| 1 | //===---- URI.h - File URIs with schemes -------------------------*- 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 | #include "URI.h" |
| 10 | #include "support/Logger.h" |
| 11 | #include "llvm/ADT/StringExtras.h" |
| 12 | #include "llvm/ADT/Twine.h" |
| 13 | #include "llvm/Support/Error.h" |
| 14 | #include "llvm/Support/Path.h" |
| 15 | #include <algorithm> |
| 16 | |
| 17 | LLVM_INSTANTIATE_REGISTRY(clang::clangd::URISchemeRegistry) |
| 18 | |
| 19 | namespace clang { |
| 20 | namespace clangd { |
| 21 | namespace { |
| 22 | |
| 23 | bool isWindowsPath(llvm::StringRef Path) { |
| 24 | return Path.size() > 1 && llvm::isAlpha(C: Path[0]) && Path[1] == ':'; |
| 25 | } |
| 26 | |
| 27 | bool isNetworkPath(llvm::StringRef Path) { |
| 28 | return Path.size() > 2 && Path[0] == Path[1] && |
| 29 | llvm::sys::path::is_separator(value: Path[0]); |
| 30 | } |
| 31 | |
| 32 | /// This manages file paths in the file system. All paths in the scheme |
| 33 | /// are absolute (with leading '/'). |
| 34 | /// Note that this scheme is hardcoded into the library and not registered in |
| 35 | /// registry. |
| 36 | class FileSystemScheme : public URIScheme { |
| 37 | public: |
| 38 | llvm::Expected<std::string> |
| 39 | getAbsolutePath(llvm::StringRef Authority, llvm::StringRef Body, |
| 40 | llvm::StringRef /*HintPath*/) const override { |
| 41 | if (!Body.starts_with(Prefix: "/" )) |
| 42 | return error(Fmt: "File scheme: expect body to be an absolute path starting " |
| 43 | "with '/': {0}" , |
| 44 | Vals&: Body); |
| 45 | llvm::SmallString<128> Path; |
| 46 | if (!Authority.empty()) { |
| 47 | // Windows UNC paths e.g. file://server/share => \\server\share |
| 48 | ("//" + Authority).toVector(Out&: Path); |
| 49 | } else if (isWindowsPath(Path: Body.substr(Start: 1))) { |
| 50 | // Windows paths e.g. file:///X:/path => X:\path |
| 51 | Body.consume_front(Prefix: "/" ); |
| 52 | } |
| 53 | Path.append(RHS: Body); |
| 54 | llvm::sys::path::native(path&: Path); |
| 55 | return std::string(Path); |
| 56 | } |
| 57 | |
| 58 | llvm::Expected<URI> |
| 59 | uriFromAbsolutePath(llvm::StringRef AbsolutePath) const override { |
| 60 | std::string Body; |
| 61 | llvm::StringRef Authority; |
| 62 | llvm::StringRef Root = llvm::sys::path::root_name(path: AbsolutePath); |
| 63 | if (isNetworkPath(Path: Root)) { |
| 64 | // Windows UNC paths e.g. \\server\share => file://server/share |
| 65 | Authority = Root.drop_front(N: 2); |
| 66 | AbsolutePath.consume_front(Prefix: Root); |
| 67 | } else if (isWindowsPath(Path: Root)) { |
| 68 | // Windows paths e.g. X:\path => file:///X:/path |
| 69 | Body = "/" ; |
| 70 | } |
| 71 | Body += llvm::sys::path::convert_to_slash(path: AbsolutePath); |
| 72 | return URI("file" , Authority, Body); |
| 73 | } |
| 74 | }; |
| 75 | |
| 76 | llvm::Expected<std::unique_ptr<URIScheme>> |
| 77 | findSchemeByName(llvm::StringRef Scheme) { |
| 78 | if (Scheme == "file" ) |
| 79 | return std::make_unique<FileSystemScheme>(); |
| 80 | |
| 81 | for (const auto &URIScheme : URISchemeRegistry::entries()) { |
| 82 | if (URIScheme.getName() != Scheme) |
| 83 | continue; |
| 84 | return URIScheme.instantiate(); |
| 85 | } |
| 86 | return error(Fmt: "Can't find scheme: {0}" , Vals&: Scheme); |
| 87 | } |
| 88 | |
| 89 | bool shouldEscape(unsigned char C) { |
| 90 | // Unreserved characters. |
| 91 | if ((C >= 'a' && C <= 'z') || (C >= 'A' && C <= 'Z') || |
| 92 | (C >= '0' && C <= '9')) |
| 93 | return false; |
| 94 | switch (C) { |
| 95 | case '-': |
| 96 | case '_': |
| 97 | case '.': |
| 98 | case '~': |
| 99 | case '/': // '/' is only reserved when parsing. |
| 100 | // ':' is only reserved for relative URI paths, which clangd doesn't produce. |
| 101 | case ':': |
| 102 | return false; |
| 103 | } |
| 104 | return true; |
| 105 | } |
| 106 | |
| 107 | /// Encodes a string according to percent-encoding. |
| 108 | /// - Unreserved characters are not escaped. |
| 109 | /// - Reserved characters always escaped with exceptions like '/'. |
| 110 | /// - All other characters are escaped. |
| 111 | void percentEncode(llvm::StringRef Content, std::string &Out) { |
| 112 | for (unsigned char C : Content) |
| 113 | if (shouldEscape(C)) { |
| 114 | Out.push_back(c: '%'); |
| 115 | Out.push_back(c: llvm::hexdigit(X: C / 16)); |
| 116 | Out.push_back(c: llvm::hexdigit(X: C % 16)); |
| 117 | } else { |
| 118 | Out.push_back(c: C); |
| 119 | } |
| 120 | } |
| 121 | |
| 122 | /// Decodes a string according to percent-encoding. |
| 123 | std::string percentDecode(llvm::StringRef Content) { |
| 124 | std::string Result; |
| 125 | for (auto I = Content.begin(), E = Content.end(); I != E; ++I) { |
| 126 | if (*I != '%') { |
| 127 | Result += *I; |
| 128 | continue; |
| 129 | } |
| 130 | if (*I == '%' && I + 2 < Content.end() && llvm::isHexDigit(C: *(I + 1)) && |
| 131 | llvm::isHexDigit(C: *(I + 2))) { |
| 132 | Result.push_back(c: llvm::hexFromNibbles(MSB: *(I + 1), LSB: *(I + 2))); |
| 133 | I += 2; |
| 134 | } else |
| 135 | Result.push_back(c: *I); |
| 136 | } |
| 137 | return Result; |
| 138 | } |
| 139 | |
| 140 | bool isValidScheme(llvm::StringRef Scheme) { |
| 141 | if (Scheme.empty()) |
| 142 | return false; |
| 143 | if (!llvm::isAlpha(C: Scheme[0])) |
| 144 | return false; |
| 145 | return llvm::all_of(Range: llvm::drop_begin(RangeOrContainer&: Scheme), P: [](char C) { |
| 146 | return llvm::isAlnum(C) || C == '+' || C == '.' || C == '-'; |
| 147 | }); |
| 148 | } |
| 149 | |
| 150 | } // namespace |
| 151 | |
| 152 | URI::URI(llvm::StringRef Scheme, llvm::StringRef Authority, |
| 153 | llvm::StringRef Body) |
| 154 | : Scheme(Scheme), Authority(Authority), Body(Body) { |
| 155 | assert(!Scheme.empty()); |
| 156 | assert((Authority.empty() || Body.starts_with("/" )) && |
| 157 | "URI body must start with '/' when authority is present." ); |
| 158 | } |
| 159 | |
| 160 | std::string URI::toString() const { |
| 161 | std::string Result; |
| 162 | percentEncode(Content: Scheme, Out&: Result); |
| 163 | Result.push_back(c: ':'); |
| 164 | if (Authority.empty() && Body.empty()) |
| 165 | return Result; |
| 166 | // If authority if empty, we only print body if it starts with "/"; otherwise, |
| 167 | // the URI is invalid. |
| 168 | if (!Authority.empty() || llvm::StringRef(Body).starts_with(Prefix: "/" )) { |
| 169 | Result.append(s: "//" ); |
| 170 | percentEncode(Content: Authority, Out&: Result); |
| 171 | } |
| 172 | percentEncode(Content: Body, Out&: Result); |
| 173 | return Result; |
| 174 | } |
| 175 | |
| 176 | llvm::Expected<URI> URI::parse(llvm::StringRef OrigUri) { |
| 177 | URI U; |
| 178 | llvm::StringRef Uri = OrigUri; |
| 179 | |
| 180 | auto Pos = Uri.find(C: ':'); |
| 181 | if (Pos == llvm::StringRef::npos) |
| 182 | return error(Fmt: "Scheme must be provided in URI: {0}" , Vals&: OrigUri); |
| 183 | auto SchemeStr = Uri.substr(Start: 0, N: Pos); |
| 184 | U.Scheme = percentDecode(Content: SchemeStr); |
| 185 | if (!isValidScheme(Scheme: U.Scheme)) |
| 186 | return error(Fmt: "Invalid scheme: {0} (decoded: {1})" , Vals&: SchemeStr, Vals&: U.Scheme); |
| 187 | Uri = Uri.substr(Start: Pos + 1); |
| 188 | if (Uri.consume_front(Prefix: "//" )) { |
| 189 | Pos = Uri.find(C: '/'); |
| 190 | U.Authority = percentDecode(Content: Uri.substr(Start: 0, N: Pos)); |
| 191 | Uri = Uri.substr(Start: Pos); |
| 192 | } |
| 193 | U.Body = percentDecode(Content: Uri); |
| 194 | return U; |
| 195 | } |
| 196 | |
| 197 | llvm::Expected<std::string> URI::resolve(llvm::StringRef FileURI, |
| 198 | llvm::StringRef HintPath) { |
| 199 | auto Uri = URI::parse(OrigUri: FileURI); |
| 200 | if (!Uri) |
| 201 | return Uri.takeError(); |
| 202 | auto Path = URI::resolve(U: *Uri, HintPath); |
| 203 | if (!Path) |
| 204 | return Path.takeError(); |
| 205 | return *Path; |
| 206 | } |
| 207 | |
| 208 | llvm::Expected<URI> URI::create(llvm::StringRef AbsolutePath, |
| 209 | llvm::StringRef Scheme) { |
| 210 | if (!llvm::sys::path::is_absolute(path: AbsolutePath)) |
| 211 | return error(Fmt: "Not a valid absolute path: {0}" , Vals&: AbsolutePath); |
| 212 | auto S = findSchemeByName(Scheme); |
| 213 | if (!S) |
| 214 | return S.takeError(); |
| 215 | return S->get()->uriFromAbsolutePath(AbsolutePath); |
| 216 | } |
| 217 | |
| 218 | URI URI::create(llvm::StringRef AbsolutePath) { |
| 219 | if (!llvm::sys::path::is_absolute(path: AbsolutePath)) |
| 220 | llvm_unreachable( |
| 221 | ("Not a valid absolute path: " + AbsolutePath).str().c_str()); |
| 222 | for (auto &Entry : URISchemeRegistry::entries()) { |
| 223 | auto URI = Entry.instantiate()->uriFromAbsolutePath(AbsolutePath); |
| 224 | // For some paths, conversion to different URI schemes is impossible. These |
| 225 | // should be just skipped. |
| 226 | if (!URI) { |
| 227 | // Ignore the error. |
| 228 | llvm::consumeError(Err: URI.takeError()); |
| 229 | continue; |
| 230 | } |
| 231 | return std::move(*URI); |
| 232 | } |
| 233 | // Fallback to file: scheme which should work for any paths. |
| 234 | return URI::createFile(AbsolutePath); |
| 235 | } |
| 236 | |
| 237 | URI URI::createFile(llvm::StringRef AbsolutePath) { |
| 238 | auto U = FileSystemScheme().uriFromAbsolutePath(AbsolutePath); |
| 239 | if (!U) |
| 240 | llvm_unreachable(llvm::toString(U.takeError()).c_str()); |
| 241 | return std::move(*U); |
| 242 | } |
| 243 | |
| 244 | llvm::Expected<std::string> URI::resolve(const URI &Uri, |
| 245 | llvm::StringRef HintPath) { |
| 246 | auto S = findSchemeByName(Scheme: Uri.Scheme); |
| 247 | if (!S) |
| 248 | return S.takeError(); |
| 249 | return S->get()->getAbsolutePath(Authority: Uri.Authority, Body: Uri.Body, HintPath); |
| 250 | } |
| 251 | |
| 252 | llvm::Expected<std::string> URI::resolvePath(llvm::StringRef AbsPath, |
| 253 | llvm::StringRef HintPath) { |
| 254 | if (!llvm::sys::path::is_absolute(path: AbsPath)) |
| 255 | llvm_unreachable(("Not a valid absolute path: " + AbsPath).str().c_str()); |
| 256 | for (auto &Entry : URISchemeRegistry::entries()) { |
| 257 | auto S = Entry.instantiate(); |
| 258 | auto U = S->uriFromAbsolutePath(AbsolutePath: AbsPath); |
| 259 | // For some paths, conversion to different URI schemes is impossible. These |
| 260 | // should be just skipped. |
| 261 | if (!U) { |
| 262 | // Ignore the error. |
| 263 | llvm::consumeError(Err: U.takeError()); |
| 264 | continue; |
| 265 | } |
| 266 | return S->getAbsolutePath(Authority: U->Authority, Body: U->Body, HintPath); |
| 267 | } |
| 268 | // Fallback to file: scheme which doesn't do any canonicalization. |
| 269 | return std::string(AbsPath); |
| 270 | } |
| 271 | |
| 272 | llvm::Expected<std::string> URI::includeSpelling(const URI &Uri) { |
| 273 | auto S = findSchemeByName(Scheme: Uri.Scheme); |
| 274 | if (!S) |
| 275 | return S.takeError(); |
| 276 | return S->get()->getIncludeSpelling(U: Uri); |
| 277 | } |
| 278 | |
| 279 | } // namespace clangd |
| 280 | } // namespace clang |
| 281 | |