |
| 404 | for (auto& part: path) { |
| 405 | for (char c: part) { |
| 406 | result = (result * 33) ^ c; |
| 407 | } |
| 408 | result = (result * 33) ^ '/'; |
| 409 | } |
| 410 | return result; |
| 411 | } |
| 412 | |
| 413 | void reportError(SourcePos start, SourcePos end, kj::StringPtr message) const override { |
| 414 | kj::getExceptionCallback().onRecoverableException(kj::Exception( |
| 415 | kj::Exception::Type::FAILED, path.toString(), start.line, |
| 416 | kj::heapString(message))); |
| 417 | } |
| 418 | |
| 419 | private: |
| 420 | const kj::ReadableDirectory& baseDir; |
| 421 | kj::Path path; |
| 422 | kj::ArrayPtr<const kj::ReadableDirectory* const> importPath; |
| 423 | kj::Own<const kj::ReadableFile> file; |
| 424 | kj::String displayName; |
| 425 | bool displayNameOverridden; |
| 426 | }; |
| 427 | |
| 428 | kj::Own<SchemaFile> SchemaFile::newFromDirectory( |
| 429 | const kj::ReadableDirectory& baseDir, kj::Path path, |
| 430 | kj::ArrayPtr<const kj::ReadableDirectory* const> importPath, |
| 431 | kj::Maybe<kj::String> displayNameOverride) { |
| 432 | return kj::heap<DiskSchemaFile>(baseDir, kj::mv(path), importPath, baseDir.openFile(path), |
| 433 | kj::mv(displayNameOverride)); |
| 434 | } |
| 435 | |
| 436 | } // namespace capnp |
| 437 | |