| 1 | //===-- Mustache.cpp ------------------------------------------------------===// |
| 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 | #include "llvm/Support/Mustache.h" |
| 9 | #include "llvm/ADT/SmallVector.h" |
| 10 | #include "llvm/Support/Debug.h" |
| 11 | #include "llvm/Support/raw_ostream.h" |
| 12 | #include <sstream> |
| 13 | |
| 14 | #define DEBUG_TYPE "mustache" |
| 15 | |
| 16 | using namespace llvm; |
| 17 | using namespace llvm::mustache; |
| 18 | |
| 19 | namespace { |
| 20 | |
| 21 | using Accessor = ArrayRef<StringRef>; |
| 22 | |
| 23 | static bool isFalsey(const json::Value &V) { |
| 24 | return V.getAsNull() || (V.getAsBoolean() && !V.getAsBoolean().value()) || |
| 25 | (V.getAsArray() && V.getAsArray()->empty()); |
| 26 | } |
| 27 | |
| 28 | static bool isContextFalsey(const json::Value *V) { |
| 29 | // A missing context (represented by a nullptr) is defined as falsey. |
| 30 | if (!V) |
| 31 | return true; |
| 32 | return isFalsey(V: *V); |
| 33 | } |
| 34 | |
| 35 | static void splitAndTrim(StringRef Str, SmallVectorImpl<StringRef> &Tokens) { |
| 36 | size_t CurrentPos = 0; |
| 37 | while (CurrentPos < Str.size()) { |
| 38 | // Find the next delimiter. |
| 39 | size_t DelimiterPos = Str.find(C: '.', From: CurrentPos); |
| 40 | |
| 41 | // If no delimiter is found, process the rest of the string. |
| 42 | if (DelimiterPos == StringRef::npos) |
| 43 | DelimiterPos = Str.size(); |
| 44 | |
| 45 | // Get the current part, which may have whitespace. |
| 46 | StringRef Part = Str.slice(Start: CurrentPos, End: DelimiterPos); |
| 47 | |
| 48 | // Manually trim the part without creating a new string object. |
| 49 | size_t Start = Part.find_first_not_of(Chars: " \t\r\n" ); |
| 50 | if (Start != StringRef::npos) { |
| 51 | size_t End = Part.find_last_not_of(Chars: " \t\r\n" ); |
| 52 | Tokens.push_back(Elt: Part.slice(Start, End: End + 1)); |
| 53 | } |
| 54 | |
| 55 | // Move past the delimiter for the next iteration. |
| 56 | CurrentPos = DelimiterPos + 1; |
| 57 | } |
| 58 | } |
| 59 | |
| 60 | static Accessor splitMustacheString(StringRef Str, MustacheContext &Ctx) { |
| 61 | // We split the mustache string into an accessor. |
| 62 | // For example: |
| 63 | // "a.b.c" would be split into {"a", "b", "c"} |
| 64 | // We make an exception for a single dot which |
| 65 | // refers to the current context. |
| 66 | SmallVector<StringRef> Tokens; |
| 67 | if (Str == "." ) { |
| 68 | // "." is a special accessor that refers to the current context. |
| 69 | // It's a literal, so it doesn't need to be saved. |
| 70 | Tokens.push_back(Elt: "." ); |
| 71 | } else { |
| 72 | splitAndTrim(Str, Tokens); |
| 73 | } |
| 74 | // Now, allocate memory for the array of StringRefs in the arena. |
| 75 | StringRef *ArenaTokens = Ctx.Allocator.Allocate<StringRef>(Num: Tokens.size()); |
| 76 | // Copy the StringRefs from the stack vector to the arena. |
| 77 | llvm::copy(Range&: Tokens, Out: ArenaTokens); |
| 78 | // Return an ArrayRef pointing to the stable arena memory. |
| 79 | return ArrayRef<StringRef>(ArenaTokens, Tokens.size()); |
| 80 | } |
| 81 | } // namespace |
| 82 | |
| 83 | namespace llvm::mustache { |
| 84 | |
| 85 | class MustacheOutputStream : public raw_ostream { |
| 86 | public: |
| 87 | MustacheOutputStream() = default; |
| 88 | ~MustacheOutputStream() override = default; |
| 89 | |
| 90 | virtual void suspendIndentation() {} |
| 91 | virtual void resumeIndentation() {} |
| 92 | |
| 93 | private: |
| 94 | void anchor() override; |
| 95 | }; |
| 96 | |
| 97 | void MustacheOutputStream::anchor() {} |
| 98 | |
| 99 | class RawMustacheOutputStream : public MustacheOutputStream { |
| 100 | public: |
| 101 | RawMustacheOutputStream(raw_ostream &OS) : OS(OS) { SetUnbuffered(); } |
| 102 | |
| 103 | private: |
| 104 | raw_ostream &OS; |
| 105 | |
| 106 | void write_impl(const char *Ptr, size_t Size) override { |
| 107 | OS.write(Ptr, Size); |
| 108 | } |
| 109 | uint64_t current_pos() const override { return OS.tell(); } |
| 110 | }; |
| 111 | |
| 112 | class Token { |
| 113 | public: |
| 114 | enum class Type { |
| 115 | Text, |
| 116 | Variable, |
| 117 | Partial, |
| 118 | SectionOpen, |
| 119 | SectionClose, |
| 120 | InvertSectionOpen, |
| 121 | UnescapeVariable, |
| 122 | , |
| 123 | SetDelimiter, |
| 124 | }; |
| 125 | |
| 126 | Token(StringRef Str) |
| 127 | : TokenType(Type::Text), RawBody(Str), TokenBody(RawBody), |
| 128 | AccessorValue({}), Indentation(0) {}; |
| 129 | |
| 130 | Token(StringRef RawBody, StringRef TokenBody, char Identifier, |
| 131 | MustacheContext &Ctx) |
| 132 | : RawBody(RawBody), TokenBody(TokenBody), Indentation(0) { |
| 133 | TokenType = getTokenType(Identifier); |
| 134 | if (TokenType == Type::Comment) |
| 135 | return; |
| 136 | StringRef AccessorStr(this->TokenBody); |
| 137 | if (TokenType != Type::Variable) |
| 138 | AccessorStr = AccessorStr.substr(Start: 1); |
| 139 | AccessorValue = splitMustacheString(Str: StringRef(AccessorStr).trim(), Ctx); |
| 140 | } |
| 141 | |
| 142 | ArrayRef<StringRef> getAccessor() const { return AccessorValue; } |
| 143 | |
| 144 | Type getType() const { return TokenType; } |
| 145 | |
| 146 | void setIndentation(size_t NewIndentation) { Indentation = NewIndentation; } |
| 147 | |
| 148 | size_t getIndentation() const { return Indentation; } |
| 149 | |
| 150 | static Type getTokenType(char Identifier) { |
| 151 | switch (Identifier) { |
| 152 | case '#': |
| 153 | return Type::SectionOpen; |
| 154 | case '/': |
| 155 | return Type::SectionClose; |
| 156 | case '^': |
| 157 | return Type::InvertSectionOpen; |
| 158 | case '!': |
| 159 | return Type::Comment; |
| 160 | case '>': |
| 161 | return Type::Partial; |
| 162 | case '&': |
| 163 | return Type::UnescapeVariable; |
| 164 | case '=': |
| 165 | return Type::SetDelimiter; |
| 166 | default: |
| 167 | return Type::Variable; |
| 168 | } |
| 169 | } |
| 170 | |
| 171 | Type TokenType; |
| 172 | // RawBody is the original string that was tokenized. |
| 173 | StringRef RawBody; |
| 174 | // TokenBody is the original string with the identifier removed. |
| 175 | StringRef TokenBody; |
| 176 | ArrayRef<StringRef> AccessorValue; |
| 177 | size_t Indentation; |
| 178 | }; |
| 179 | |
| 180 | using EscapeMap = DenseMap<char, std::string>; |
| 181 | |
| 182 | class ASTNode : public ilist_node<ASTNode> { |
| 183 | public: |
| 184 | enum Type { |
| 185 | Root, |
| 186 | Text, |
| 187 | Partial, |
| 188 | Variable, |
| 189 | UnescapeVariable, |
| 190 | Section, |
| 191 | InvertSection, |
| 192 | }; |
| 193 | |
| 194 | ASTNode(MustacheContext &Ctx) |
| 195 | : Ctx(Ctx), Ty(Type::Root), Parent(nullptr), ParentContext(nullptr) {} |
| 196 | |
| 197 | ASTNode(MustacheContext &Ctx, StringRef Body, ASTNode *Parent) |
| 198 | : Ctx(Ctx), Ty(Type::Text), Body(Body), Parent(Parent), |
| 199 | ParentContext(nullptr) {} |
| 200 | |
| 201 | // Constructor for Section/InvertSection/Variable/UnescapeVariable Nodes |
| 202 | ASTNode(MustacheContext &Ctx, Type Ty, ArrayRef<StringRef> Accessor, |
| 203 | ASTNode *Parent) |
| 204 | : Ctx(Ctx), Ty(Ty), Parent(Parent), AccessorValue(Accessor), |
| 205 | ParentContext(nullptr) {} |
| 206 | |
| 207 | void addChild(AstPtr Child) { Children.push_back(val: Child); }; |
| 208 | |
| 209 | void setRawBody(StringRef NewBody) { RawBody = NewBody; }; |
| 210 | |
| 211 | void setIndentation(size_t NewIndentation) { Indentation = NewIndentation; }; |
| 212 | |
| 213 | void render(const llvm::json::Value &Data, MustacheOutputStream &OS); |
| 214 | |
| 215 | private: |
| 216 | void renderLambdas(const llvm::json::Value &Contexts, |
| 217 | MustacheOutputStream &OS, Lambda &L); |
| 218 | |
| 219 | void renderSectionLambdas(const llvm::json::Value &Contexts, |
| 220 | MustacheOutputStream &OS, SectionLambda &L); |
| 221 | |
| 222 | void renderPartial(const llvm::json::Value &Contexts, |
| 223 | MustacheOutputStream &OS, ASTNode *Partial); |
| 224 | |
| 225 | void renderChild(const llvm::json::Value &Context, MustacheOutputStream &OS); |
| 226 | |
| 227 | const llvm::json::Value *findContext(); |
| 228 | |
| 229 | void renderRoot(const json::Value &CurrentCtx, MustacheOutputStream &OS); |
| 230 | void renderText(MustacheOutputStream &OS); |
| 231 | void renderPartial(const json::Value &CurrentCtx, MustacheOutputStream &OS); |
| 232 | void renderVariable(const json::Value &CurrentCtx, MustacheOutputStream &OS); |
| 233 | void renderUnescapeVariable(const json::Value &CurrentCtx, |
| 234 | MustacheOutputStream &OS); |
| 235 | void renderSection(const json::Value &CurrentCtx, MustacheOutputStream &OS); |
| 236 | void renderInvertSection(const json::Value &CurrentCtx, |
| 237 | MustacheOutputStream &OS); |
| 238 | |
| 239 | MustacheContext &Ctx; |
| 240 | Type Ty; |
| 241 | size_t Indentation = 0; |
| 242 | StringRef RawBody; |
| 243 | StringRef Body; |
| 244 | ASTNode *Parent; |
| 245 | ASTNodeList Children; |
| 246 | const ArrayRef<StringRef> AccessorValue; |
| 247 | const llvm::json::Value *ParentContext; |
| 248 | }; |
| 249 | |
| 250 | // A wrapper for arena allocator for ASTNodes |
| 251 | static AstPtr createRootNode(MustacheContext &Ctx) { |
| 252 | return new (Ctx.Allocator.Allocate<ASTNode>()) ASTNode(Ctx); |
| 253 | } |
| 254 | |
| 255 | static AstPtr createNode(MustacheContext &Ctx, ASTNode::Type T, |
| 256 | ArrayRef<StringRef> A, ASTNode *Parent) { |
| 257 | return new (Ctx.Allocator.Allocate<ASTNode>()) ASTNode(Ctx, T, A, Parent); |
| 258 | } |
| 259 | |
| 260 | static AstPtr createTextNode(MustacheContext &Ctx, StringRef Body, |
| 261 | ASTNode *Parent) { |
| 262 | return new (Ctx.Allocator.Allocate<ASTNode>()) ASTNode(Ctx, Body, Parent); |
| 263 | } |
| 264 | |
| 265 | // Function to check if there is meaningful text behind. |
| 266 | // We determine if a token has meaningful text behind |
| 267 | // if the right of previous token contains anything that is |
| 268 | // not a newline. |
| 269 | // For example: |
| 270 | // "Stuff {{#Section}}" (returns true) |
| 271 | // vs |
| 272 | // "{{#Section}} \n" (returns false) |
| 273 | // We make an exception for when previous token is empty |
| 274 | // and the current token is the second token. |
| 275 | // For example: |
| 276 | // "{{#Section}}" |
| 277 | static bool hasTextBehind(size_t Idx, const ArrayRef<Token> &Tokens) { |
| 278 | if (Idx == 0) |
| 279 | return true; |
| 280 | |
| 281 | size_t PrevIdx = Idx - 1; |
| 282 | if (Tokens[PrevIdx].getType() != Token::Type::Text) |
| 283 | return true; |
| 284 | |
| 285 | const Token &PrevToken = Tokens[PrevIdx]; |
| 286 | StringRef TokenBody = StringRef(PrevToken.RawBody).rtrim(Chars: " \r\t\v" ); |
| 287 | return !TokenBody.ends_with(Suffix: "\n" ) && !(TokenBody.empty() && Idx == 1); |
| 288 | } |
| 289 | |
| 290 | // Function to check if there's no meaningful text ahead. |
| 291 | // We determine if a token has text ahead if the left of previous |
| 292 | // token does not start with a newline. |
| 293 | static bool hasTextAhead(size_t Idx, const ArrayRef<Token> &Tokens) { |
| 294 | if (Idx >= Tokens.size() - 1) |
| 295 | return true; |
| 296 | |
| 297 | size_t NextIdx = Idx + 1; |
| 298 | if (Tokens[NextIdx].getType() != Token::Type::Text) |
| 299 | return true; |
| 300 | |
| 301 | const Token &NextToken = Tokens[NextIdx]; |
| 302 | StringRef TokenBody = StringRef(NextToken.RawBody).ltrim(Chars: " " ); |
| 303 | return !TokenBody.starts_with(Prefix: "\r\n" ) && !TokenBody.starts_with(Prefix: "\n" ); |
| 304 | } |
| 305 | |
| 306 | static bool requiresCleanUp(Token::Type T) { |
| 307 | // We must clean up all the tokens that could contain child nodes. |
| 308 | return T == Token::Type::SectionOpen || T == Token::Type::InvertSectionOpen || |
| 309 | T == Token::Type::SectionClose || T == Token::Type::Comment || |
| 310 | T == Token::Type::Partial || T == Token::Type::SetDelimiter; |
| 311 | } |
| 312 | |
| 313 | // Adjust next token body if there is no text ahead. |
| 314 | // For example: |
| 315 | // The template string |
| 316 | // "{{! Comment }} \nLine 2" |
| 317 | // would be considered as no text ahead and should be rendered as |
| 318 | // " Line 2" |
| 319 | static void (SmallVectorImpl<Token> &Tokens, size_t Idx) { |
| 320 | Token &NextToken = Tokens[Idx + 1]; |
| 321 | StringRef NextTokenBody = NextToken.TokenBody; |
| 322 | // Cut off the leading newline which could be \n or \r\n. |
| 323 | if (NextTokenBody.starts_with(Prefix: "\r\n" )) |
| 324 | NextToken.TokenBody = NextTokenBody.substr(Start: 2); |
| 325 | else if (NextTokenBody.starts_with(Prefix: "\n" )) |
| 326 | NextToken.TokenBody = NextTokenBody.substr(Start: 1); |
| 327 | } |
| 328 | |
| 329 | // Adjust previous token body if there no text behind. |
| 330 | // For example: |
| 331 | // The template string |
| 332 | // " \t{{#section}}A{{/section}}" |
| 333 | // would be considered as having no text ahead and would be render as: |
| 334 | // "A" |
| 335 | void stripTokenBefore(SmallVectorImpl<Token> &Tokens, size_t Idx, |
| 336 | Token &CurrentToken, Token::Type CurrentType) { |
| 337 | Token &PrevToken = Tokens[Idx - 1]; |
| 338 | StringRef PrevTokenBody = PrevToken.TokenBody; |
| 339 | StringRef Unindented = PrevTokenBody.rtrim(Chars: " \r\t\v" ); |
| 340 | size_t Indentation = PrevTokenBody.size() - Unindented.size(); |
| 341 | PrevToken.TokenBody = Unindented; |
| 342 | CurrentToken.setIndentation(Indentation); |
| 343 | } |
| 344 | |
| 345 | struct Tag { |
| 346 | enum class Kind { |
| 347 | None, |
| 348 | Normal, // {{...}} |
| 349 | Triple, // {{{...}}} |
| 350 | }; |
| 351 | |
| 352 | Kind TagKind = Kind::None; |
| 353 | StringRef Content; // The content between the delimiters. |
| 354 | StringRef FullMatch; // The entire tag, including delimiters. |
| 355 | size_t StartPosition = StringRef::npos; |
| 356 | }; |
| 357 | |
| 358 | [[maybe_unused]] static const char *tagKindToString(Tag::Kind K) { |
| 359 | switch (K) { |
| 360 | case Tag::Kind::None: |
| 361 | return "None" ; |
| 362 | case Tag::Kind::Normal: |
| 363 | return "Normal" ; |
| 364 | case Tag::Kind::Triple: |
| 365 | return "Triple" ; |
| 366 | } |
| 367 | llvm_unreachable("Unknown Tag::Kind" ); |
| 368 | } |
| 369 | |
| 370 | [[maybe_unused]] static const char *jsonKindToString(json::Value::Kind K) { |
| 371 | switch (K) { |
| 372 | case json::Value::Kind::Null: |
| 373 | return "JSON_KIND_NULL" ; |
| 374 | case json::Value::Kind::Boolean: |
| 375 | return "JSON_KIND_BOOLEAN" ; |
| 376 | case json::Value::Kind::Number: |
| 377 | return "JSON_KIND_NUMBER" ; |
| 378 | case json::Value::Kind::String: |
| 379 | return "JSON_KIND_STRING" ; |
| 380 | case json::Value::Kind::Array: |
| 381 | return "JSON_KIND_ARRAY" ; |
| 382 | case json::Value::Kind::Object: |
| 383 | return "JSON_KIND_OBJECT" ; |
| 384 | } |
| 385 | llvm_unreachable("Unknown json::Value::Kind" ); |
| 386 | } |
| 387 | |
| 388 | // Simple tokenizer that splits the template into tokens. |
| 389 | static SmallVector<Token> tokenize(StringRef Template, MustacheContext &Ctx) { |
| 390 | LLVM_DEBUG(dbgs() << "[Tokenize Template] \"" << Template << "\"\n" ); |
| 391 | SmallVector<Token> Tokens; |
| 392 | SmallString<8> Open("{{" ); |
| 393 | SmallString<8> Close("}}" ); |
| 394 | size_t Cursor = 0; |
| 395 | size_t TextStart = 0; |
| 396 | |
| 397 | const StringLiteral TripleOpen("{{{" ); |
| 398 | const StringLiteral TripleClose("}}}" ); |
| 399 | |
| 400 | while (Cursor < Template.size()) { |
| 401 | StringRef TemplateSuffix = Template.substr(Start: Cursor); |
| 402 | StringRef TagOpen, TagClose; |
| 403 | Tag::Kind Kind; |
| 404 | |
| 405 | // Determine which tag we've encountered. |
| 406 | if (TemplateSuffix.starts_with(Prefix: TripleOpen)) { |
| 407 | Kind = Tag::Kind::Triple; |
| 408 | TagOpen = TripleOpen; |
| 409 | TagClose = TripleClose; |
| 410 | } else if (TemplateSuffix.starts_with(Prefix: Open)) { |
| 411 | Kind = Tag::Kind::Normal; |
| 412 | TagOpen = Open; |
| 413 | TagClose = Close; |
| 414 | } else { |
| 415 | // Not at a tag, continue scanning. |
| 416 | ++Cursor; |
| 417 | continue; |
| 418 | } |
| 419 | |
| 420 | // Found a tag, first add the preceding text. |
| 421 | if (Cursor > TextStart) |
| 422 | Tokens.emplace_back(Args: Template.slice(Start: TextStart, End: Cursor)); |
| 423 | |
| 424 | // Find the closing tag. |
| 425 | size_t EndPos = Template.find(Str: TagClose, From: Cursor + TagOpen.size()); |
| 426 | if (EndPos == StringRef::npos) { |
| 427 | // No closing tag, the rest is text. |
| 428 | Tokens.emplace_back(Args: Template.substr(Start: Cursor)); |
| 429 | TextStart = Cursor = Template.size(); |
| 430 | break; |
| 431 | } |
| 432 | |
| 433 | // Extract tag content and full match. |
| 434 | size_t ContentStart = Cursor + TagOpen.size(); |
| 435 | StringRef Content = Template.substr(Start: ContentStart, N: EndPos - ContentStart); |
| 436 | StringRef FullMatch = |
| 437 | Template.substr(Start: Cursor, N: (EndPos + TagClose.size()) - Cursor); |
| 438 | |
| 439 | // Process the tag (inlined logic from processTag). |
| 440 | LLVM_DEBUG(dbgs() << "[Tag] " << FullMatch << ", Content: " << Content |
| 441 | << ", Kind: " << tagKindToString(Kind) << "\n" ); |
| 442 | if (Kind == Tag::Kind::Triple) { |
| 443 | Tokens.emplace_back(Args&: FullMatch, Args: Ctx.Saver.save(S: "&" + Content), Args: '&', Args&: Ctx); |
| 444 | } else { // Normal Tag |
| 445 | StringRef Interpolated = Content; |
| 446 | if (!Interpolated.trim().starts_with(Prefix: "=" )) { |
| 447 | char Front = Interpolated.empty() ? ' ' : Interpolated.trim().front(); |
| 448 | Tokens.emplace_back(Args&: FullMatch, Args&: Interpolated, Args&: Front, Args&: Ctx); |
| 449 | } else { // Set Delimiter |
| 450 | Tokens.emplace_back(Args&: FullMatch, Args&: Interpolated, Args: '=', Args&: Ctx); |
| 451 | StringRef DelimSpec = Interpolated.trim(); |
| 452 | DelimSpec = DelimSpec.drop_front(N: 1); |
| 453 | DelimSpec = DelimSpec.take_until(F: [](char C) { return C == '='; }); |
| 454 | DelimSpec = DelimSpec.trim(); |
| 455 | |
| 456 | auto [NewOpen, NewClose] = DelimSpec.split(Separator: ' '); |
| 457 | LLVM_DEBUG(dbgs() << "[Set Delimiter] NewOpen: " << NewOpen |
| 458 | << ", NewClose: " << NewClose << "\n" ); |
| 459 | Open = NewOpen; |
| 460 | Close = NewClose; |
| 461 | } |
| 462 | } |
| 463 | |
| 464 | // Move past the tag for the next iteration. |
| 465 | Cursor += FullMatch.size(); |
| 466 | TextStart = Cursor; |
| 467 | } |
| 468 | |
| 469 | // Add any remaining text after the last tag. |
| 470 | if (TextStart < Template.size()) |
| 471 | Tokens.emplace_back(Args: Template.substr(Start: TextStart)); |
| 472 | |
| 473 | // Fix up white spaces for standalone tags. |
| 474 | size_t LastIdx = Tokens.size() - 1; |
| 475 | for (size_t Idx = 0, End = Tokens.size(); Idx < End; ++Idx) { |
| 476 | Token &CurrentToken = Tokens[Idx]; |
| 477 | Token::Type CurrentType = CurrentToken.getType(); |
| 478 | if (!requiresCleanUp(T: CurrentType)) |
| 479 | continue; |
| 480 | |
| 481 | bool HasTextBehind = hasTextBehind(Idx, Tokens); |
| 482 | bool HasTextAhead = hasTextAhead(Idx, Tokens); |
| 483 | |
| 484 | if ((!HasTextAhead && !HasTextBehind) || (!HasTextAhead && Idx == 0)) |
| 485 | stripTokenAhead(Tokens, Idx); |
| 486 | |
| 487 | if ((!HasTextBehind && !HasTextAhead) || (!HasTextBehind && Idx == LastIdx)) |
| 488 | stripTokenBefore(Tokens, Idx, CurrentToken, CurrentType); |
| 489 | } |
| 490 | return Tokens; |
| 491 | } |
| 492 | |
| 493 | // Custom stream to escape strings. |
| 494 | class EscapeStringStream : public MustacheOutputStream { |
| 495 | public: |
| 496 | explicit EscapeStringStream(llvm::raw_ostream &WrappedStream, |
| 497 | EscapeMap &Escape) |
| 498 | : Escape(Escape), EscapeChars(Escape.keys().begin(), Escape.keys().end()), |
| 499 | WrappedStream(WrappedStream) { |
| 500 | SetUnbuffered(); |
| 501 | } |
| 502 | |
| 503 | protected: |
| 504 | void write_impl(const char *Ptr, size_t Size) override { |
| 505 | StringRef Data(Ptr, Size); |
| 506 | size_t Start = 0; |
| 507 | while (Start < Size) { |
| 508 | // Find the next character that needs to be escaped. |
| 509 | size_t Next = Data.find_first_of(Chars: EscapeChars.str(), From: Start); |
| 510 | |
| 511 | // If no escapable characters are found, write the rest of the string. |
| 512 | if (Next == StringRef::npos) { |
| 513 | WrappedStream << Data.substr(Start); |
| 514 | return; |
| 515 | } |
| 516 | |
| 517 | // Write the chunk of text before the escapable character. |
| 518 | if (Next > Start) |
| 519 | WrappedStream << Data.substr(Start, N: Next - Start); |
| 520 | |
| 521 | // Look up and write the escaped version of the character. |
| 522 | WrappedStream << Escape[Data[Next]]; |
| 523 | Start = Next + 1; |
| 524 | } |
| 525 | } |
| 526 | |
| 527 | uint64_t current_pos() const override { return WrappedStream.tell(); } |
| 528 | |
| 529 | private: |
| 530 | EscapeMap &Escape; |
| 531 | SmallString<8> EscapeChars; |
| 532 | llvm::raw_ostream &WrappedStream; |
| 533 | }; |
| 534 | |
| 535 | // Custom stream to add indentation used to for rendering partials. |
| 536 | class AddIndentationStringStream : public MustacheOutputStream { |
| 537 | public: |
| 538 | explicit AddIndentationStringStream(raw_ostream &WrappedStream, |
| 539 | size_t Indentation) |
| 540 | : Indentation(Indentation), WrappedStream(WrappedStream), |
| 541 | NeedsIndent(true), IsSuspended(false) { |
| 542 | SetUnbuffered(); |
| 543 | } |
| 544 | |
| 545 | void suspendIndentation() override { IsSuspended = true; } |
| 546 | void resumeIndentation() override { IsSuspended = false; } |
| 547 | |
| 548 | protected: |
| 549 | void write_impl(const char *Ptr, size_t Size) override { |
| 550 | llvm::StringRef Data(Ptr, Size); |
| 551 | SmallString<0> Indent; |
| 552 | Indent.resize(N: Indentation, NV: ' '); |
| 553 | |
| 554 | for (char C : Data) { |
| 555 | LLVM_DEBUG(dbgs() << "[Indentation Stream] NeedsIndent:" << NeedsIndent |
| 556 | << ", C:'" << C << "', Indentation:" << Indentation |
| 557 | << "\n" ); |
| 558 | if (NeedsIndent && C != '\n') { |
| 559 | WrappedStream << Indent; |
| 560 | NeedsIndent = false; |
| 561 | } |
| 562 | WrappedStream << C; |
| 563 | if (C == '\n' && !IsSuspended) |
| 564 | NeedsIndent = true; |
| 565 | } |
| 566 | } |
| 567 | |
| 568 | uint64_t current_pos() const override { return WrappedStream.tell(); } |
| 569 | |
| 570 | private: |
| 571 | size_t Indentation; |
| 572 | raw_ostream &WrappedStream; |
| 573 | bool NeedsIndent; |
| 574 | bool IsSuspended; |
| 575 | }; |
| 576 | |
| 577 | class Parser { |
| 578 | public: |
| 579 | Parser(StringRef TemplateStr, MustacheContext &Ctx) |
| 580 | : Ctx(Ctx), TemplateStr(TemplateStr) {} |
| 581 | |
| 582 | AstPtr parse(); |
| 583 | |
| 584 | private: |
| 585 | void parseMustache(ASTNode *Parent); |
| 586 | void parseSection(ASTNode *Parent, ASTNode::Type Ty, const Accessor &A); |
| 587 | |
| 588 | MustacheContext &Ctx; |
| 589 | SmallVector<Token> Tokens; |
| 590 | size_t CurrentPtr; |
| 591 | StringRef TemplateStr; |
| 592 | }; |
| 593 | |
| 594 | void Parser::parseSection(ASTNode *Parent, ASTNode::Type Ty, |
| 595 | const Accessor &A) { |
| 596 | AstPtr CurrentNode = createNode(Ctx, T: Ty, A, Parent); |
| 597 | size_t Start = CurrentPtr; |
| 598 | parseMustache(Parent: CurrentNode); |
| 599 | const size_t End = CurrentPtr - 1; |
| 600 | |
| 601 | size_t RawBodySize = 0; |
| 602 | for (size_t I = Start; I < End; ++I) |
| 603 | RawBodySize += Tokens[I].RawBody.size(); |
| 604 | |
| 605 | SmallString<128> RawBody; |
| 606 | RawBody.reserve(N: RawBodySize); |
| 607 | for (std::size_t I = Start; I < End; ++I) |
| 608 | RawBody += Tokens[I].RawBody; |
| 609 | |
| 610 | CurrentNode->setRawBody(Ctx.Saver.save(S: StringRef(RawBody))); |
| 611 | Parent->addChild(Child: CurrentNode); |
| 612 | } |
| 613 | |
| 614 | AstPtr Parser::parse() { |
| 615 | Tokens = tokenize(Template: TemplateStr, Ctx); |
| 616 | CurrentPtr = 0; |
| 617 | AstPtr RootNode = createRootNode(Ctx); |
| 618 | parseMustache(Parent: RootNode); |
| 619 | return RootNode; |
| 620 | } |
| 621 | |
| 622 | void Parser::parseMustache(ASTNode *Parent) { |
| 623 | |
| 624 | while (CurrentPtr < Tokens.size()) { |
| 625 | Token CurrentToken = Tokens[CurrentPtr]; |
| 626 | CurrentPtr++; |
| 627 | ArrayRef<StringRef> A = CurrentToken.getAccessor(); |
| 628 | AstPtr CurrentNode; |
| 629 | |
| 630 | switch (CurrentToken.getType()) { |
| 631 | case Token::Type::Text: { |
| 632 | CurrentNode = createTextNode(Ctx, Body: CurrentToken.TokenBody, Parent); |
| 633 | Parent->addChild(Child: CurrentNode); |
| 634 | break; |
| 635 | } |
| 636 | case Token::Type::Variable: { |
| 637 | CurrentNode = createNode(Ctx, T: ASTNode::Variable, A, Parent); |
| 638 | Parent->addChild(Child: CurrentNode); |
| 639 | break; |
| 640 | } |
| 641 | case Token::Type::UnescapeVariable: { |
| 642 | CurrentNode = createNode(Ctx, T: ASTNode::UnescapeVariable, A, Parent); |
| 643 | Parent->addChild(Child: CurrentNode); |
| 644 | break; |
| 645 | } |
| 646 | case Token::Type::Partial: { |
| 647 | CurrentNode = createNode(Ctx, T: ASTNode::Partial, A, Parent); |
| 648 | CurrentNode->setIndentation(CurrentToken.getIndentation()); |
| 649 | Parent->addChild(Child: CurrentNode); |
| 650 | break; |
| 651 | } |
| 652 | case Token::Type::SectionOpen: { |
| 653 | parseSection(Parent, Ty: ASTNode::Section, A); |
| 654 | break; |
| 655 | } |
| 656 | case Token::Type::InvertSectionOpen: { |
| 657 | parseSection(Parent, Ty: ASTNode::InvertSection, A); |
| 658 | break; |
| 659 | } |
| 660 | case Token::Type::Comment: |
| 661 | case Token::Type::SetDelimiter: |
| 662 | break; |
| 663 | case Token::Type::SectionClose: |
| 664 | return; |
| 665 | } |
| 666 | } |
| 667 | } |
| 668 | static void toMustacheString(const json::Value &Data, raw_ostream &OS) { |
| 669 | LLVM_DEBUG(dbgs() << "[To Mustache String] Kind: " |
| 670 | << jsonKindToString(Data.kind()) << ", Data: " << Data |
| 671 | << "\n" ); |
| 672 | switch (Data.kind()) { |
| 673 | case json::Value::Null: |
| 674 | return; |
| 675 | case json::Value::Number: { |
| 676 | auto Num = *Data.getAsNumber(); |
| 677 | std::ostringstream SS; |
| 678 | SS << Num; |
| 679 | OS << SS.str(); |
| 680 | return; |
| 681 | } |
| 682 | case json::Value::String: { |
| 683 | OS << *Data.getAsString(); |
| 684 | return; |
| 685 | } |
| 686 | |
| 687 | case json::Value::Array: { |
| 688 | auto Arr = *Data.getAsArray(); |
| 689 | if (Arr.empty()) |
| 690 | return; |
| 691 | [[fallthrough]]; |
| 692 | } |
| 693 | case json::Value::Object: |
| 694 | case json::Value::Boolean: { |
| 695 | llvm::json::OStream JOS(OS, 2); |
| 696 | JOS.value(V: Data); |
| 697 | break; |
| 698 | } |
| 699 | } |
| 700 | } |
| 701 | |
| 702 | void ASTNode::renderRoot(const json::Value &CurrentCtx, |
| 703 | MustacheOutputStream &OS) { |
| 704 | renderChild(Context: CurrentCtx, OS); |
| 705 | } |
| 706 | |
| 707 | void ASTNode::renderText(MustacheOutputStream &OS) { OS << Body; } |
| 708 | |
| 709 | void ASTNode::renderPartial(const json::Value &CurrentCtx, |
| 710 | MustacheOutputStream &OS) { |
| 711 | LLVM_DEBUG(dbgs() << "[Render Partial] Accessor:" << AccessorValue[0] |
| 712 | << ", Indentation:" << Indentation << "\n" ); |
| 713 | auto Partial = Ctx.Partials.find(Key: AccessorValue[0]); |
| 714 | if (Partial != Ctx.Partials.end()) |
| 715 | renderPartial(Contexts: CurrentCtx, OS, Partial: Partial->getValue()); |
| 716 | } |
| 717 | |
| 718 | void ASTNode::renderVariable(const json::Value &CurrentCtx, |
| 719 | MustacheOutputStream &OS) { |
| 720 | auto Lambda = Ctx.Lambdas.find(Key: AccessorValue[0]); |
| 721 | if (Lambda != Ctx.Lambdas.end()) { |
| 722 | renderLambdas(Contexts: CurrentCtx, OS, L&: Lambda->getValue()); |
| 723 | } else if (const json::Value *ContextPtr = findContext()) { |
| 724 | EscapeStringStream ES(OS, Ctx.Escapes); |
| 725 | toMustacheString(Data: *ContextPtr, OS&: ES); |
| 726 | } |
| 727 | } |
| 728 | |
| 729 | void ASTNode::renderUnescapeVariable(const json::Value &CurrentCtx, |
| 730 | MustacheOutputStream &OS) { |
| 731 | LLVM_DEBUG(dbgs() << "[Render UnescapeVariable] Accessor:" << AccessorValue[0] |
| 732 | << "\n" ); |
| 733 | auto Lambda = Ctx.Lambdas.find(Key: AccessorValue[0]); |
| 734 | if (Lambda != Ctx.Lambdas.end()) { |
| 735 | renderLambdas(Contexts: CurrentCtx, OS, L&: Lambda->getValue()); |
| 736 | } else if (const json::Value *ContextPtr = findContext()) { |
| 737 | OS.suspendIndentation(); |
| 738 | toMustacheString(Data: *ContextPtr, OS); |
| 739 | OS.resumeIndentation(); |
| 740 | } |
| 741 | } |
| 742 | |
| 743 | void ASTNode::renderSection(const json::Value &CurrentCtx, |
| 744 | MustacheOutputStream &OS) { |
| 745 | auto SectionLambda = Ctx.SectionLambdas.find(Key: AccessorValue[0]); |
| 746 | if (SectionLambda != Ctx.SectionLambdas.end()) { |
| 747 | renderSectionLambdas(Contexts: CurrentCtx, OS, L&: SectionLambda->getValue()); |
| 748 | return; |
| 749 | } |
| 750 | |
| 751 | const json::Value *ContextPtr = findContext(); |
| 752 | if (isContextFalsey(V: ContextPtr)) |
| 753 | return; |
| 754 | |
| 755 | if (const json::Array *Arr = ContextPtr->getAsArray()) { |
| 756 | for (const json::Value &V : *Arr) |
| 757 | renderChild(Context: V, OS); |
| 758 | return; |
| 759 | } |
| 760 | renderChild(Context: *ContextPtr, OS); |
| 761 | } |
| 762 | |
| 763 | void ASTNode::renderInvertSection(const json::Value &CurrentCtx, |
| 764 | MustacheOutputStream &OS) { |
| 765 | bool IsLambda = Ctx.SectionLambdas.contains(Key: AccessorValue[0]); |
| 766 | const json::Value *ContextPtr = findContext(); |
| 767 | if (isContextFalsey(V: ContextPtr) && !IsLambda) { |
| 768 | renderChild(Context: CurrentCtx, OS); |
| 769 | } |
| 770 | } |
| 771 | |
| 772 | void ASTNode::render(const llvm::json::Value &Data, MustacheOutputStream &OS) { |
| 773 | if (Ty != Root && Ty != Text && AccessorValue.empty()) |
| 774 | return; |
| 775 | // Set the parent context to the incoming context so that we |
| 776 | // can walk up the context tree correctly in findContext(). |
| 777 | ParentContext = &Data; |
| 778 | |
| 779 | switch (Ty) { |
| 780 | case Root: |
| 781 | renderRoot(CurrentCtx: Data, OS); |
| 782 | return; |
| 783 | case Text: |
| 784 | renderText(OS); |
| 785 | return; |
| 786 | case Partial: |
| 787 | renderPartial(CurrentCtx: Data, OS); |
| 788 | return; |
| 789 | case Variable: |
| 790 | renderVariable(CurrentCtx: Data, OS); |
| 791 | return; |
| 792 | case UnescapeVariable: |
| 793 | renderUnescapeVariable(CurrentCtx: Data, OS); |
| 794 | return; |
| 795 | case Section: |
| 796 | renderSection(CurrentCtx: Data, OS); |
| 797 | return; |
| 798 | case InvertSection: |
| 799 | renderInvertSection(CurrentCtx: Data, OS); |
| 800 | return; |
| 801 | } |
| 802 | llvm_unreachable("Invalid ASTNode type" ); |
| 803 | } |
| 804 | |
| 805 | const json::Value *ASTNode::findContext() { |
| 806 | // The mustache spec allows for dot notation to access nested values |
| 807 | // a single dot refers to the current context. |
| 808 | // We attempt to find the JSON context in the current node, if it is not |
| 809 | // found, then we traverse the parent nodes to find the context until we |
| 810 | // reach the root node or the context is found. |
| 811 | if (AccessorValue.empty()) |
| 812 | return nullptr; |
| 813 | if (AccessorValue[0] == "." ) |
| 814 | return ParentContext; |
| 815 | |
| 816 | const json::Object *CurrentContext = ParentContext->getAsObject(); |
| 817 | StringRef CurrentAccessor = AccessorValue[0]; |
| 818 | ASTNode *CurrentParent = Parent; |
| 819 | |
| 820 | while (!CurrentContext || !CurrentContext->get(K: CurrentAccessor)) { |
| 821 | if (CurrentParent->Ty != Root) { |
| 822 | CurrentContext = CurrentParent->ParentContext->getAsObject(); |
| 823 | CurrentParent = CurrentParent->Parent; |
| 824 | continue; |
| 825 | } |
| 826 | return nullptr; |
| 827 | } |
| 828 | const json::Value *Context = nullptr; |
| 829 | for (auto [Idx, Acc] : enumerate(First: AccessorValue)) { |
| 830 | const json::Value *CurrentValue = CurrentContext->get(K: Acc); |
| 831 | if (!CurrentValue) |
| 832 | return nullptr; |
| 833 | if (Idx < AccessorValue.size() - 1) { |
| 834 | CurrentContext = CurrentValue->getAsObject(); |
| 835 | if (!CurrentContext) |
| 836 | return nullptr; |
| 837 | } else { |
| 838 | Context = CurrentValue; |
| 839 | } |
| 840 | } |
| 841 | return Context; |
| 842 | } |
| 843 | |
| 844 | void ASTNode::renderChild(const json::Value &Contexts, |
| 845 | MustacheOutputStream &OS) { |
| 846 | for (ASTNode &Child : Children) |
| 847 | Child.render(Data: Contexts, OS); |
| 848 | } |
| 849 | |
| 850 | void ASTNode::renderPartial(const json::Value &Contexts, |
| 851 | MustacheOutputStream &OS, ASTNode *Partial) { |
| 852 | LLVM_DEBUG(dbgs() << "[Render Partial Indentation] Indentation: " << Indentation << "\n" ); |
| 853 | AddIndentationStringStream IS(OS, Indentation); |
| 854 | Partial->render(Data: Contexts, OS&: IS); |
| 855 | } |
| 856 | |
| 857 | void ASTNode::renderLambdas(const llvm::json::Value &Contexts, |
| 858 | MustacheOutputStream &OS, Lambda &L) { |
| 859 | json::Value LambdaResult = L(); |
| 860 | std::string LambdaStr; |
| 861 | raw_string_ostream Output(LambdaStr); |
| 862 | toMustacheString(Data: LambdaResult, OS&: Output); |
| 863 | Parser P(LambdaStr, Ctx); |
| 864 | AstPtr LambdaNode = P.parse(); |
| 865 | |
| 866 | EscapeStringStream ES(OS, Ctx.Escapes); |
| 867 | if (Ty == Variable) { |
| 868 | LambdaNode->render(Data: Contexts, OS&: ES); |
| 869 | return; |
| 870 | } |
| 871 | LambdaNode->render(Data: Contexts, OS); |
| 872 | } |
| 873 | |
| 874 | void ASTNode::renderSectionLambdas(const llvm::json::Value &Contexts, |
| 875 | MustacheOutputStream &OS, SectionLambda &L) { |
| 876 | json::Value Return = L(RawBody.str()); |
| 877 | if (isFalsey(V: Return)) |
| 878 | return; |
| 879 | std::string LambdaStr; |
| 880 | raw_string_ostream Output(LambdaStr); |
| 881 | toMustacheString(Data: Return, OS&: Output); |
| 882 | Parser P(LambdaStr, Ctx); |
| 883 | AstPtr LambdaNode = P.parse(); |
| 884 | LambdaNode->render(Data: Contexts, OS); |
| 885 | } |
| 886 | |
| 887 | void Template::render(const llvm::json::Value &Data, llvm::raw_ostream &OS) { |
| 888 | RawMustacheOutputStream MOS(OS); |
| 889 | Tree->render(Data, OS&: MOS); |
| 890 | } |
| 891 | |
| 892 | void Template::registerPartial(std::string Name, std::string Partial) { |
| 893 | StringRef SavedPartial = Ctx.Saver.save(S: Partial); |
| 894 | Parser P(SavedPartial, Ctx); |
| 895 | AstPtr PartialTree = P.parse(); |
| 896 | Ctx.Partials.insert(KV: std::make_pair(x&: Name, y&: PartialTree)); |
| 897 | } |
| 898 | |
| 899 | void Template::registerLambda(std::string Name, Lambda L) { |
| 900 | Ctx.Lambdas[Name] = std::move(L); |
| 901 | } |
| 902 | |
| 903 | void Template::registerLambda(std::string Name, SectionLambda L) { |
| 904 | Ctx.SectionLambdas[Name] = std::move(L); |
| 905 | } |
| 906 | |
| 907 | void Template::overrideEscapeCharacters(EscapeMap E) { |
| 908 | Ctx.Escapes = std::move(E); |
| 909 | } |
| 910 | |
| 911 | Template::Template(StringRef TemplateStr, MustacheContext &Ctx) : Ctx(Ctx) { |
| 912 | Parser P(TemplateStr, Ctx); |
| 913 | Tree = P.parse(); |
| 914 | // The default behavior is to escape html entities. |
| 915 | const EscapeMap HtmlEntities = {{'&', "&" }, |
| 916 | {'<', "<" }, |
| 917 | {'>', ">" }, |
| 918 | {'"', """ }, |
| 919 | {'\'', "'" }}; |
| 920 | overrideEscapeCharacters(E: HtmlEntities); |
| 921 | } |
| 922 | |
| 923 | Template::Template(Template &&Other) noexcept |
| 924 | : Ctx(Other.Ctx), Tree(Other.Tree) { |
| 925 | Other.Tree = nullptr; |
| 926 | } |
| 927 | |
| 928 | Template::~Template() = default; |
| 929 | |
| 930 | } // namespace llvm::mustache |
| 931 | |
| 932 | #undef DEBUG_TYPE |
| 933 | |