1 | // Protocol Buffers - Google's data interchange format |
2 | // Copyright 2008 Google Inc. All rights reserved. |
3 | // https://developers.google.com/protocol-buffers/ |
4 | // |
5 | // Redistribution and use in source and binary forms, with or without |
6 | // modification, are permitted provided that the following conditions are |
7 | // met: |
8 | // |
9 | // * Redistributions of source code must retain the above copyright |
10 | // notice, this list of conditions and the following disclaimer. |
11 | // * Redistributions in binary form must reproduce the above |
12 | // copyright notice, this list of conditions and the following disclaimer |
13 | // in the documentation and/or other materials provided with the |
14 | // distribution. |
15 | // * Neither the name of Google Inc. nor the names of its |
16 | // contributors may be used to endorse or promote products derived from |
17 | // this software without specific prior written permission. |
18 | // |
19 | // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
20 | // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
21 | // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR |
22 | // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT |
23 | // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, |
24 | // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT |
25 | // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
26 | // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
27 | // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
28 | // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
29 | // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
30 | |
31 | #include <google/protobuf/util/internal/json_stream_parser.h> |
32 | |
33 | #include <algorithm> |
34 | #include <cctype> |
35 | #include <cmath> |
36 | #include <memory> |
37 | #include <stack> |
38 | #include <string> |
39 | |
40 | #include <google/protobuf/stubs/common.h> |
41 | #include <google/protobuf/stubs/logging.h> |
42 | #include <google/protobuf/stubs/strutil.h> |
43 | #include <google/protobuf/stubs/status.h> |
44 | #include <google/protobuf/util/internal/object_writer.h> |
45 | #include <google/protobuf/util/internal/json_escaping.h> |
46 | |
47 | |
48 | namespace google { |
49 | namespace protobuf { |
50 | namespace util { |
51 | |
52 | namespace converter { |
53 | |
54 | // Number of digits in an escaped UTF-16 code unit ('\\' 'u' X X X X) |
55 | static const int kUnicodeEscapedLength = 6; |
56 | |
57 | static const int kDefaultMaxRecursionDepth = 100; |
58 | |
59 | // These cannot be constexpr for portability with VS2015. |
60 | static const StringPiece kKeywordTrue = "true" ; |
61 | static const StringPiece kKeywordFalse = "false" ; |
62 | static const StringPiece kKeywordNull = "null" ; |
63 | |
64 | inline bool IsLetter(char c) { |
65 | return ('a' <= c && c <= 'z') || ('A' <= c && c <= 'Z') || (c == '_') || |
66 | (c == '$'); |
67 | } |
68 | |
69 | inline bool IsAlphanumeric(char c) { |
70 | return IsLetter(c) || ('0' <= c && c <= '9'); |
71 | } |
72 | |
73 | // Indicates a character may not be part of an unquoted key. |
74 | inline bool IsKeySeparator(char c) { |
75 | return (ascii_isspace(c) || c == '"' || c == '\'' || c == '{' || |
76 | c == '}' || c == '[' || c == ']' || c == ':' || c == ','); |
77 | } |
78 | |
79 | inline void ReplaceInvalidCodePoints(StringPiece str, |
80 | const std::string& replacement, |
81 | std::string* dst) { |
82 | while (!str.empty()) { |
83 | int n_valid_bytes = internal::UTF8SpnStructurallyValid(str); |
84 | StringPiece valid_part = str.substr(pos: 0, n: n_valid_bytes); |
85 | StrAppend(dest: dst, a: valid_part); |
86 | |
87 | if (n_valid_bytes == str.size()) { |
88 | break; |
89 | } |
90 | |
91 | // Append replacement value. |
92 | StrAppend(dest: dst, a: replacement); |
93 | |
94 | // Move past valid bytes + one invalid byte. |
95 | str.remove_prefix(n: n_valid_bytes + 1); |
96 | } |
97 | } |
98 | |
99 | static bool ConsumeKey(StringPiece* input, StringPiece* key) { |
100 | if (input->empty() || !IsLetter(c: (*input)[0])) return false; |
101 | int len = 1; |
102 | for (; len < input->size(); ++len) { |
103 | if (!IsAlphanumeric(c: (*input)[len])) { |
104 | break; |
105 | } |
106 | } |
107 | *key = StringPiece(input->data(), len); |
108 | *input = StringPiece(input->data() + len, input->size() - len); |
109 | return true; |
110 | } |
111 | |
112 | // Same as 'ConsumeKey', but allows a widened set of key characters. |
113 | static bool ConsumeKeyPermissive(StringPiece* input, |
114 | StringPiece* key) { |
115 | if (input->empty() || !IsLetter(c: (*input)[0])) return false; |
116 | int len = 1; |
117 | for (; len < input->size(); ++len) { |
118 | if (IsKeySeparator(c: (*input)[len])) { |
119 | break; |
120 | } |
121 | } |
122 | *key = StringPiece(input->data(), len); |
123 | *input = StringPiece(input->data() + len, input->size() - len); |
124 | return true; |
125 | } |
126 | |
127 | static bool MatchKey(StringPiece input) { |
128 | return !input.empty() && IsLetter(c: input[0]); |
129 | } |
130 | |
131 | JsonStreamParser::JsonStreamParser(ObjectWriter* ow) |
132 | : ow_(ow), |
133 | stack_(), |
134 | leftover_(), |
135 | json_(), |
136 | p_(), |
137 | key_(), |
138 | key_storage_(), |
139 | finishing_(false), |
140 | seen_non_whitespace_(false), |
141 | allow_no_root_element_(false), |
142 | parsed_(), |
143 | parsed_storage_(), |
144 | string_open_(0), |
145 | chunk_storage_(), |
146 | coerce_to_utf8_(false), |
147 | utf8_replacement_character_(" " ), |
148 | allow_empty_null_(false), |
149 | allow_permissive_key_naming_(false), |
150 | loose_float_number_conversion_(false), |
151 | recursion_depth_(0), |
152 | max_recursion_depth_(kDefaultMaxRecursionDepth) { |
153 | // Initialize the stack with a single value to be parsed. |
154 | stack_.push(x: VALUE); |
155 | } |
156 | |
157 | JsonStreamParser::~JsonStreamParser() {} |
158 | |
159 | |
160 | util::Status JsonStreamParser::Parse(StringPiece json) { |
161 | StringPiece chunk = json; |
162 | // If we have leftovers from a previous chunk, append the new chunk to it |
163 | // and create a new StringPiece pointing at the string's data. This could |
164 | // be large but we rely on the chunks to be small, assuming they are |
165 | // fragments of a Cord. |
166 | if (!leftover_.empty()) { |
167 | // Don't point chunk to leftover_ because leftover_ will be updated in |
168 | // ParseChunk(chunk). |
169 | chunk_storage_.swap(s&: leftover_); |
170 | StrAppend(dest: &chunk_storage_, a: json); |
171 | chunk = StringPiece(chunk_storage_); |
172 | } |
173 | |
174 | // Find the structurally valid UTF8 prefix and parse only that. |
175 | int n = internal::UTF8SpnStructurallyValid(str: chunk); |
176 | if (n > 0) { |
177 | util::Status status = ParseChunk(chunk: chunk.substr(pos: 0, n)); |
178 | |
179 | // Any leftover characters are stashed in leftover_ for later parsing when |
180 | // there is more data available. |
181 | StrAppend(dest: &leftover_, a: chunk.substr(pos: n)); |
182 | return status; |
183 | } else { |
184 | leftover_.assign(s: chunk.data(), n: chunk.size()); |
185 | return util::Status(); |
186 | } |
187 | } |
188 | |
189 | util::Status JsonStreamParser::FinishParse() { |
190 | // If we do not expect anything and there is nothing left to parse we're all |
191 | // done. |
192 | if (stack_.empty() && leftover_.empty()) { |
193 | return util::Status(); |
194 | } |
195 | |
196 | // Lifetime needs to last until RunParser returns, so keep this variable |
197 | // outside of the coerce_to_utf8 block. |
198 | std::unique_ptr<std::string> scratch; |
199 | |
200 | bool is_valid_utf8 = internal::IsStructurallyValidUTF8(str: leftover_); |
201 | if (coerce_to_utf8_ && !is_valid_utf8) { |
202 | scratch.reset(p: new std::string); |
203 | scratch->reserve(res_arg: leftover_.size() * utf8_replacement_character_.size()); |
204 | ReplaceInvalidCodePoints(str: leftover_, replacement: utf8_replacement_character_, |
205 | dst: scratch.get()); |
206 | p_ = json_ = *scratch; |
207 | } else { |
208 | p_ = json_ = leftover_; |
209 | if (!is_valid_utf8) { |
210 | return ReportFailure(message: "Encountered non UTF-8 code points." , |
211 | parse_code: ParseErrorType::NON_UTF_8); |
212 | } |
213 | } |
214 | |
215 | // Parse the remainder in finishing mode, which reports errors for things like |
216 | // unterminated strings or unknown tokens that would normally be retried. |
217 | finishing_ = true; |
218 | util::Status result = RunParser(); |
219 | if (result.ok()) { |
220 | SkipWhitespace(); |
221 | if (!p_.empty()) { |
222 | result = |
223 | ReportFailure(message: "Parsing terminated before end of input." , |
224 | parse_code: ParseErrorType::PARSING_TERMINATED_BEFORE_END_OF_INPUT); |
225 | } |
226 | } |
227 | return result; |
228 | } |
229 | |
230 | util::Status JsonStreamParser::ParseChunk(StringPiece chunk) { |
231 | // Do not do any work if the chunk is empty. |
232 | if (chunk.empty()) return util::Status(); |
233 | |
234 | p_ = json_ = chunk; |
235 | |
236 | finishing_ = false; |
237 | util::Status result = RunParser(); |
238 | if (!result.ok()) return result; |
239 | |
240 | SkipWhitespace(); |
241 | if (p_.empty()) { |
242 | // If we parsed everything we had, clear the leftover. |
243 | leftover_.clear(); |
244 | } else { |
245 | // If we do not expect anything i.e. stack is empty, and we have non-empty |
246 | // string left to parse, we report an error. |
247 | if (stack_.empty()) { |
248 | return ReportFailure( |
249 | message: "Parsing terminated before end of input." , |
250 | parse_code: ParseErrorType::PARSING_TERMINATED_BEFORE_END_OF_INPUT); |
251 | } |
252 | // If we expect future data i.e. stack is non-empty, and we have some |
253 | // unparsed data left, we save it for later parse. |
254 | leftover_ = std::string(p_); |
255 | } |
256 | return util::Status(); |
257 | } |
258 | |
259 | bool JsonStreamParser::IsInputAllWhiteSpaces(TokenType type) { |
260 | // Conclude the whole input is full of white spaces by: |
261 | // - it is at the finishing stage |
262 | // - we have run out of the input data |
263 | // - haven't seen non-whitespace char so far |
264 | if (finishing_ && p_.empty() && type == UNKNOWN && !seen_non_whitespace_) { |
265 | return true; |
266 | } |
267 | return false; |
268 | } |
269 | |
270 | util::Status JsonStreamParser::RunParser() { |
271 | while (!stack_.empty()) { |
272 | ParseType type = stack_.top(); |
273 | TokenType t = (string_open_ == 0) ? GetNextTokenType() : BEGIN_STRING; |
274 | stack_.pop(); |
275 | util::Status result; |
276 | switch (type) { |
277 | case VALUE: |
278 | if (allow_no_root_element_ && IsInputAllWhiteSpaces(type: t)) { |
279 | return util::Status(); |
280 | } |
281 | result = ParseValue(type: t); |
282 | break; |
283 | |
284 | case OBJ_MID: |
285 | result = ParseObjectMid(type: t); |
286 | break; |
287 | |
288 | case ENTRY: |
289 | result = ParseEntry(type: t); |
290 | break; |
291 | |
292 | case ENTRY_MID: |
293 | result = ParseEntryMid(type: t); |
294 | break; |
295 | |
296 | case ARRAY_VALUE: |
297 | result = ParseArrayValue(type: t); |
298 | break; |
299 | |
300 | case ARRAY_MID: |
301 | result = ParseArrayMid(type: t); |
302 | break; |
303 | |
304 | default: |
305 | result = |
306 | util::InternalError(message: StrCat(a: "Unknown parse type: " , b: type)); |
307 | break; |
308 | } |
309 | if (!result.ok()) { |
310 | // If we were cancelled, save our state and try again later. |
311 | if (!finishing_ && util::IsCancelled(status: result)) { |
312 | stack_.push(x: type); |
313 | // If we have a key we still need to render, make sure to save off the |
314 | // contents in our own storage. |
315 | if (!key_.empty() && key_storage_.empty()) { |
316 | StrAppend(dest: &key_storage_, a: key_); |
317 | key_ = StringPiece(key_storage_); |
318 | } |
319 | result = util::Status(); |
320 | } |
321 | return result; |
322 | } |
323 | } |
324 | return util::Status(); |
325 | } |
326 | |
327 | util::Status JsonStreamParser::ParseValue(TokenType type) { |
328 | switch (type) { |
329 | case BEGIN_OBJECT: |
330 | return HandleBeginObject(); |
331 | case BEGIN_ARRAY: |
332 | return HandleBeginArray(); |
333 | case BEGIN_STRING: |
334 | return ParseString(); |
335 | case BEGIN_NUMBER: |
336 | return ParseNumber(); |
337 | case BEGIN_TRUE: |
338 | return ParseTrue(); |
339 | case BEGIN_FALSE: |
340 | return ParseFalse(); |
341 | case BEGIN_NULL: |
342 | return ParseNull(); |
343 | case UNKNOWN: |
344 | return ReportUnknown(message: "Expected a value." , parse_code: ParseErrorType::EXPECTED_VALUE); |
345 | default: { |
346 | // Special case for having been cut off while parsing, wait for more data. |
347 | // This handles things like 'fals' being at the end of the string, we |
348 | // don't know if the next char would be e, completing it, or something |
349 | // else, making it invalid. |
350 | if (!finishing_ && p_.length() < kKeywordFalse.length()) { |
351 | return util::CancelledError(message: "" ); |
352 | } |
353 | |
354 | if (allow_empty_null_ && IsEmptyNullAllowed(type)) { |
355 | return ParseEmptyNull(); |
356 | } |
357 | return ReportFailure(message: "Unexpected token." , |
358 | parse_code: ParseErrorType::UNEXPECTED_TOKEN); |
359 | } |
360 | } |
361 | } |
362 | |
363 | util::Status JsonStreamParser::ParseString() { |
364 | util::Status result = ParseStringHelper(); |
365 | if (result.ok()) { |
366 | ow_->RenderString(name: key_, value: parsed_); |
367 | key_ = StringPiece(); |
368 | parsed_ = StringPiece(); |
369 | parsed_storage_.clear(); |
370 | } |
371 | return result; |
372 | } |
373 | |
374 | util::Status JsonStreamParser::ParseStringHelper() { |
375 | // If we haven't seen the start quote, grab it and remember it for later. |
376 | if (string_open_ == 0) { |
377 | string_open_ = *p_.data(); |
378 | GOOGLE_DCHECK(string_open_ == '\"' || string_open_ == '\''); |
379 | Advance(); |
380 | } |
381 | // Track where we last copied data from so we can minimize copying. |
382 | const char* last = p_.data(); |
383 | while (!p_.empty()) { |
384 | const char* data = p_.data(); |
385 | if (*data == '\\') { |
386 | // We're about to handle an escape, copy all bytes from last to data. |
387 | if (last < data) { |
388 | parsed_storage_.append(s: last, n: data - last); |
389 | } |
390 | // If we ran out of string after the \, cancel or report an error |
391 | // depending on if we expect more data later. |
392 | if (p_.length() == 1) { |
393 | if (!finishing_) { |
394 | return util::CancelledError(message: "" ); |
395 | } |
396 | return ReportFailure(message: "Closing quote expected in string." , |
397 | parse_code: ParseErrorType::EXPECTED_CLOSING_QUOTE); |
398 | } |
399 | // Parse a unicode escape if we found \u in the string. |
400 | if (data[1] == 'u') { |
401 | util::Status result = ParseUnicodeEscape(); |
402 | if (!result.ok()) { |
403 | return result; |
404 | } |
405 | // Move last pointer past the unicode escape and continue. |
406 | last = p_.data(); |
407 | continue; |
408 | } |
409 | // Handle the standard set of backslash-escaped characters. |
410 | switch (data[1]) { |
411 | case 'b': |
412 | parsed_storage_.push_back(c: '\b'); |
413 | break; |
414 | case 'f': |
415 | parsed_storage_.push_back(c: '\f'); |
416 | break; |
417 | case 'n': |
418 | parsed_storage_.push_back(c: '\n'); |
419 | break; |
420 | case 'r': |
421 | parsed_storage_.push_back(c: '\r'); |
422 | break; |
423 | case 't': |
424 | parsed_storage_.push_back(c: '\t'); |
425 | break; |
426 | case 'v': |
427 | parsed_storage_.push_back(c: '\v'); |
428 | break; |
429 | default: |
430 | parsed_storage_.push_back(c: data[1]); |
431 | } |
432 | // We handled two characters, so advance past them and continue. |
433 | p_.remove_prefix(n: 2); |
434 | last = p_.data(); |
435 | continue; |
436 | } |
437 | // If we found the closing quote note it, advance past it, and return. |
438 | if (*data == string_open_) { |
439 | // If we didn't copy anything, reuse the input buffer. |
440 | if (parsed_storage_.empty()) { |
441 | parsed_ = StringPiece(last, data - last); |
442 | } else { |
443 | if (last < data) { |
444 | parsed_storage_.append(s: last, n: data - last); |
445 | } |
446 | parsed_ = StringPiece(parsed_storage_); |
447 | } |
448 | // Clear the quote char so next time we try to parse a string we'll |
449 | // start fresh. |
450 | string_open_ = 0; |
451 | Advance(); |
452 | return util::Status(); |
453 | } |
454 | // Normal character, just advance past it. |
455 | Advance(); |
456 | } |
457 | // If we ran out of characters, copy over what we have so far. |
458 | if (last < p_.data()) { |
459 | parsed_storage_.append(s: last, n: p_.data() - last); |
460 | } |
461 | // If we didn't find the closing quote but we expect more data, cancel for now |
462 | if (!finishing_) { |
463 | return util::CancelledError(message: "" ); |
464 | } |
465 | // End of string reached without a closing quote, report an error. |
466 | string_open_ = 0; |
467 | return ReportFailure(message: "Closing quote expected in string." , |
468 | parse_code: ParseErrorType::EXPECTED_CLOSING_QUOTE); |
469 | } |
470 | |
471 | // Converts a unicode escaped character to a decimal value stored in a char32 |
472 | // for use in UTF8 encoding utility. We assume that str begins with \uhhhh and |
473 | // convert that from the hex number to a decimal value. |
474 | // |
475 | // There are some security exploits with UTF-8 that we should be careful of: |
476 | // - http://www.unicode.org/reports/tr36/#UTF-8_Exploit |
477 | // - http://sites/intl-eng/design-guide/core-application |
478 | util::Status JsonStreamParser::ParseUnicodeEscape() { |
479 | if (p_.length() < kUnicodeEscapedLength) { |
480 | if (!finishing_) { |
481 | return util::CancelledError(message: "" ); |
482 | } |
483 | return ReportFailure(message: "Illegal hex string." , |
484 | parse_code: ParseErrorType::ILLEGAL_HEX_STRING); |
485 | } |
486 | GOOGLE_DCHECK_EQ('\\', p_.data()[0]); |
487 | GOOGLE_DCHECK_EQ('u', p_.data()[1]); |
488 | uint32_t code = 0; |
489 | for (int i = 2; i < kUnicodeEscapedLength; ++i) { |
490 | if (!isxdigit(p_.data()[i])) { |
491 | return ReportFailure(message: "Invalid escape sequence." , |
492 | parse_code: ParseErrorType::INVALID_ESCAPE_SEQUENCE); |
493 | } |
494 | code = (code << 4) + hex_digit_to_int(c: p_.data()[i]); |
495 | } |
496 | if (code >= JsonEscaping::kMinHighSurrogate && |
497 | code <= JsonEscaping::kMaxHighSurrogate) { |
498 | if (p_.length() < 2 * kUnicodeEscapedLength) { |
499 | if (!finishing_) { |
500 | return util::CancelledError(message: "" ); |
501 | } |
502 | if (!coerce_to_utf8_) { |
503 | return ReportFailure(message: "Missing low surrogate." , |
504 | parse_code: ParseErrorType::MISSING_LOW_SURROGATE); |
505 | } |
506 | } else if (p_.data()[kUnicodeEscapedLength] == '\\' && |
507 | p_.data()[kUnicodeEscapedLength + 1] == 'u') { |
508 | uint32_t low_code = 0; |
509 | for (int i = kUnicodeEscapedLength + 2; i < 2 * kUnicodeEscapedLength; |
510 | ++i) { |
511 | if (!isxdigit(p_.data()[i])) { |
512 | return ReportFailure(message: "Invalid escape sequence." , |
513 | parse_code: ParseErrorType::INVALID_ESCAPE_SEQUENCE); |
514 | } |
515 | low_code = (low_code << 4) + hex_digit_to_int(c: p_.data()[i]); |
516 | } |
517 | if (low_code >= JsonEscaping::kMinLowSurrogate && |
518 | low_code <= JsonEscaping::kMaxLowSurrogate) { |
519 | // Convert UTF-16 surrogate pair to 21-bit Unicode codepoint. |
520 | code = (((code & 0x3FF) << 10) | (low_code & 0x3FF)) + |
521 | JsonEscaping::kMinSupplementaryCodePoint; |
522 | // Advance past the first code unit escape. |
523 | p_.remove_prefix(n: kUnicodeEscapedLength); |
524 | } else if (!coerce_to_utf8_) { |
525 | return ReportFailure(message: "Invalid low surrogate." , |
526 | parse_code: ParseErrorType::INVALID_LOW_SURROGATE); |
527 | } |
528 | } else if (!coerce_to_utf8_) { |
529 | return ReportFailure(message: "Missing low surrogate." , |
530 | parse_code: ParseErrorType::MISSING_LOW_SURROGATE); |
531 | } |
532 | } |
533 | if (!coerce_to_utf8_ && !IsValidCodePoint(code_point: code)) { |
534 | return ReportFailure(message: "Invalid unicode code point." , |
535 | parse_code: ParseErrorType::INVALID_UNICODE); |
536 | } |
537 | char buf[UTFmax]; |
538 | int len = EncodeAsUTF8Char(code_point: code, output: buf); |
539 | // Advance past the [final] code unit escape. |
540 | p_.remove_prefix(n: kUnicodeEscapedLength); |
541 | parsed_storage_.append(s: buf, n: len); |
542 | return util::Status(); |
543 | } |
544 | |
545 | util::Status JsonStreamParser::ParseNumber() { |
546 | NumberResult number; |
547 | util::Status result = ParseNumberHelper(result: &number); |
548 | if (result.ok()) { |
549 | switch (number.type) { |
550 | case NumberResult::DOUBLE: |
551 | ow_->RenderDouble(name: key_, value: number.double_val); |
552 | key_ = StringPiece(); |
553 | break; |
554 | |
555 | case NumberResult::INT: |
556 | ow_->RenderInt64(name: key_, value: number.int_val); |
557 | key_ = StringPiece(); |
558 | break; |
559 | |
560 | case NumberResult::UINT: |
561 | ow_->RenderUint64(name: key_, value: number.uint_val); |
562 | key_ = StringPiece(); |
563 | break; |
564 | |
565 | default: |
566 | return ReportFailure(message: "Unable to parse number." , |
567 | parse_code: ParseErrorType::UNABLE_TO_PARSE_NUMBER); |
568 | } |
569 | } |
570 | return result; |
571 | } |
572 | |
573 | util::Status JsonStreamParser::ParseDoubleHelper(const std::string& number, |
574 | NumberResult* result) { |
575 | if (!safe_strtod(str: number, value: &result->double_val)) { |
576 | return ReportFailure(message: "Unable to parse number." , |
577 | parse_code: ParseErrorType::UNABLE_TO_PARSE_NUMBER); |
578 | } |
579 | if (!loose_float_number_conversion_ && !std::isfinite(x: result->double_val)) { |
580 | return ReportFailure(message: "Number exceeds the range of double." , |
581 | parse_code: ParseErrorType::NUMBER_EXCEEDS_RANGE_DOUBLE); |
582 | } |
583 | result->type = NumberResult::DOUBLE; |
584 | return util::Status(); |
585 | } |
586 | |
587 | util::Status JsonStreamParser::ParseNumberHelper(NumberResult* result) { |
588 | const char* data = p_.data(); |
589 | int length = p_.length(); |
590 | |
591 | // Look for the first non-numeric character, or the end of the string. |
592 | int index = 0; |
593 | bool floating = false; |
594 | bool negative = data[index] == '-'; |
595 | // Find the first character that cannot be part of the number. Along the way |
596 | // detect if the number needs to be parsed as a double. |
597 | // Note that this restricts numbers to the JSON specification, so for example |
598 | // we do not support hex or octal notations. |
599 | for (; index < length; ++index) { |
600 | char c = data[index]; |
601 | if (isdigit(c)) continue; |
602 | if (c == '.' || c == 'e' || c == 'E') { |
603 | floating = true; |
604 | continue; |
605 | } |
606 | if (c == '+' || c == '-' || c == 'x') continue; |
607 | // Not a valid number character, break out. |
608 | break; |
609 | } |
610 | |
611 | // If the entire input is a valid number, and we may have more content in the |
612 | // future, we abort for now and resume when we know more. |
613 | if (index == length && !finishing_) { |
614 | return util::CancelledError(message: "" ); |
615 | } |
616 | |
617 | // Create a string containing just the number, so we can use safe_strtoX |
618 | std::string number = std::string(p_.substr(pos: 0, n: index)); |
619 | |
620 | // Floating point number, parse as a double. |
621 | if (floating) { |
622 | util::Status status = ParseDoubleHelper(number, result); |
623 | if (status.ok()) { |
624 | p_.remove_prefix(n: index); |
625 | } |
626 | return status; |
627 | } |
628 | |
629 | // Positive non-floating point number, parse as a uint64_t. |
630 | if (!negative) { |
631 | // Octal/Hex numbers are not valid JSON values. |
632 | if (number.length() >= 2 && number[0] == '0') { |
633 | return ReportFailure( |
634 | message: "Octal/hex numbers are not valid JSON values." , |
635 | parse_code: ParseErrorType::OCTAL_OR_HEX_ARE_NOT_VALID_JSON_VALUES); |
636 | } |
637 | if (safe_strtou64(str: number, value: &result->uint_val)) { |
638 | result->type = NumberResult::UINT; |
639 | p_.remove_prefix(n: index); |
640 | return util::Status(); |
641 | } else { |
642 | // If the value is too large, parse it as double. |
643 | util::Status status = ParseDoubleHelper(number, result); |
644 | if (status.ok()) { |
645 | p_.remove_prefix(n: index); |
646 | } |
647 | return status; |
648 | } |
649 | } |
650 | |
651 | // Octal/Hex numbers are not valid JSON values. |
652 | if (number.length() >= 3 && number[1] == '0') { |
653 | return ReportFailure( |
654 | message: "Octal/hex numbers are not valid JSON values." , |
655 | parse_code: ParseErrorType::OCTAL_OR_HEX_ARE_NOT_VALID_JSON_VALUES); |
656 | } |
657 | // Negative non-floating point number, parse as an int64_t. |
658 | if (safe_strto64(str: number, value: &result->int_val)) { |
659 | result->type = NumberResult::INT; |
660 | p_.remove_prefix(n: index); |
661 | return util::Status(); |
662 | } else { |
663 | // If the value is too large, parse it as double. |
664 | util::Status status = ParseDoubleHelper(number, result); |
665 | if (status.ok()) { |
666 | p_.remove_prefix(n: index); |
667 | } |
668 | return status; |
669 | } |
670 | } |
671 | |
672 | util::Status JsonStreamParser::HandleBeginObject() { |
673 | GOOGLE_DCHECK_EQ('{', *p_.data()); |
674 | Advance(); |
675 | ow_->StartObject(name: key_); |
676 | auto status = IncrementRecursionDepth(key: key_); |
677 | if (!status.ok()) { |
678 | return status; |
679 | } |
680 | key_ = StringPiece(); |
681 | stack_.push(x: ENTRY); |
682 | return util::Status(); |
683 | } |
684 | |
685 | util::Status JsonStreamParser::ParseObjectMid(TokenType type) { |
686 | if (type == UNKNOWN) { |
687 | return ReportUnknown(message: "Expected , or } after key:value pair." , |
688 | parse_code: ParseErrorType::EXPECTED_COMMA_OR_BRACES); |
689 | } |
690 | |
691 | // Object is complete, advance past the comma and render the EndObject. |
692 | if (type == END_OBJECT) { |
693 | Advance(); |
694 | ow_->EndObject(); |
695 | --recursion_depth_; |
696 | return util::Status(); |
697 | } |
698 | // Found a comma, advance past it and get ready for an entry. |
699 | if (type == VALUE_SEPARATOR) { |
700 | Advance(); |
701 | stack_.push(x: ENTRY); |
702 | return util::Status(); |
703 | } |
704 | // Illegal token after key:value pair. |
705 | return ReportFailure(message: "Expected , or } after key:value pair." , |
706 | parse_code: ParseErrorType::EXPECTED_COMMA_OR_BRACES); |
707 | } |
708 | |
709 | util::Status JsonStreamParser::ParseEntry(TokenType type) { |
710 | if (type == UNKNOWN) { |
711 | return ReportUnknown(message: "Expected an object key or }." , |
712 | parse_code: ParseErrorType::EXPECTED_OBJECT_KEY_OR_BRACES); |
713 | } |
714 | |
715 | // Close the object and return. This allows for trailing commas. |
716 | if (type == END_OBJECT) { |
717 | ow_->EndObject(); |
718 | Advance(); |
719 | --recursion_depth_; |
720 | return util::Status(); |
721 | } |
722 | |
723 | util::Status result; |
724 | if (type == BEGIN_STRING) { |
725 | // Key is a string (standard JSON), parse it and store the string. |
726 | result = ParseStringHelper(); |
727 | if (result.ok()) { |
728 | key_storage_.clear(); |
729 | if (!parsed_storage_.empty()) { |
730 | parsed_storage_.swap(s&: key_storage_); |
731 | key_ = StringPiece(key_storage_); |
732 | } else { |
733 | key_ = parsed_; |
734 | } |
735 | parsed_ = StringPiece(); |
736 | } |
737 | } else if (type == BEGIN_KEY) { |
738 | // Key is a bare key (back compat), create a StringPiece pointing to it. |
739 | result = ParseKey(); |
740 | } else if (type == BEGIN_NULL || type == BEGIN_TRUE || type == BEGIN_FALSE) { |
741 | // Key may be a bare key that begins with a reserved word. |
742 | result = ParseKey(); |
743 | if (result.ok() && (key_ == kKeywordNull || key_ == kKeywordTrue || |
744 | key_ == kKeywordFalse)) { |
745 | result = ReportFailure(message: "Expected an object key or }." , |
746 | parse_code: ParseErrorType::EXPECTED_OBJECT_KEY_OR_BRACES); |
747 | } |
748 | } else { |
749 | // Unknown key type, report an error. |
750 | result = ReportFailure(message: "Expected an object key or }." , |
751 | parse_code: ParseErrorType::EXPECTED_OBJECT_KEY_OR_BRACES); |
752 | } |
753 | // On success we next expect an entry mid ':' then an object mid ',' or '}' |
754 | if (result.ok()) { |
755 | stack_.push(x: OBJ_MID); |
756 | stack_.push(x: ENTRY_MID); |
757 | } |
758 | return result; |
759 | } |
760 | |
761 | util::Status JsonStreamParser::ParseEntryMid(TokenType type) { |
762 | if (type == UNKNOWN) { |
763 | return ReportUnknown(message: "Expected : between key:value pair." , |
764 | parse_code: ParseErrorType::EXPECTED_COLON); |
765 | } |
766 | if (type == ENTRY_SEPARATOR) { |
767 | Advance(); |
768 | stack_.push(x: VALUE); |
769 | return util::Status(); |
770 | } |
771 | return ReportFailure(message: "Expected : between key:value pair." , |
772 | parse_code: ParseErrorType::EXPECTED_COLON); |
773 | } |
774 | |
775 | util::Status JsonStreamParser::HandleBeginArray() { |
776 | GOOGLE_DCHECK_EQ('[', *p_.data()); |
777 | Advance(); |
778 | ow_->StartList(name: key_); |
779 | key_ = StringPiece(); |
780 | stack_.push(x: ARRAY_VALUE); |
781 | return util::Status(); |
782 | } |
783 | |
784 | util::Status JsonStreamParser::ParseArrayValue(TokenType type) { |
785 | if (type == UNKNOWN) { |
786 | return ReportUnknown(message: "Expected a value or ] within an array." , |
787 | parse_code: ParseErrorType::EXPECTED_VALUE_OR_BRACKET); |
788 | } |
789 | |
790 | if (type == END_ARRAY) { |
791 | ow_->EndList(); |
792 | Advance(); |
793 | return util::Status(); |
794 | } |
795 | |
796 | // The ParseValue call may push something onto the stack so we need to make |
797 | // sure an ARRAY_MID is after it, so we push it on now. Also, the parsing of |
798 | // empty-null array value is relying on this ARRAY_MID token. |
799 | stack_.push(x: ARRAY_MID); |
800 | util::Status result = ParseValue(type); |
801 | if (util::IsCancelled(status: result)) { |
802 | // If we were cancelled, pop back off the ARRAY_MID so we don't try to |
803 | // push it on again when we try over. |
804 | stack_.pop(); |
805 | } |
806 | return result; |
807 | } |
808 | |
809 | util::Status JsonStreamParser::ParseArrayMid(TokenType type) { |
810 | if (type == UNKNOWN) { |
811 | return ReportUnknown(message: "Expected , or ] after array value." , |
812 | parse_code: ParseErrorType::EXPECTED_COMMA_OR_BRACKET); |
813 | } |
814 | |
815 | if (type == END_ARRAY) { |
816 | ow_->EndList(); |
817 | Advance(); |
818 | return util::Status(); |
819 | } |
820 | |
821 | // Found a comma, advance past it and expect an array value next. |
822 | if (type == VALUE_SEPARATOR) { |
823 | Advance(); |
824 | stack_.push(x: ARRAY_VALUE); |
825 | return util::Status(); |
826 | } |
827 | // Illegal token after array value. |
828 | return ReportFailure(message: "Expected , or ] after array value." , |
829 | parse_code: ParseErrorType::EXPECTED_COMMA_OR_BRACKET); |
830 | } |
831 | |
832 | util::Status JsonStreamParser::ParseTrue() { |
833 | ow_->RenderBool(name: key_, value: true); |
834 | key_ = StringPiece(); |
835 | p_.remove_prefix(n: kKeywordTrue.length()); |
836 | return util::Status(); |
837 | } |
838 | |
839 | util::Status JsonStreamParser::ParseFalse() { |
840 | ow_->RenderBool(name: key_, value: false); |
841 | key_ = StringPiece(); |
842 | p_.remove_prefix(n: kKeywordFalse.length()); |
843 | return util::Status(); |
844 | } |
845 | |
846 | util::Status JsonStreamParser::ParseNull() { |
847 | ow_->RenderNull(name: key_); |
848 | key_ = StringPiece(); |
849 | p_.remove_prefix(n: kKeywordNull.length()); |
850 | return util::Status(); |
851 | } |
852 | |
853 | util::Status JsonStreamParser::ParseEmptyNull() { |
854 | ow_->RenderNull(name: key_); |
855 | key_ = StringPiece(); |
856 | return util::Status(); |
857 | } |
858 | |
859 | bool JsonStreamParser::IsEmptyNullAllowed(TokenType type) { |
860 | if (stack_.empty()) return false; |
861 | return (stack_.top() == ARRAY_MID && type == VALUE_SEPARATOR) || |
862 | stack_.top() == OBJ_MID; |
863 | } |
864 | |
865 | util::Status JsonStreamParser::ReportFailure(StringPiece message, |
866 | ParseErrorType parse_code) { |
867 | (void)parse_code; // Parameter is used in Google-internal code. |
868 | static const int kContextLength = 20; |
869 | const char* p_start = p_.data(); |
870 | const char* json_start = json_.data(); |
871 | const char* begin = std::max(p_start - kContextLength, json_start); |
872 | const char* end = |
873 | std::min(p_start + kContextLength, json_start + json_.size()); |
874 | StringPiece segment(begin, end - begin); |
875 | std::string location(p_start - begin, ' '); |
876 | location.push_back(c: '^'); |
877 | auto status = util::InvalidArgumentError( |
878 | message: StrCat(a: message, b: "\n" , c: segment, d: "\n" , e: location)); |
879 | return status; |
880 | } |
881 | |
882 | util::Status JsonStreamParser::ReportUnknown(StringPiece message, |
883 | ParseErrorType parse_code) { |
884 | // If we aren't finishing the parse, cancel parsing and try later. |
885 | if (!finishing_) { |
886 | return util::CancelledError(message: "" ); |
887 | } |
888 | if (p_.empty()) { |
889 | return ReportFailure(message: StrCat(a: "Unexpected end of string. " , b: message), |
890 | parse_code); |
891 | } |
892 | return ReportFailure(message, parse_code); |
893 | } |
894 | |
895 | util::Status JsonStreamParser::IncrementRecursionDepth( |
896 | StringPiece key) const { |
897 | if (++recursion_depth_ > max_recursion_depth_) { |
898 | return util::InvalidArgumentError(message: StrCat( |
899 | a: "Message too deep. Max recursion depth reached for key '" , b: key, c: "'" )); |
900 | } |
901 | return util::Status(); |
902 | } |
903 | |
904 | void JsonStreamParser::SkipWhitespace() { |
905 | while (!p_.empty() && ascii_isspace(c: *p_.data())) { |
906 | Advance(); |
907 | } |
908 | if (!p_.empty() && !ascii_isspace(c: *p_.data())) { |
909 | seen_non_whitespace_ = true; |
910 | } |
911 | } |
912 | |
913 | void JsonStreamParser::Advance() { |
914 | // Advance by moving one UTF8 character while making sure we don't go beyond |
915 | // the length of StringPiece. |
916 | p_.remove_prefix(n: std::min<int>( |
917 | p_.length(), UTF8FirstLetterNumBytes(src: p_.data(), len: p_.length()))); |
918 | } |
919 | |
920 | util::Status JsonStreamParser::ParseKey() { |
921 | StringPiece original = p_; |
922 | |
923 | if (allow_permissive_key_naming_) { |
924 | if (!ConsumeKeyPermissive(input: &p_, key: &key_)) { |
925 | return ReportFailure(message: "Invalid key or variable name." , |
926 | parse_code: ParseErrorType::INVALID_KEY_OR_VARIABLE_NAME); |
927 | } |
928 | } else { |
929 | if (!ConsumeKey(input: &p_, key: &key_)) { |
930 | return ReportFailure(message: "Invalid key or variable name." , |
931 | parse_code: ParseErrorType::INVALID_KEY_OR_VARIABLE_NAME); |
932 | } |
933 | } |
934 | |
935 | // If we consumed everything but expect more data, reset p_ and cancel since |
936 | // we can't know if the key was complete or not. |
937 | if (!finishing_ && p_.empty()) { |
938 | p_ = original; |
939 | return util::CancelledError(message: "" ); |
940 | } |
941 | // Since we aren't using the key storage, clear it out. |
942 | key_storage_.clear(); |
943 | return util::Status(); |
944 | } |
945 | |
946 | JsonStreamParser::TokenType JsonStreamParser::GetNextTokenType() { |
947 | SkipWhitespace(); |
948 | |
949 | int size = p_.size(); |
950 | if (size == 0) { |
951 | // If we ran out of data, report unknown and we'll place the previous parse |
952 | // type onto the stack and try again when we have more data. |
953 | return UNKNOWN; |
954 | } |
955 | // TODO(sven): Split this method based on context since different contexts |
956 | // support different tokens. Would slightly speed up processing? |
957 | const char* data = p_.data(); |
958 | StringPiece data_view = StringPiece(data, size); |
959 | if (*data == '\"' || *data == '\'') return BEGIN_STRING; |
960 | if (*data == '-' || ('0' <= *data && *data <= '9')) { |
961 | return BEGIN_NUMBER; |
962 | } |
963 | if (size >= kKeywordTrue.length() && |
964 | HasPrefixString(str: data_view, prefix: kKeywordTrue)) { |
965 | return BEGIN_TRUE; |
966 | } |
967 | if (size >= kKeywordFalse.length() && |
968 | HasPrefixString(str: data_view, prefix: kKeywordFalse)) { |
969 | return BEGIN_FALSE; |
970 | } |
971 | if (size >= kKeywordNull.length() && |
972 | HasPrefixString(str: data_view, prefix: kKeywordNull)) { |
973 | return BEGIN_NULL; |
974 | } |
975 | if (*data == '{') return BEGIN_OBJECT; |
976 | if (*data == '}') return END_OBJECT; |
977 | if (*data == '[') return BEGIN_ARRAY; |
978 | if (*data == ']') return END_ARRAY; |
979 | if (*data == ':') return ENTRY_SEPARATOR; |
980 | if (*data == ',') return VALUE_SEPARATOR; |
981 | if (MatchKey(input: p_)) { |
982 | return BEGIN_KEY; |
983 | } |
984 | |
985 | // We don't know that we necessarily have an invalid token here, just that we |
986 | // can't parse what we have so far. So we don't report an error and just |
987 | // return UNKNOWN so we can try again later when we have more data, or if we |
988 | // finish and we have leftovers. |
989 | return UNKNOWN; |
990 | } |
991 | |
992 | } // namespace converter |
993 | } // namespace util |
994 | } // namespace protobuf |
995 | } // namespace google |
996 | |