json/include/nlohmann/detail/input/parser.hpp

823 lines
28 KiB
C++
Raw Normal View History

2018-01-10 09:18:31 +00:00
#pragma once
2017-08-14 15:02:40 +00:00
#include <cassert> // assert
#include <cmath> // isfinite
#include <cstdint> // uint8_t
#include <functional> // function
#include <string> // string
#include <utility> // move
2017-08-14 15:02:40 +00:00
#include <nlohmann/detail/exceptions.hpp>
#include <nlohmann/detail/macro_scope.hpp>
#include <nlohmann/detail/input/input_adapters.hpp>
2018-02-26 19:08:12 +00:00
#include <nlohmann/detail/input/json_sax.hpp>
#include <nlohmann/detail/input/lexer.hpp>
#include <nlohmann/detail/value_t.hpp>
2017-08-14 15:02:40 +00:00
namespace nlohmann
{
namespace detail
{
////////////
// parser //
////////////
/*!
@brief syntax analysis
This class implements a recursive decent parser.
*/
template<typename BasicJsonType>
class parser
{
using number_integer_t = typename BasicJsonType::number_integer_t;
using number_unsigned_t = typename BasicJsonType::number_unsigned_t;
using number_float_t = typename BasicJsonType::number_float_t;
using string_t = typename BasicJsonType::string_t;
2017-08-14 15:02:40 +00:00
using lexer_t = lexer<BasicJsonType>;
using token_type = typename lexer_t::token_type;
public:
enum class parse_event_t : uint8_t
{
/// the parser read `{` and started to process a JSON object
object_start,
/// the parser read `}` and finished processing a JSON object
object_end,
/// the parser read `[` and started to process a JSON array
array_start,
/// the parser read `]` and finished processing a JSON array
array_end,
/// the parser read a key of a value in an object
key,
/// the parser finished reading a JSON value
value
};
2018-02-26 22:39:23 +00:00
using json_sax_t = json_sax<BasicJsonType>;
2018-02-24 17:04:07 +00:00
2017-08-14 15:02:40 +00:00
using parser_callback_t =
std::function<bool(int depth, parse_event_t event, BasicJsonType& parsed)>;
/// a parser reading from an input adapter
2018-03-20 21:39:08 +00:00
explicit parser(detail::input_adapter_t&& adapter,
2017-08-14 15:02:40 +00:00
const parser_callback_t cb = nullptr,
const bool allow_exceptions_ = true)
2018-03-20 21:39:08 +00:00
: callback(cb), m_lexer(std::move(adapter)), allow_exceptions(allow_exceptions_)
2018-03-06 17:17:07 +00:00
{
// read first token
get_token();
}
2018-02-24 17:04:07 +00:00
2017-08-14 15:02:40 +00:00
/*!
@brief public parser interface
@param[in] strict whether to expect the last token to be EOF
@param[in,out] result parsed JSON value
@throw parse_error.101 in case of an unexpected token
@throw parse_error.102 if to_unicode fails or surrogate error
@throw parse_error.103 if to_unicode fails
*/
void parse(const bool strict, BasicJsonType& result)
{
2018-03-06 17:17:07 +00:00
if (callback)
{
/*
json_sax_dom_callback_parser<BasicJsonType> sdp(result, callback, allow_exceptions);
sax_parse_internal(&sdp);
result.assert_invariant();
// in strict mode, input must be completely read
if (strict and (get_token() != token_type::end_of_input))
{
sdp.parse_error(m_lexer.get_position(),
m_lexer.get_token_string(),
parse_error::create(101, m_lexer.get_position(), exception_message(token_type::end_of_input)));
}
// in case of an error, return discarded value
if (sdp.is_errored())
{
result = value_t::discarded;
return;
}
*/
2018-03-06 17:17:07 +00:00
parse_internal(true, result);
result.assert_invariant();
2017-08-14 15:02:40 +00:00
2018-03-06 17:17:07 +00:00
// in strict mode, input must be completely read
if (strict)
{
get_token();
expect(token_type::end_of_input);
}
2017-08-14 15:02:40 +00:00
2018-03-06 17:17:07 +00:00
// in case of an error, return discarded value
if (errored)
{
result = value_t::discarded;
return;
}
2017-08-14 15:02:40 +00:00
2018-03-06 17:17:07 +00:00
// set top-level value to null if it was discarded by the callback
// function
if (result.is_discarded())
{
result = nullptr;
}
2017-08-14 15:02:40 +00:00
}
2018-03-06 17:17:07 +00:00
else
2017-08-14 15:02:40 +00:00
{
2018-03-06 17:17:07 +00:00
json_sax_dom_parser<BasicJsonType> sdp(result, allow_exceptions);
sax_parse_internal(&sdp);
result.assert_invariant();
// in strict mode, input must be completely read
if (strict and (get_token() != token_type::end_of_input))
{
sdp.parse_error(m_lexer.get_position(),
m_lexer.get_token_string(),
2018-03-11 21:47:25 +00:00
parse_error::create(101, m_lexer.get_position(), exception_message(token_type::end_of_input)));
2018-03-06 17:17:07 +00:00
}
// in case of an error, return discarded value
if (sdp.is_errored())
{
result = value_t::discarded;
return;
}
2017-08-14 15:02:40 +00:00
}
}
/*!
@brief public accept interface
@param[in] strict whether to expect the last token to be EOF
@return whether the input is a proper JSON text
*/
bool accept(const bool strict = true)
{
json_sax_acceptor<BasicJsonType> sax_acceptor;
2018-03-20 21:39:08 +00:00
return sax_parse(&sax_acceptor, strict);
}
bool sax_parse(json_sax_t* sax, const bool strict = true)
{
const bool result = sax_parse_internal(sax);
2018-03-20 21:39:08 +00:00
// strict mode: next byte must be EOF
if (result and strict and (get_token() != token_type::end_of_input))
2017-08-14 15:02:40 +00:00
{
2018-03-20 21:39:08 +00:00
return sax->parse_error(m_lexer.get_position(),
m_lexer.get_token_string(),
parse_error::create(101, m_lexer.get_position(), exception_message(token_type::end_of_input)));
2017-08-14 15:02:40 +00:00
}
2018-03-20 21:39:08 +00:00
return result;
2018-02-24 17:04:07 +00:00
}
2017-08-14 15:02:40 +00:00
private:
/*!
@brief the actual parser
@throw parse_error.101 in case of an unexpected token
@throw parse_error.102 if to_unicode fails or surrogate error
@throw parse_error.103 if to_unicode fails
*/
void parse_internal(bool keep, BasicJsonType& result)
{
// never parse after a parse error was detected
assert(not errored);
2018-03-10 10:24:00 +00:00
// this function is only called when a callback is given
assert(callback);
2017-08-14 15:02:40 +00:00
// start with a discarded value
if (not result.is_discarded())
{
result.m_value.destroy(result.m_type);
result.m_type = value_t::discarded;
}
switch (last_token)
{
case token_type::begin_object:
{
if (keep)
{
2018-03-10 10:24:00 +00:00
keep = callback(depth++, parse_event_t::object_start, result);
2017-08-14 15:02:40 +00:00
2018-03-10 10:24:00 +00:00
if (keep)
2017-08-14 15:02:40 +00:00
{
// explicitly set result to object to cope with {}
result.m_type = value_t::object;
result.m_value = value_t::object;
}
}
// read next token
get_token();
// closing } -> we are done
if (last_token == token_type::end_object)
{
2018-03-10 10:24:00 +00:00
if (keep and not callback(--depth, parse_event_t::object_end, result))
2017-08-14 15:02:40 +00:00
{
result.m_value.destroy(result.m_type);
result.m_type = value_t::discarded;
}
break;
}
// parse values
string_t key;
2017-08-14 15:02:40 +00:00
BasicJsonType value;
while (true)
{
// store key
if (not expect(token_type::value_string))
{
return;
}
key = m_lexer.move_string();
bool keep_tag = false;
if (keep)
{
2018-03-10 10:24:00 +00:00
BasicJsonType k(key);
keep_tag = callback(depth, parse_event_t::key, k);
2017-08-14 15:02:40 +00:00
}
// parse separator (:)
get_token();
if (not expect(token_type::name_separator))
{
return;
}
// parse and add value
get_token();
value.m_value.destroy(value.m_type);
value.m_type = value_t::discarded;
parse_internal(keep, value);
if (JSON_UNLIKELY(errored))
{
return;
}
if (keep and keep_tag and not value.is_discarded())
{
result.m_value.object->emplace(std::move(key), std::move(value));
}
// comma -> next value
get_token();
if (last_token == token_type::value_separator)
{
get_token();
continue;
}
// closing }
if (not expect(token_type::end_object))
{
return;
}
break;
}
2018-03-10 10:24:00 +00:00
if (keep and not callback(--depth, parse_event_t::object_end, result))
2017-08-14 15:02:40 +00:00
{
result.m_value.destroy(result.m_type);
result.m_type = value_t::discarded;
}
break;
}
case token_type::begin_array:
{
if (keep)
{
2018-03-10 10:24:00 +00:00
keep = callback(depth++, parse_event_t::array_start, result);
2017-08-14 15:02:40 +00:00
2018-03-10 10:24:00 +00:00
if (keep)
2017-08-14 15:02:40 +00:00
{
// explicitly set result to array to cope with []
result.m_type = value_t::array;
result.m_value = value_t::array;
}
}
// read next token
get_token();
// closing ] -> we are done
if (last_token == token_type::end_array)
{
2018-03-10 10:24:00 +00:00
if (not callback(--depth, parse_event_t::array_end, result))
2017-08-14 15:02:40 +00:00
{
result.m_value.destroy(result.m_type);
result.m_type = value_t::discarded;
}
break;
}
// parse values
BasicJsonType value;
while (true)
{
// parse value
value.m_value.destroy(value.m_type);
value.m_type = value_t::discarded;
parse_internal(keep, value);
if (JSON_UNLIKELY(errored))
{
return;
}
if (keep and not value.is_discarded())
{
result.m_value.array->push_back(std::move(value));
}
// comma -> next value
get_token();
if (last_token == token_type::value_separator)
{
get_token();
continue;
}
// closing ]
if (not expect(token_type::end_array))
{
return;
}
break;
}
2018-03-10 10:24:00 +00:00
if (keep and not callback(--depth, parse_event_t::array_end, result))
2017-08-14 15:02:40 +00:00
{
result.m_value.destroy(result.m_type);
result.m_type = value_t::discarded;
}
break;
}
case token_type::literal_null:
{
result.m_type = value_t::null;
break;
}
case token_type::value_string:
{
result.m_type = value_t::string;
result.m_value = m_lexer.move_string();
break;
}
case token_type::literal_true:
{
result.m_type = value_t::boolean;
result.m_value = true;
break;
}
case token_type::literal_false:
{
result.m_type = value_t::boolean;
result.m_value = false;
break;
}
case token_type::value_unsigned:
{
result.m_type = value_t::number_unsigned;
result.m_value = m_lexer.get_number_unsigned();
break;
}
case token_type::value_integer:
{
result.m_type = value_t::number_integer;
result.m_value = m_lexer.get_number_integer();
break;
}
case token_type::value_float:
{
result.m_type = value_t::number_float;
result.m_value = m_lexer.get_number_float();
// throw in case of infinity or NAN
if (JSON_UNLIKELY(not std::isfinite(result.m_value.number_float)))
{
if (allow_exceptions)
{
JSON_THROW(out_of_range::create(406, "number overflow parsing '" +
m_lexer.get_token_string() + "'"));
}
expect(token_type::uninitialized);
}
break;
}
case token_type::parse_error:
{
// using "uninitialized" to avoid "expected" message
if (not expect(token_type::uninitialized))
{
return;
}
break; // LCOV_EXCL_LINE
}
default:
{
// the last token was unexpected; we expected a value
if (not expect(token_type::literal_or_value))
{
return;
}
break; // LCOV_EXCL_LINE
}
}
2018-03-10 10:24:00 +00:00
if (keep and not callback(depth, parse_event_t::value, result))
2017-08-14 15:02:40 +00:00
{
result.m_value.destroy(result.m_type);
2017-08-14 15:02:40 +00:00
result.m_type = value_t::discarded;
}
}
2018-03-06 17:17:07 +00:00
bool sax_parse_internal(json_sax_t* sax)
2018-02-24 17:04:07 +00:00
{
// two values for the structured values
enum class parse_state_t { array_value, object_value };
// stack to remember the hieararchy of structured values we are parsing
std::vector<parse_state_t> states;
// value to avoid a goto (see comment where set to true)
bool skip_to_state_evaluation = false;
while (true)
2018-02-24 17:04:07 +00:00
{
if (not skip_to_state_evaluation)
2018-02-24 17:04:07 +00:00
{
// invariant: get_token() was called before each iteration
switch (last_token)
2018-02-24 17:04:07 +00:00
{
case token_type::begin_object:
{
2018-03-18 15:07:21 +00:00
if (JSON_UNLIKELY(not sax->start_object()))
{
return false;
}
2018-02-24 17:04:07 +00:00
// read next token
get_token();
2018-02-24 17:04:07 +00:00
// closing } -> we are done
if (last_token == token_type::end_object)
{
2018-03-18 15:07:21 +00:00
if (JSON_UNLIKELY(not sax->end_object()))
{
return false;
}
break;
}
2018-02-24 17:04:07 +00:00
// parse key
if (JSON_UNLIKELY(last_token != token_type::value_string))
{
return sax->parse_error(m_lexer.get_position(),
m_lexer.get_token_string(),
parse_error::create(101, m_lexer.get_position(), exception_message(token_type::value_string)));
}
else
{
2018-03-18 15:07:21 +00:00
if (JSON_UNLIKELY(not sax->key(m_lexer.move_string())))
{
return false;
}
}
// parse separator (:)
get_token();
if (JSON_UNLIKELY(last_token != token_type::name_separator))
{
return sax->parse_error(m_lexer.get_position(),
m_lexer.get_token_string(),
parse_error::create(101, m_lexer.get_position(), exception_message(token_type::name_separator)));
}
// remember we are now inside an object
states.push_back(parse_state_t::object_value);
// parse values
get_token();
continue;
}
case token_type::begin_array:
2018-02-24 17:04:07 +00:00
{
2018-03-18 15:07:21 +00:00
if (JSON_UNLIKELY(not sax->start_array()))
2018-02-24 17:04:07 +00:00
{
return false;
}
// read next token
get_token();
// closing ] -> we are done
if (last_token == token_type::end_array)
{
2018-03-18 15:07:21 +00:00
if (JSON_UNLIKELY(not sax->end_array()))
{
return false;
}
break;
}
// remember we are now inside an array
states.push_back(parse_state_t::array_value);
// parse values (no need to call get_token)
continue;
2018-02-24 17:04:07 +00:00
}
case token_type::value_float:
2018-02-24 17:04:07 +00:00
{
const auto res = m_lexer.get_number_float();
if (JSON_UNLIKELY(not std::isfinite(res)))
{
return sax->parse_error(m_lexer.get_position(),
m_lexer.get_token_string(),
out_of_range::create(406, "number overflow parsing '" + m_lexer.get_token_string() + "'"));
}
else
{
2018-03-18 15:07:21 +00:00
if (JSON_UNLIKELY(not sax->number_float(res, m_lexer.move_string())))
{
return false;
}
break;
}
2018-02-24 17:04:07 +00:00
}
case token_type::literal_false:
2018-02-24 17:04:07 +00:00
{
2018-03-18 15:07:21 +00:00
if (JSON_UNLIKELY(not sax->boolean(false)))
{
return false;
}
break;
2018-02-24 17:04:07 +00:00
}
case token_type::literal_null:
2018-02-24 17:04:07 +00:00
{
2018-03-18 15:07:21 +00:00
if (JSON_UNLIKELY(not sax->null()))
{
return false;
}
break;
2018-02-24 17:04:07 +00:00
}
case token_type::literal_true:
2018-02-24 17:04:07 +00:00
{
2018-03-18 15:07:21 +00:00
if (JSON_UNLIKELY(not sax->boolean(true)))
{
return false;
}
break;
2018-02-24 17:04:07 +00:00
}
2018-03-11 21:47:25 +00:00
case token_type::value_integer:
{
2018-03-18 15:07:21 +00:00
if (JSON_UNLIKELY(not sax->number_integer(m_lexer.get_number_integer())))
{
return false;
}
break;
2018-02-24 17:04:07 +00:00
}
case token_type::value_string:
2018-02-24 17:04:07 +00:00
{
2018-03-18 15:07:21 +00:00
if (JSON_UNLIKELY(not sax->string(m_lexer.move_string())))
{
return false;
}
break;
2018-02-24 17:04:07 +00:00
}
case token_type::value_unsigned:
2018-02-24 17:04:07 +00:00
{
2018-03-18 15:07:21 +00:00
if (JSON_UNLIKELY(not sax->number_unsigned(m_lexer.get_number_unsigned())))
{
return false;
}
break;
2018-02-24 17:04:07 +00:00
}
case token_type::parse_error:
2018-02-24 17:04:07 +00:00
{
// using "uninitialized" to avoid "expected" message
return sax->parse_error(m_lexer.get_position(),
m_lexer.get_token_string(),
parse_error::create(101, m_lexer.get_position(), exception_message(token_type::uninitialized)));
2018-02-24 17:04:07 +00:00
}
default: // the last token was unexpected
2018-02-24 17:04:07 +00:00
{
return sax->parse_error(m_lexer.get_position(),
m_lexer.get_token_string(),
parse_error::create(101, m_lexer.get_position(), exception_message(token_type::literal_or_value)));
2018-02-24 17:04:07 +00:00
}
}
}
else
2018-02-24 17:04:07 +00:00
{
skip_to_state_evaluation = false;
2018-02-24 17:04:07 +00:00
}
// we reached this line after we successfully parsed a value
if (states.empty())
2018-02-24 17:04:07 +00:00
{
// empty stack: we reached the end of the hieararchy: done
return true;
2018-02-24 17:04:07 +00:00
}
else
2018-02-24 17:04:07 +00:00
{
get_token();
switch (states.back())
{
case parse_state_t::array_value:
{
// comma -> next value
if (last_token == token_type::value_separator)
{
// parse a new value
get_token();
continue;
}
2018-02-24 17:04:07 +00:00
// closing ]
if (JSON_LIKELY(last_token == token_type::end_array))
{
2018-03-18 15:07:21 +00:00
if (JSON_UNLIKELY(not sax->end_array()))
{
return false;
}
// We are done with this array. Before we can parse
// a new value, we need to evaluate the new state
// first. By setting skip_to_state_evaluation to
// false, we are effectively jumping to the
// beginning of this switch.
assert(not states.empty());
states.pop_back();
skip_to_state_evaluation = true;
continue;
}
else
{
return sax->parse_error(m_lexer.get_position(),
m_lexer.get_token_string(),
parse_error::create(101, m_lexer.get_position(), exception_message(token_type::end_array)));
}
}
2018-02-24 17:04:07 +00:00
case parse_state_t::object_value:
{
// comma -> next value
if (last_token == token_type::value_separator)
{
get_token();
// parse key
if (JSON_UNLIKELY(last_token != token_type::value_string))
{
return sax->parse_error(m_lexer.get_position(),
m_lexer.get_token_string(),
parse_error::create(101, m_lexer.get_position(), exception_message(token_type::value_string)));
}
else
{
2018-03-18 15:07:21 +00:00
if (JSON_UNLIKELY(not sax->key(m_lexer.move_string())))
{
return false;
}
}
// parse separator (:)
get_token();
if (JSON_UNLIKELY(last_token != token_type::name_separator))
{
return sax->parse_error(m_lexer.get_position(),
m_lexer.get_token_string(),
parse_error::create(101, m_lexer.get_position(), exception_message(token_type::name_separator)));
}
// parse values
get_token();
continue;
}
// closing }
if (JSON_LIKELY(last_token == token_type::end_object))
{
2018-03-18 15:07:21 +00:00
if (JSON_UNLIKELY(not sax->end_object()))
{
return false;
}
// We are done with this object. Before we can
// parse a new value, we need to evaluate the new
// state first. By setting skip_to_state_evaluation
// to false, we are effectively jumping to the
// beginning of this switch.
assert(not states.empty());
states.pop_back();
skip_to_state_evaluation = true;
continue;
}
else
{
return sax->parse_error(m_lexer.get_position(),
m_lexer.get_token_string(),
parse_error::create(101, m_lexer.get_position(), exception_message(token_type::end_object)));
}
}
}
2018-02-24 17:04:07 +00:00
}
}
}
2017-08-14 15:02:40 +00:00
/// get next token from lexer
token_type get_token()
{
return (last_token = m_lexer.scan());
}
/*!
@throw parse_error.101 if expected token did not occur
*/
bool expect(token_type t)
{
if (JSON_UNLIKELY(t != last_token))
{
errored = true;
if (allow_exceptions)
{
JSON_THROW(parse_error::create(101, m_lexer.get_position(), exception_message(t)));
2017-08-14 15:02:40 +00:00
}
else
{
return false;
}
}
return true;
}
std::string exception_message(const token_type expected)
2017-08-14 15:02:40 +00:00
{
std::string error_msg = "syntax error - ";
if (last_token == token_type::parse_error)
{
error_msg += std::string(m_lexer.get_error_message()) + "; last read: '" +
m_lexer.get_token_string() + "'";
}
else
{
error_msg += "unexpected " + std::string(lexer_t::token_type_name(last_token));
}
if (expected != token_type::uninitialized)
{
error_msg += "; expected " + std::string(lexer_t::token_type_name(expected));
}
return error_msg;
2017-08-14 15:02:40 +00:00
}
private:
/// current level of recursion
int depth = 0;
/// callback function
const parser_callback_t callback = nullptr;
/// the type of the last read token
token_type last_token = token_type::uninitialized;
/// the lexer
lexer_t m_lexer;
/// whether a syntax error occurred
bool errored = false;
/// whether to throw exceptions in case of errors
const bool allow_exceptions = true;
};
}
}