/*
- * Copyright 2015 Facebook, Inc.
+ * Copyright 2017 Facebook, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
*/
#include <folly/json.h>
+
+#include <algorithm>
#include <cassert>
+#include <functional>
+
#include <boost/next_prior.hpp>
#include <boost/algorithm/string.hpp>
#include <folly/Range.h>
#include <folly/String.h>
#include <folly/Unicode.h>
+#include <folly/portability/Constexpr.h>
namespace folly {
namespace json {
namespace {
-char32_t decodeUtf8(
- const unsigned char*& p,
- const unsigned char* const e,
- bool skipOnError) {
- /* The following encodings are valid, except for the 5 and 6 byte
- * combinations:
- * 0xxxxxxx
- * 110xxxxx 10xxxxxx
- * 1110xxxx 10xxxxxx 10xxxxxx
- * 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx
- * 111110xx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx
- * 1111110x 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx
- */
-
- auto skip = [&] { ++p; return U'\ufffd'; };
-
- if (p >= e) {
- if (skipOnError) return skip();
- throw std::runtime_error("folly::decodeUtf8 empty/invalid string");
- }
-
- unsigned char fst = *p;
- if (!(fst & 0x80)) {
- // trivial case
- return *p++;
- }
-
- static const uint32_t bitMask[] = {
- (1 << 7) - 1,
- (1 << 11) - 1,
- (1 << 16) - 1,
- (1 << 21) - 1
- };
-
- // upper control bits are masked out later
- uint32_t d = fst;
-
- if ((fst & 0xC0) != 0xC0) {
- if (skipOnError) return skip();
- throw std::runtime_error(to<std::string>("folly::decodeUtf8 i=0 d=", d));
- }
-
- fst <<= 1;
-
- for (unsigned int i = 1; i != 3 && p + i < e; ++i) {
- unsigned char tmp = p[i];
-
- if ((tmp & 0xC0) != 0x80) {
- if (skipOnError) return skip();
- throw std::runtime_error(
- to<std::string>("folly::decodeUtf8 i=", i, " tmp=", (uint32_t)tmp));
- }
-
- d = (d << 6) | (tmp & 0x3F);
- fst <<= 1;
-
- if (!(fst & 0x80)) {
- d &= bitMask[i];
-
- // overlong, could have been encoded with i bytes
- if ((d & ~bitMask[i - 1]) == 0) {
- if (skipOnError) return skip();
- throw std::runtime_error(
- to<std::string>("folly::decodeUtf8 i=", i, " d=", d));
- }
-
- // check for surrogates only needed for 3 bytes
- if (i == 2) {
- if ((d >= 0xD800 && d <= 0xDFFF) || d > 0x10FFFF) {
- if (skipOnError) return skip();
- throw std::runtime_error(
- to<std::string>("folly::decodeUtf8 i=", i, " d=", d));
- }
- }
-
- p += i + 1;
- return d;
- }
- }
-
- if (skipOnError) return skip();
- throw std::runtime_error("folly::decodeUtf8 encoding length maxed out");
-}
-
struct Printer {
- explicit Printer(fbstring& out,
- unsigned* indentLevel,
- serialization_opts const* opts)
- : out_(out)
- , indentLevel_(indentLevel)
- , opts_(*opts)
- {}
+ explicit Printer(
+ std::string& out,
+ unsigned* indentLevel,
+ serialization_opts const* opts)
+ : out_(out), indentLevel_(indentLevel), opts_(*opts) {}
void operator()(dynamic const& v) const {
switch (v.type()) {
indent();
newline();
if (opts_.sort_keys) {
- std::vector<std::pair<dynamic, dynamic>> items(
- o.items().begin(), o.items().end());
- std::sort(items.begin(), items.end());
- printKVPairs(items.begin(), items.end());
+ using ref = std::reference_wrapper<decltype(o.items())::value_type const>;
+ std::vector<ref> refs(o.items().begin(), o.items().end());
+ std::sort(refs.begin(), refs.end(), [](ref a, ref b) {
+ // Only compare keys. No ordering among identical keys.
+ return a.get().first < b.get().first;
+ });
+ printKVPairs(refs.cbegin(), refs.cend());
} else {
printKVPairs(o.items().begin(), o.items().end());
}
void newline() const {
if (indentLevel_) {
- out_ += to<fbstring>('\n', fbstring(*indentLevel_ * 2, ' '));
+ out_ += to<std::string>('\n', std::string(*indentLevel_ * 2, ' '));
}
}
}
private:
- fbstring& out_;
- unsigned* const indentLevel_;
- serialization_opts const& opts_;
+ std::string& out_;
+ unsigned* const indentLevel_;
+ serialization_opts const& opts_;
};
- //////////////////////////////////////////////////////////////////////
-
- struct ParseError : std::runtime_error {
- explicit ParseError(int line)
- : std::runtime_error(to<std::string>("json parse error on line ", line))
- {}
-
- explicit ParseError(int line, std::string const& context,
- std::string const& expected)
- : std::runtime_error(to<std::string>("json parse error on line ", line,
- !context.empty() ? to<std::string>(" near `", context, '\'')
- : "",
- ": ", expected))
- {}
+//////////////////////////////////////////////////////////////////////
- explicit ParseError(std::string const& msg)
- : std::runtime_error("json parse error: " + msg)
- {}
- };
+struct ParseError : std::runtime_error {
+ explicit ParseError(
+ unsigned int line,
+ std::string const& context,
+ std::string const& expected)
+ : std::runtime_error(to<std::string>(
+ "json parse error on line ",
+ line,
+ !context.empty() ? to<std::string>(" near `", context, '\'') : "",
+ ": ",
+ expected)) {}
+};
// Wraps our input buffer with some helper functions.
struct Input {
return opts_;
}
-private:
+ void incrementRecursionLevel() {
+ if (currentRecursionLevel_ > opts_.recursion_limit) {
+ error("recursion limit exceeded");
+ }
+ currentRecursionLevel_++;
+ }
+
+ void decrementRecursionLevel() {
+ currentRecursionLevel_--;
+ }
+
+ private:
void storeCurrent() {
current_ = range_.empty() ? EOF : range_.front();
}
json::serialization_opts const& opts_;
unsigned lineNum_;
int current_;
+ unsigned int currentRecursionLevel_{0};
+};
+
+class RecursionGuard {
+ public:
+ explicit RecursionGuard(Input& in) : in_(in) {
+ in_.incrementRecursionLevel();
+ }
+
+ ~RecursionGuard() {
+ in_.decrementRecursionLevel();
+ }
+
+ private:
+ Input& in_;
};
dynamic parseValue(Input& in);
-fbstring parseString(Input& in);
+std::string parseString(Input& in);
dynamic parseNumber(Input& in);
dynamic parseObject(Input& in) {
assert(*in == '[');
++in;
- dynamic ret = {};
+ dynamic ret = dynamic::array;
in.skipWhitespace();
if (*in == ']') {
dynamic parseNumber(Input& in) {
bool const negative = (*in == '-');
- if (negative) {
- if (in.consume("-Infinity")) {
+ if (negative && in.consume("-Infinity")) {
+ if (in.getOpts().parse_numbers_as_strings) {
+ return "-Infinity";
+ } else {
return -std::numeric_limits<double>::infinity();
}
}
}
auto const wasE = *in == 'e' || *in == 'E';
+
+ constexpr const char* maxInt = "9223372036854775807";
+ constexpr const char* minInt = "-9223372036854775808";
+ constexpr auto maxIntLen = constexpr_strlen(maxInt);
+ constexpr auto minIntLen = constexpr_strlen(minInt);
+
+ if (*in != '.' && !wasE && in.getOpts().parse_numbers_as_strings) {
+ return integral;
+ }
+
if (*in != '.' && !wasE) {
- auto val = to<int64_t>(integral);
- in.skipWhitespace();
- return val;
+ if (LIKELY(!in.getOpts().double_fallback || integral.size() < maxIntLen) ||
+ (!negative && integral.size() == maxIntLen && integral <= maxInt) ||
+ (negative && integral.size() == minIntLen && integral <= minInt)) {
+ auto val = to<int64_t>(integral);
+ in.skipWhitespace();
+ return val;
+ } else {
+ auto val = to<double>(integral);
+ in.skipWhitespace();
+ return val;
+ }
}
auto end = !wasE ? (++in, in.skipDigits().end()) : in.begin();
end = expPart.end();
}
auto fullNum = range(integral.begin(), end);
-
+ if (in.getOpts().parse_numbers_as_strings) {
+ return fullNum;
+ }
auto val = to<double>(fullNum);
return val;
}
-fbstring decodeUnicodeEscape(Input& in) {
- auto hexVal = [&] (char c) -> unsigned {
- return c >= '0' && c <= '9' ? c - '0' :
+std::string decodeUnicodeEscape(Input& in) {
+ auto hexVal = [&] (int c) -> uint16_t {
+ return uint16_t(
+ c >= '0' && c <= '9' ? c - '0' :
c >= 'a' && c <= 'f' ? c - 'a' + 10 :
c >= 'A' && c <= 'F' ? c - 'A' + 10 :
- (in.error("invalid hex digit"), 0);
+ (in.error("invalid hex digit"), 0));
};
auto readHex = [&]() -> uint16_t {
in.error("expected 4 hex digits");
}
- uint16_t ret = hexVal(*in) * 4096;
+ uint16_t ret = uint16_t(hexVal(*in) * 4096);
++in;
ret += hexVal(*in) * 256;
++in;
return codePointToUtf8(codePoint);
}
-fbstring parseString(Input& in) {
+std::string parseString(Input& in) {
assert(*in == '\"');
++in;
- fbstring ret;
+ std::string ret;
for (;;) {
auto range = in.skipWhile(
[] (char c) { return c != '\"' && c != '\\'; }
case 'r': ret.push_back('\r'); ++in; break;
case 't': ret.push_back('\t'); ++in; break;
case 'u': ++in; ret += decodeUnicodeEscape(in); break;
- default: in.error(to<fbstring>("unknown escape ", *in,
- " in string").c_str());
+ default:
+ in.error(to<std::string>("unknown escape ", *in, " in string").c_str());
}
continue;
}
in.error("null byte in string");
}
- ret.push_back(*in);
+ ret.push_back(char(*in));
++in;
}
}
dynamic parseValue(Input& in) {
+ RecursionGuard guard(in);
+
in.skipWhitespace();
return *in == '[' ? parseArray(in) :
*in == '{' ? parseObject(in) :
in.consume("true") ? true :
in.consume("false") ? false :
in.consume("null") ? nullptr :
- in.consume("Infinity") ? std::numeric_limits<double>::infinity() :
- in.consume("NaN") ? std::numeric_limits<double>::quiet_NaN() :
+ in.consume("Infinity") ?
+ (in.getOpts().parse_numbers_as_strings ? (dynamic)"Infinity" :
+ (dynamic)std::numeric_limits<double>::infinity()) :
+ in.consume("NaN") ?
+ (in.getOpts().parse_numbers_as_strings ? (dynamic)"NaN" :
+ (dynamic)std::numeric_limits<double>::quiet_NaN()) :
in.error("expected json value");
}
//////////////////////////////////////////////////////////////////////
-fbstring serialize(dynamic const& dyn, serialization_opts const& opts) {
- fbstring ret;
+std::string serialize(dynamic const& dyn, serialization_opts const& opts) {
+ std::string ret;
unsigned indentLevel = 0;
Printer p(ret, opts.pretty_formatting ? &indentLevel : nullptr, &opts);
p(dyn);
}
// Escape a string so that it is legal to print it in JSON text.
-void escapeString(StringPiece input,
- fbstring& out,
- const serialization_opts& opts) {
- auto hexDigit = [] (int c) -> char {
+void escapeString(
+ StringPiece input,
+ std::string& out,
+ const serialization_opts& opts) {
+ auto hexDigit = [] (uint8_t c) -> char {
return c < 10 ? c + '0' : c - 10 + 'a';
};
- out.reserve(out.size() + input.size() + 2);
out.push_back('\"');
auto* p = reinterpret_cast<const unsigned char*>(input.begin());
if (q == p) {
// calling utf8_decode has the side effect of
// checking that utf8 encodings are valid
- char32_t v = decodeUtf8(q, e, opts.skip_invalid_utf8);
+ char32_t v = utf8ToCodePoint(q, e, opts.skip_invalid_utf8);
if (opts.skip_invalid_utf8 && v == U'\ufffd') {
- out.append("\ufffd");
+ out.append(u8"\ufffd");
p = q;
continue;
}
if (opts.encode_non_ascii && (*p & 0x80)) {
// note that this if condition captures utf8 chars
// with value > 127, so size > 1 byte
- char32_t v = decodeUtf8(p, e, opts.skip_invalid_utf8);
+ char32_t v = utf8ToCodePoint(p, e, opts.skip_invalid_utf8);
out.append("\\u");
- out.push_back(hexDigit(v >> 12));
+ out.push_back(hexDigit(uint8_t(v >> 12)));
out.push_back(hexDigit((v >> 8) & 0x0f));
out.push_back(hexDigit((v >> 4) & 0x0f));
out.push_back(hexDigit(v & 0x0f));
} else if (*p == '\\' || *p == '\"') {
out.push_back('\\');
- out.push_back(*p++);
+ out.push_back(char(*p++));
} else if (*p <= 0x1f) {
switch (*p) {
case '\b': out.append("\\b"); p++; break;
// note that this if condition captures non readable chars
// with value < 32, so size = 1 byte (e.g control chars).
out.append("\\u00");
- out.push_back(hexDigit((*p & 0xf0) >> 4));
- out.push_back(hexDigit(*p & 0xf));
+ out.push_back(hexDigit(uint8_t((*p & 0xf0) >> 4)));
+ out.push_back(hexDigit(uint8_t(*p & 0xf)));
p++;
}
} else {
- out.push_back(*p++);
+ out.push_back(char(*p++));
}
}
out.push_back('\"');
}
-fbstring stripComments(StringPiece jsonC) {
- fbstring result;
+std::string stripComments(StringPiece jsonC) {
+ std::string result;
enum class State {
None,
InString,
return ret;
}
-fbstring toJson(dynamic const& dyn) {
+std::string toJson(dynamic const& dyn) {
return json::serialize(dyn, json::serialization_opts());
}
-fbstring toPrettyJson(dynamic const& dyn) {
+std::string toPrettyJson(dynamic const& dyn) {
json::serialization_opts opts;
opts.pretty_formatting = true;
return json::serialize(dyn, opts);
out << json::serialize(*this, opts);
}
+void PrintTo(const dynamic& dyn, std::ostream* os) {
+ json::serialization_opts opts;
+ opts.allow_nan_inf = true;
+ opts.allow_non_string_keys = true;
+ opts.pretty_formatting = true;
+ opts.sort_keys = true;
+ *os << json::serialize(dyn, opts);
+}
+
//////////////////////////////////////////////////////////////////////
}