Merge pull request #2297 from nlohmann/issue2286

Add support for high-precision numbers in UBJSON encoding

doc/mkdocs/docs/features/binary_formats/ubjson.md

@@ -28,6 +28,7 @@ number_unsigned | 128..255                          | uint8       | `U`
 number_unsigned | 256..32767                        | int16       | `I`
 number_unsigned | 32768..2147483647                 | int32       | `l`
 number_unsigned | 2147483648..9223372036854775807   | int64       | `L`
+number_unsigned | 2147483649..18446744073709551615  | high-precision | `H`
 number_float    | *any value*                       | float64     | `D`
 string          | *with shortest length indicator*  | string      | `S`
 array           | *see notes on optimized format*   | array       | `[`
@@ -44,7 +45,6 @@ object          | *see notes on optimized format*   | map         | `{`
 	The following values can **not** be converted to a UBJSON value:
 
       - strings with more than 9223372036854775807 bytes (theoretical)
-      - unsigned integer numbers above 9223372036854775807
 
 !!! info "Unused UBJSON markers"
 

doc/mkdocs/docs/home/exceptions.md

@@ -279,6 +279,16 @@ The parsing of the corresponding BSON record type is not implemented (yet).
     [json.exception.parse_error.114] parse error at byte 5: Unsupported BSON record type 0xFF
     ```
 
+### json.exception.parse_error.115
+
+A UBJSON high-precision number could not be parsed.
+
+!!! failure "Example message"
+
+    ```
+    [json.exception.parse_error.115] parse error at byte 5: syntax error while parsing UBJSON high-precision number: invalid number text: 1A
+    ```
+
 ## Iterator errors
 
 This exception is thrown if iterators passed to a library function do not match
@@ -765,6 +775,10 @@ UBJSON and BSON only support integer numbers up to 9223372036854775807.
     number overflow serializing '9223372036854775808'
     ```
 
+!!! note
+
+    Since version 3.9.0, integer numbers beyond int64 are serialized as high-precision UBJSON numbers, and this exception does not further occur. 
+
 ### json.exception.out_of_range.408
 
 The size (following `#`) of an UBJSON array or object exceeds the maximal capacity.

include/nlohmann/detail/exceptions.hpp

@@ -97,6 +97,7 @@ json.exception.parse_error.110 | parse error at 1: cannot read 2 bytes from vect
 json.exception.parse_error.112 | parse error at 1: error reading CBOR; last byte: 0xF8 | Not all types of CBOR or MessagePack are supported. This exception occurs if an unsupported byte was read.
 json.exception.parse_error.113 | parse error at 2: expected a CBOR string; last byte: 0x98 | While parsing a map key, a value that is not a string has been read.
 json.exception.parse_error.114 | parse error: Unsupported BSON record type 0x0F | The parsing of the corresponding BSON record type is not implemented (yet).
+json.exception.parse_error.115 | parse error at byte 5: syntax error while parsing UBJSON high-precision number: invalid number text: 1A | A UBJSON high-precision number could not be parsed.
 
 @note For an input with n bytes, 1 is the index of the first character and n+1
       is the index of the terminating null byte or the end of file. This also
@@ -285,7 +286,7 @@ json.exception.out_of_range.403 | key 'foo' not found | The provided key was not
 json.exception.out_of_range.404 | unresolved reference token 'foo' | A reference token in a JSON Pointer could not be resolved.
 json.exception.out_of_range.405 | JSON pointer has no parent | The JSON Patch operations 'remove' and 'add' can not be applied to the root element of the JSON value.
 json.exception.out_of_range.406 | number overflow parsing '10E1000' | A parsed number could not be stored as without changing it to NaN or INF.
-json.exception.out_of_range.407 | number overflow serializing '9223372036854775808' | UBJSON and BSON only support integer numbers up to 9223372036854775807. |
+json.exception.out_of_range.407 | number overflow serializing '9223372036854775808' | UBJSON and BSON only support integer numbers up to 9223372036854775807. (until version 3.8.0) |
 json.exception.out_of_range.408 | excessive array size: 8658170730974374167 | The size (following `#`) of an UBJSON array or object exceeds the maximal capacity. |
 json.exception.out_of_range.409 | BSON key cannot contain code point U+0000 (at byte 2) | Key identifiers to be serialized to BSON cannot contain code point U+0000, since the key is stored as zero-terminated c-string |
 

include/nlohmann/detail/input/binary_reader.hpp

@@ -15,6 +15,7 @@
 #include <nlohmann/detail/exceptions.hpp>
 #include <nlohmann/detail/input/input_adapters.hpp>
 #include <nlohmann/detail/input/json_sax.hpp>
+#include <nlohmann/detail/input/lexer.hpp>
 #include <nlohmann/detail/macro_scope.hpp>
 #include <nlohmann/detail/meta/is_sax.hpp>
 #include <nlohmann/detail/value_t.hpp>
@@ -2004,6 +2005,11 @@ class binary_reader
                 return get_number(input_format_t::ubjson, number) && sax->number_float(static_cast<number_float_t>(number), "");
             }
 
+            case 'H':
+            {
+                return get_ubjson_high_precision_number();
+            }
+
             case 'C':  // char
             {
                 get();
@@ -2180,6 +2186,55 @@ class binary_reader
     // Note, no reader for UBJSON binary types is implemented because they do
     // not exist
 
+    bool get_ubjson_high_precision_number()
+    {
+        // get size of following number string
+        std::size_t size{};
+        auto res = get_ubjson_size_value(size);
+        if (JSON_HEDLEY_UNLIKELY(!res))
+        {
+            return res;
+        }
+
+        // get number string
+        std::vector<char> number_vector;
+        for (std::size_t i = 0; i < size; ++i)
+        {
+            get();
+            if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::ubjson, "number")))
+            {
+                return false;
+            }
+            number_vector.push_back(static_cast<char>(current));
+        }
+
+        // parse number string
+        auto number_ia = detail::input_adapter(std::forward<decltype(number_vector)>(number_vector));
+        auto number_lexer = detail::lexer<BasicJsonType, decltype(number_ia)>(std::move(number_ia), false);
+        const auto result_number = number_lexer.scan();
+        const auto number_string = number_lexer.get_token_string();
+        const auto result_remainder = number_lexer.scan();
+
+        using token_type = typename detail::lexer_base<BasicJsonType>::token_type;
+
+        if (JSON_HEDLEY_UNLIKELY(result_remainder != token_type::end_of_input))
+        {
+            return sax->parse_error(chars_read, number_string, parse_error::create(115, chars_read, exception_message(input_format_t::ubjson, "invalid number text: " + number_lexer.get_token_string(), "high-precision number")));
+        }
+
+        switch (result_number)
+        {
+            case token_type::value_integer:
+                return sax->number_integer(number_lexer.get_number_integer());
+            case token_type::value_unsigned:
+                return sax->number_unsigned(number_lexer.get_number_unsigned());
+            case token_type::value_float:
+                return sax->number_float(number_lexer.get_number_float(), std::move(number_string));
+            default:
+                return sax->parse_error(chars_read, number_string, parse_error::create(115, chars_read, exception_message(input_format_t::ubjson, "invalid number text: " + number_lexer.get_token_string(), "high-precision number")));
+        }
+    }
+
     ///////////////////////
     // Utility functions //
     ///////////////////////

include/nlohmann/detail/output/binary_writer.hpp

@@ -1319,7 +1319,17 @@ class binary_writer
         }
         else
         {
-            JSON_THROW(out_of_range::create(407, "integer number " + std::to_string(n) + " cannot be represented by UBJSON as it does not fit int64"));
+            if (add_prefix)
+            {
+                oa->write_character(to_char_type('H'));  // high-precision number
+            }
+
+            const auto number = BasicJsonType(n).dump();
+            write_number_with_ubjson_prefix(number.size(), true);
+            for (std::size_t i = 0; i < number.size(); ++i)
+            {
+                oa->write_character(to_char_type(static_cast<std::uint8_t>(number[i])));
+            }
         }
     }
 
@@ -1373,19 +1383,23 @@ class binary_writer
         // LCOV_EXCL_START
         else
         {
-            JSON_THROW(out_of_range::create(407, "integer number " + std::to_string(n) + " cannot be represented by UBJSON as it does not fit int64"));
+            if (add_prefix)
+            {
+                oa->write_character(to_char_type('H'));  // high-precision number
+            }
+
+            const auto number = BasicJsonType(n).dump();
+            write_number_with_ubjson_prefix(number.size(), true);
+            for (std::size_t i = 0; i < number.size(); ++i)
+            {
+                oa->write_character(to_char_type(static_cast<std::uint8_t>(number[i])));
+            }
         }
         // LCOV_EXCL_STOP
     }
 
     /*!
     @brief determine the type prefix of container values
-
-    @note This function does not need to be 100% accurate when it comes to
-          integer limits. In case a number exceeds the limits of int64_t,
-          this will be detected by a later call to function
-          write_number_with_ubjson_prefix. Therefore, we return 'L' for any
-          value that does not fit the previous limits.
     */
     CharType ubjson_prefix(const BasicJsonType& j) const noexcept
     {
@@ -1415,8 +1429,12 @@ class binary_writer
                 {
                     return 'l';
                 }
-                // no check and assume int64_t (see note above)
-                return 'L';
+                if ((std::numeric_limits<std::int64_t>::min)() <= j.m_value.number_integer && j.m_value.number_integer <= (std::numeric_limits<std::int64_t>::max)())
+                {
+                    return 'L';
+                }
+                // anything else is treated as high-precision number
+                return 'H'; // LCOV_EXCL_LINE
             }
 
             case value_t::number_unsigned:
@@ -1437,8 +1455,12 @@ class binary_writer
                 {
                     return 'l';
                 }
-                // no check and assume int64_t (see note above)
-                return 'L';
+                if (j.m_value.number_unsigned <= static_cast<std::uint64_t>((std::numeric_limits<std::int64_t>::max)()))
+                {
+                    return 'L';
+                }
+                // anything else is treated as high-precision number
+                return 'H'; // LCOV_EXCL_LINE
             }
 
             case value_t::number_float:

include/nlohmann/json.hpp

@@ -7201,6 +7201,7 @@ class basic_json
     number_unsigned | 256..32767                        | int16       | `I`
     number_unsigned | 32768..2147483647                 | int32       | `l`
     number_unsigned | 2147483648..9223372036854775807   | int64       | `L`
+    number_unsigned | 2147483649..18446744073709551615  | high-precision | `H`
     number_float    | *any value*                       | float64     | `D`
     string          | *with shortest length indicator*  | string      | `S`
     array           | *see notes on optimized format*   | array       | `[`
@@ -7211,7 +7212,6 @@ class basic_json
 
     @note The following values can **not** be converted to a UBJSON value:
           - strings with more than 9223372036854775807 bytes (theoretical)
-          - unsigned integer numbers above 9223372036854775807
 
     @note The following markers are not used in the conversion:
           - `Z`: no-op values are not created.
@@ -7687,6 +7687,7 @@ class basic_json
     int16       | number_integer                          | `I`
     int32       | number_integer                          | `l`
     int64       | number_integer                          | `L`
+    high-precision number | number_integer, number_unsigned, or number_float - depends on number string | 'H'
     string      | string                                  | `S`
     char        | string                                  | `C`
     array       | array (optimized values are supported)  | `[`

test/src/unit-ubjson.cpp

@@ -32,6 +32,7 @@ SOFTWARE.
 #include <nlohmann/json.hpp>
 using nlohmann::json;
 
+#include <iostream>
 #include <fstream>
 #include <set>
 #include <test_data.hpp>
@@ -768,6 +769,65 @@ TEST_CASE("UBJSON")
                     CHECK(json::from_ubjson(result, true, false) == j);
                 }
             }
+
+            SECTION("high-precision number")
+            {
+                SECTION("unsigned integer number")
+                {
+                    std::vector<uint8_t> vec = {'H', 'i', 0x14, '1', '2', '3', '4', '5', '6', '7', '8', '9', '0', '1', '2', '3', '4', '5', '6', '7', '8', '9', '0'};
+                    const auto j = json::from_ubjson(vec);
+                    CHECK(j.is_number_unsigned());
+                    CHECK(j.dump() == "12345678901234567890");
+                }
+
+                SECTION("signed integer number")
+                {
+                    std::vector<uint8_t> vec = {'H', 'i', 0x13, '-', '1', '2', '3', '4', '5', '6', '7', '8', '9', '0', '1', '2', '3', '4', '5', '6', '7', '8'};
+                    const auto j = json::from_ubjson(vec);
+                    CHECK(j.is_number_integer());
+                    CHECK(j.dump() == "-123456789012345678");
+                }
+
+                SECTION("floating-point number")
+                {
+                    std::vector<uint8_t> vec = {'H', 'i', 0x16, '3', '.', '1', '4', '1', '5', '9',  '2', '6', '5', '3', '5', '8', '9',  '7', '9', '3', '2', '3', '8', '4',  '6'};
+                    const auto j = json::from_ubjson(vec);
+                    CHECK(j.is_number_float());
+                    CHECK(j.dump() == "3.141592653589793");
+                }
+
+                SECTION("errors")
+                {
+                    // error while parsing length
+                    std::vector<uint8_t> vec0 = {'H', 'i'};
+                    CHECK(json::from_ubjson(vec0, true, false).is_discarded());
+                    // error while parsing string
+                    std::vector<uint8_t> vec1 = {'H', 'i', '1'};
+                    CHECK(json::from_ubjson(vec1, true, false).is_discarded());
+
+                    std::vector<uint8_t> vec2 = {'H', 'i', 2, '1', 'A', '3'};
+                    CHECK_THROWS_WITH_AS(json::from_ubjson(vec2), "[json.exception.parse_error.115] parse error at byte 5: syntax error while parsing UBJSON high-precision number: invalid number text: 1A", json::parse_error);
+                    std::vector<uint8_t> vec3 = {'H', 'i', 2, '1', '.'};
+                    CHECK_THROWS_WITH_AS(json::from_ubjson(vec3), "[json.exception.parse_error.115] parse error at byte 5: syntax error while parsing UBJSON high-precision number: invalid number text: 1.", json::parse_error);
+                    std::vector<uint8_t> vec4 = {'H', 2, '1', '0'};
+                    CHECK_THROWS_WITH_AS(json::from_ubjson(vec4), "[json.exception.parse_error.113] parse error at byte 2: syntax error while parsing UBJSON size: expected length type specification (U, i, I, l, L) after '#'; last byte: 0x02", json::parse_error);
+                }
+
+                SECTION("serialization")
+                {
+                    // number that does not fit int64
+                    json j = 11111111111111111111ULL;
+                    CHECK(j.is_number_unsigned());
+
+                    // number will be serialized to high-precision number
+                    const auto vec = json::to_ubjson(j);
+                    std::vector<uint8_t> expected = {'H', 'i', 0x14, '1',  '1',  '1',  '1',  '1', '1',  '1',  '1',  '1',  '1', '1',  '1',  '1',  '1',  '1', '1',  '1',  '1',  '1',  '1'};
+                    CHECK(vec == expected);
+
+                    // roundtrip
+                    CHECK(json::from_ubjson(vec) == j);
+                }
+            }
         }
 
         SECTION("string")
@@ -1540,15 +1600,6 @@ TEST_CASE("UBJSON")
             }
         }
 
-        SECTION("number out of range")
-        {
-            // larger than max int64
-            json j = 9223372036854775808llu;
-            json _;
-            CHECK_THROWS_AS(_ = json::to_ubjson(j), json::out_of_range&);
-            CHECK_THROWS_WITH(_ = json::to_ubjson(j), "[json.exception.out_of_range.407] integer number 9223372036854775808 cannot be represented by UBJSON as it does not fit int64");
-        }
-
         SECTION("excessive size")
         {
             SECTION("array")
@@ -2370,7 +2421,7 @@ TEST_CASE("all UBJSON first bytes")
     // these bytes will fail immediately with exception parse_error.112
     std::set<uint8_t> supported =
     {
-        'T', 'F', 'Z', 'U', 'i', 'I', 'l', 'L', 'd', 'D', 'C', 'S', '[', '{', 'N'
+        'T', 'F', 'Z', 'U', 'i', 'I', 'l', 'L', 'd', 'D', 'C', 'S', '[', '{', 'N', 'H'
     };
 
     for (auto i = 0; i < 256; ++i)