template<template< typename U, typename V, typename... Args > class ObjectType = std::map, template< typename U, typename... Args > class ArrayType = std::vector, class StringType = std::string, class BooleanType = bool, class NumberIntegerType = std::int64_t, class NumberUnsignedType = std::uint64_t, class NumberFloatType = double, template< typename U > class AllocatorType = std::allocator, template< typename T, typename SFINAE=
void > class JSONSerializer = adl_serializer>
class nlohmann::basic_json< ObjectType, ArrayType, StringType, BooleanType, NumberIntegerType, NumberUnsignedType, NumberFloatType, AllocatorType, JSONSerializer >::lexer
lexical analysis
This class organizes the lexical analysis during JSON deserialization. The core of it is a scanner generated by re2c that processes a buffer and recognizes tokens according to RFC 7159.
Definition at line 9559 of file json.hpp.
template<template< typename U, typename V, typename... Args > class ObjectType = std::map, template< typename U, typename... Args > class ArrayType = std::vector, class StringType = std::string, class BooleanType = bool, class NumberIntegerType = std::int64_t, class NumberUnsignedType = std::uint64_t, class NumberFloatType = double, template< typename U > class AllocatorType = std::allocator, template< typename T, typename SFINAE=
void > class JSONSerializer = adl_serializer>
| void nlohmann::basic_json< ObjectType, ArrayType, StringType, BooleanType, NumberIntegerType, NumberUnsignedType, NumberFloatType, AllocatorType, JSONSerializer >::lexer::fill_line_buffer |
( |
size_t |
n = 0 | ) |
|
|
inline |
append data from the stream to the line buffer
This function is called by the scan() function when the end of the buffer (m_limit) is reached and the m_cursor pointer cannot be incremented without leaving the limits of the line buffer. Note re2c decides when to call this function.
If the lexer reads from contiguous storage, there is no trailing null byte. Therefore, this function must make sure to add these padding null bytes.
If the lexer reads from an input stream, this function reads the next line of the input.
- Precondition
- p p p p p p u u u u u x . . . . . . ^ ^ ^ ^ m_content m_start | m_limit m_cursor
- Postcondition
- u u u u u x x x x x x x . . . . . . ^ ^ ^ | m_cursor m_limit m_start m_content
Definition at line 10849 of file json.hpp.
template<template< typename U, typename V, typename... Args > class ObjectType = std::map, template< typename U, typename... Args > class ArrayType = std::vector, class StringType = std::string, class BooleanType = bool, class NumberIntegerType = std::int64_t, class NumberUnsignedType = std::uint64_t, class NumberFloatType = double, template< typename U > class AllocatorType = std::allocator, template< typename T, typename SFINAE=
void > class JSONSerializer = adl_serializer>
| bool nlohmann::basic_json< ObjectType, ArrayType, StringType, BooleanType, NumberIntegerType, NumberUnsignedType, NumberFloatType, AllocatorType, JSONSerializer >::lexer::get_number |
( |
basic_json & |
result, |
|
|
const token_type |
token |
|
) |
| const |
|
inline |
return number value for number tokens
This function translates the last token into the most appropriate number type (either integer, unsigned integer or floating point), which is passed back to the caller via the result parameter.
integral numbers that don't fit into the the range of the respective type are parsed as number_float_t
floating-point values do not satisfy std::isfinite predicate are converted to value_t::null
throws if the entire string [m_start .. m_cursor) cannot be interpreted as a number
- Parameters
-
| [out] | result | basic_json object to receive the number. |
| [in] | token | the type of the number token |
Definition at line 11259 of file json.hpp.
template<template< typename U, typename V, typename... Args > class ObjectType = std::map, template< typename U, typename... Args > class ArrayType = std::vector, class StringType = std::string, class BooleanType = bool, class NumberIntegerType = std::int64_t, class NumberUnsignedType = std::uint64_t, class NumberFloatType = double, template< typename U > class AllocatorType = std::allocator, template< typename T, typename SFINAE=
void > class JSONSerializer = adl_serializer>
| string_t nlohmann::basic_json< ObjectType, ArrayType, StringType, BooleanType, NumberIntegerType, NumberUnsignedType, NumberFloatType, AllocatorType, JSONSerializer >::lexer::get_string |
( |
| ) |
const |
|
inline |
return string value for string tokens
The function iterates the characters between the opening and closing quotes of the string value. The complete string is the range [m_start,m_cursor). Consequently, we iterate from m_start+1 to m_cursor-1.
We differentiate two cases:
- Escaped characters. In this case, a new character is constructed according to the nature of the escape. Some escapes create new characters (e.g.,
"\\n" is replaced by "\n"), some are copied as is (e.g., "\\\\"). Furthermore, Unicode escapes of the shape "\\uxxxx" need special care. In this case, to_unicode takes care of the construction of the values.
- Unescaped characters are copied as is.
- Precondition
m_cursor - m_start >= 2, meaning the length of the last token is at least 2 bytes which is trivially true for any string (which consists of at least two quotes). " c1 c2 c3 ... "
^ ^
m_start m_cursor
@complexity Linear in the length of the string.
Lemma: The loop body will always terminate.
Proof (by contradiction): Assume the loop body does not terminate. As the loop body does not contain another loop, one of the called functions must never return. The called functions are std::strtoul and to_unicode. Neither function can loop forever, so the loop body will never loop forever which contradicts the assumption that the loop body does not terminate, q.e.d.
Lemma: The loop condition for the for loop is eventually false.
Proof (by contradiction): Assume the loop does not terminate. Due to the above lemma, this can only be due to a tautological loop condition; that is, the loop condition i < m_cursor - 1 must always be true. Let x be the change of i for any loop iteration. Then m_start + 1 + x < m_cursor - 1 must hold to loop indefinitely. This can be rephrased to m_cursor - m_start - 2 > x. With the precondition, we x <= 0, meaning that the loop condition holds indefinitely if i is always decreased. However, observe that the value of i is strictly increasing with each iteration, as it is incremented by 1 in the iteration expression and never decremented inside the loop body. Hence, the loop condition will eventually be false which contradicts the assumption that the loop condition is a tautology, q.e.d.
- Returns
- string value of current token without opening and closing quotes
- Exceptions
-
| std::out_of_range | if to_unicode fails |
Definition at line 10975 of file json.hpp.
template<template< typename U, typename V, typename... Args > class ObjectType = std::map, template< typename U, typename... Args > class ArrayType = std::vector, class StringType = std::string, class BooleanType = bool, class NumberIntegerType = std::int64_t, class NumberUnsignedType = std::uint64_t, class NumberFloatType = double, template< typename U > class AllocatorType = std::allocator, template< typename T, typename SFINAE=
void > class JSONSerializer = adl_serializer>
| token_type nlohmann::basic_json< ObjectType, ArrayType, StringType, BooleanType, NumberIntegerType, NumberUnsignedType, NumberFloatType, AllocatorType, JSONSerializer >::lexer::scan |
( |
| ) |
|
|
inline |
This function implements a scanner for JSON.
It is specified using regular expressions that try to follow RFC 7159 as close as possible. These regular expressions are then translated into a minimized deterministic finite automaton (DFA) by the tool re2c. As a result, the translated code for this function consists of a large block of code with goto jumps.
- Returns
- the class of the next token read from the buffer
@complexity Linear in the length of the input.
Proposition: The loop below will always terminate for finite input.
Proof (by contradiction): Assume a finite input. To loop forever, the loop must never hit code with a break statement. The only code snippets without a break statement are the continue statements for whitespace and byte-order-marks. To loop forever, the input must be an infinite sequence of whitespace or byte-order-marks. This contradicts the assumption of finite input, q.e.d.
Definition at line 9773 of file json.hpp.
ROOT 6.16/01 - Reference Guide Generated on Sun Dec 19 2021 22:32:56 (GVA Time) using Doxygen 1.9.3 (234637167bd5d39d32bf51f755d58253441f123a).