json_view: source positions in materialize() with JSON_DIAGNOSTIC_POSITIONS

With JSON_DIAGNOSTIC_POSITIONS, materialize() now gives every value the
start_pos()/end_pos() that basic_json::parse() would give it, so exceptions
about materialized values name the same bytes (#5864).

- json_sax_dom_parser takes the lexer type as a defaulted template
  parameter; materialize() passes position_source, which reports the
  positions it derives from the index. diagnostic_positions treats npos as
  "no position" and keeps ranges all-or-nothing.
- Strings with escapes keep the offset of their opening quote plus one in
  `next` (only with the macro); materialize() scans from there to the closing
  quote. The end of an array or object is found by scanning from the end of
  its last element over whitespace, comments and a trailing comma.
- In editable documents, values written by edits, arrays and objects whose
  elements changed, and containers whose last element has no position get
  none. Images keep the flag is_new, and load() checks that `next` of a
  string is within the text.

Nothing changes on the parse path without the macro. The view's own
exceptions still carry no byte range, so the tests that compare them with
those of materialized values skip it under the macro, as they skip the path
under JSON_DIAGNOSTICS.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

tests/src/unit-json_view.cpp

@@ -749,6 +749,20 @@ std::string exception_of(F f)
     }
     return "";
 }
+
+// the exception a call on a basic_json value throws, without the byte range of
+// the value (JSON_DIAGNOSTIC_POSITIONS), which the exceptions of views lack
+template<typename F>
+std::string json_exception_of(F f)
+{
+    std::string what = exception_of(f);
+    const std::size_t at = what.find("] (bytes ");
+    if (at != std::string::npos)
+    {
+        what.erase(at + 2, what.find(") ", at) - at);
+    }
+    return what;
+}
 #endif
 
 // compares a view with the ordered_json value materialize() gives for it:
@@ -957,24 +971,24 @@ TEST_CASE("json_view element access and iteration")
 #if !defined(JSON_NOEXCEPTION)
             if (!j.is_object())
             {
-                CHECK(exception_of([&] { static_cast<void>(v["a"]); }) == exception_of([&] { static_cast<void>(j["a"]); }));
+                CHECK(exception_of([&] { static_cast<void>(v["a"]); }) == json_exception_of([&] { static_cast<void>(j["a"]); }));
             }
             if (!j.is_array())
             {
-                CHECK(exception_of([&] { static_cast<void>(v[0]); }) == exception_of([&] { static_cast<void>(j[0]); }));
+                CHECK(exception_of([&] { static_cast<void>(v[0]); }) == json_exception_of([&] { static_cast<void>(j[0]); }));
             }
-            CHECK(exception_of([&] { static_cast<void>(v.at("a")); }) == exception_of([&] { static_cast<void>(j.at("a")); }));
-            CHECK(exception_of([&] { static_cast<void>(v.at("missing")); }) == exception_of([&] { static_cast<void>(j.at("missing")); }));
-            CHECK(exception_of([&] { static_cast<void>(v.at(0)); }) == exception_of([&] { static_cast<void>(j.at(0)); }));
-            CHECK(exception_of([&] { static_cast<void>(v.at(5)); }) == exception_of([&] { static_cast<void>(j.at(5)); }));
+            CHECK(exception_of([&] { static_cast<void>(v.at("a")); }) == json_exception_of([&] { static_cast<void>(j.at("a")); }));
+            CHECK(exception_of([&] { static_cast<void>(v.at("missing")); }) == json_exception_of([&] { static_cast<void>(j.at("missing")); }));
+            CHECK(exception_of([&] { static_cast<void>(v.at(0)); }) == json_exception_of([&] { static_cast<void>(j.at(0)); }));
+            CHECK(exception_of([&] { static_cast<void>(v.at(5)); }) == json_exception_of([&] { static_cast<void>(j.at(5)); }));
             if (!(j.is_object() && j.empty())) // (key() of an end iterator)
             {
-                CHECK(exception_of([&] { static_cast<void>(v.begin().key()); }) == exception_of([&] { static_cast<void>(j.begin().key()); }));
+                CHECK(exception_of([&] { static_cast<void>(v.begin().key()); }) == json_exception_of([&] { static_cast<void>(j.begin().key()); }));
             }
             if (!j.empty() || j.is_null())
             {
-                CHECK(exception_of([&] { static_cast<void>(v.front()); }) == exception_of([&] { static_cast<void>(j.front()); }));
-                CHECK(exception_of([&] { static_cast<void>(v.back()); }) == exception_of([&] { static_cast<void>(j.back()); }));
+                CHECK(exception_of([&] { static_cast<void>(v.front()); }) == json_exception_of([&] { static_cast<void>(j.front()); }));
+                CHECK(exception_of([&] { static_cast<void>(v.back()); }) == json_exception_of([&] { static_cast<void>(j.back()); }));
             }
 #endif
             CHECK(v.contains("a") == j.contains("a"));

tests/src/unit-json_view_from_json.cpp

@@ -326,7 +326,8 @@ TEST_CASE("json_view from_json")
         CHECK(d.root().get<std::unordered_map<std::string, int>>().at("a") == 3);
     }
 
-#if !defined(JSON_NOEXCEPTION) && !JSON_DIAGNOSTICS
+// (the exceptions of views have neither the path nor the byte range of the value)
+#if !defined(JSON_NOEXCEPTION) && !JSON_DIAGNOSTICS && !JSON_DIAGNOSTIC_POSITIONS
     SECTION("std types: the same errors as from the materialized value")
     {
         const json_document d = json_document::parse(R"([[1, "x"], {"a": 1}, "text", [1], 5, null])");
@@ -392,7 +393,7 @@ TEST_CASE("json_view from_json")
 #if !defined(JSON_NOEXCEPTION)
         const json j = d.root().materialize();
         CHECK(exception_of_call([&] { static_cast<void>(d.root()[2].get<strict_color>()); }) == "[json.exception.out_of_range.410] enum value out of range for strict_color: \"purple\"");
-#if !JSON_DIAGNOSTICS
+#if !JSON_DIAGNOSTICS && !JSON_DIAGNOSTIC_POSITIONS
         CHECK(exception_of_call([&] { static_cast<void>(d.root()[2].get<strict_color>()); }) == exception_of_call([&] { static_cast<void>(j[2].get<strict_color>()); }));
 #endif
 #endif
@@ -440,10 +441,11 @@ TEST_CASE("json_view from_json")
         CHECK(o.root().get<shape>().center.y == 2);
     }
 
-#if !defined(JSON_NOEXCEPTION) && !JSON_DIAGNOSTICS
+#if !defined(JSON_NOEXCEPTION) && !JSON_DIAGNOSTICS && !JSON_DIAGNOSTIC_POSITIONS
     SECTION("errors are those of the same conversion of a json value")
     {
-        // (with JSON_DIAGNOSTICS, those of json name the path of the value)
+        // (with JSON_DIAGNOSTICS, those of json name the path of the value,
+        // and with JSON_DIAGNOSTIC_POSITIONS its bytes)
         const json_document d = json_document::parse(R"([{"x": 1}, {"x": "1", "y": 2}, [1, 2], {"id": "7", "c": "red"}])");
         const json j = d.root().materialize();
         for (std::size_t i = 0; i < 3; ++i)

tests/src/unit-json_view_positions.cpp

@@ -0,0 +1,340 @@
+//     __ _____ _____ _____
+//  __|  |   __|     |   | |  JSON for Modern C++ (supporting code)
+// |  |  |__   |  |  | | | |  version 3.12.0
+// |_____|_____|_____|_|___|  https://github.com/nlohmann/json
+//
+// SPDX-FileCopyrightText: 2013-2026 Niels Lohmann <https://nlohmann.me>
+// SPDX-License-Identifier: MIT
+
+// source positions of materialize() (JSON_DIAGNOSTIC_POSITIONS, #5864)
+
+#include "doctest_compatibility.h"
+
+#ifndef JSON_DIAGNOSTICS
+    #define JSON_DIAGNOSTICS 1
+#endif
+#define JSON_DIAGNOSTIC_POSITIONS 1
+#include <nlohmann/json_view.hpp>
+using nlohmann::json;
+using nlohmann::ordered_json;
+using nlohmann::json_document;
+using nlohmann::json_editable_document;
+
+#include <cstdint>
+#include <cstring>
+#include <fstream>
+#include <sstream>
+#include <string>
+#include <vector>
+
+#include "json_view_test_helpers.hpp"
+#include <test_data.hpp>
+
+namespace
+{
+constexpr std::size_t npos = std::string::npos;
+
+// whether every value of a has the source positions of the value of b at the
+// same place (a and b are equal; shallow test documents only)
+template<typename BasicJsonType>
+bool same_positions(const BasicJsonType& a, const BasicJsonType& b)
+{
+    if (a.start_pos() != b.start_pos() || a.end_pos() != b.end_pos() || a.type() != b.type() || a.size() != b.size())
+    {
+        return false;
+    }
+    if (a.is_object())
+    {
+        for (auto it = a.begin(); it != a.end(); ++it)
+        {
+            if (!same_positions(it.value(), b.at(it.key())))
+            {
+                return false;
+            }
+        }
+    }
+    else if (a.is_array())
+    {
+        for (std::size_t i = 0; i < a.size(); ++i)
+        {
+            if (!same_positions(a[i], b[i]))
+            {
+                return false;
+            }
+        }
+    }
+    return true;
+}
+
+// the materialized root of text has the positions of parse()
+template<typename BasicJsonType>
+bool positions_as_parse(const std::string& text, bool comments = false, bool trailing_commas = false)
+{
+    const auto j = BasicJsonType::parse(text, nullptr, true, comments, trailing_commas);
+    const auto d = nlohmann::basic_json_document<BasicJsonType>::parse(text, true, comments, trailing_commas);
+    const auto e = nlohmann::basic_json_document<BasicJsonType, nlohmann::view_mode::editable>::parse(text, true, comments, trailing_commas);
+    return same_positions(d.root().materialize(), j) && same_positions(e.root().materialize(), j);
+}
+
+std::string range(const json& j)
+{
+    return j.start_pos() == npos && j.end_pos() == npos ? "none" : std::to_string(j.start_pos()) + "-" + std::to_string(j.end_pos());
+}
+} // namespace
+
+TEST_CASE("json_view materialize() source positions")
+{
+    SECTION("as parse()")
+    {
+        const std::vector<std::string> texts =
+        {
+            "null", "true", "false", "0", "-12", "3.25e-2", "18446744073709551615", "-9223372036854775809", R"("a")",
+            R"("\n")", R"("\u00e9")", R"("\ud83d\ude00")", "\"\xf0\x9f\x98\x80\"", R"("a\"b\\")", R"("x\\")",
+            "  [ ]  ", "{}", "[[[]]]", R"({"a":{"b":[1,{"c":[]}]}})",
+            R"({"\u0041": "v\n", "k": "\"", "\t":[ "\/" ]})",
+            "[1, 2.5, true, false, null, \"s\", \"\\u0000\", -0]",
+            "\n\t[\r\n  1 ,\n  {\n    \"a\" : [ ]\n  }\n]\n",
+            "\xef\xbb\xbf[1, {\"a\": \"b\"}]", "\xef\xbb\xbf\"\\n\"",
+        };
+        for (const auto& text : texts)
+        {
+            CAPTURE(text)
+            CHECK(positions_as_parse<json>(text));
+            CHECK(positions_as_parse<ordered_json>(text));
+        }
+    }
+
+    SECTION("comments and trailing commas")
+    {
+        const std::vector<std::string> texts =
+        {
+            "[1 /* ] */]", "[1 // ]\n]", "{\"a\": 1 /* } */ , }", "[1, 2, ]", "[[1,],[2 , /**/ ] , ]",
+            "{\"a\": [] // }\n, \"b\": {} /* } ] */ }", "[ /* empty */ ]", "{ // }\n}", "/* a */ [1] // b",
+            "[\"\\n\" /* \" */ , ]",
+        };
+        for (const auto& text : texts)
+        {
+            CAPTURE(text)
+            CHECK(positions_as_parse<json>(text, true, true));
+        }
+    }
+
+    SECTION("generated documents")
+    {
+        json_view_test::generator gen;
+        for (int i = 0; i < 300; ++i)
+        {
+            std::string text;
+            gen.value(text, 0);
+            CAPTURE(text)
+            CHECK(positions_as_parse<json>(text));
+            CHECK(positions_as_parse<ordered_json>(text));
+        }
+    }
+
+    SECTION("test data")
+    {
+        for (const char* name :
+                {
+                    "/json_testsuite/sample.json", "/nativejson-benchmark/canada.json", "/nativejson-benchmark/citm_catalog.json",
+                    "/nativejson-benchmark/twitter.json", "/json_tests/pass1.json"
+                })
+        {
+            CAPTURE(name)
+            const std::ifstream f(std::string(TEST_DATA_DIRECTORY) + name, std::ios::binary);
+            std::stringstream ss;
+            ss << f.rdbuf();
+            CHECK(positions_as_parse<json>(ss.str()));
+        }
+    }
+
+    SECTION("subtrees")
+    {
+        const std::string text = R"({"a": [1, "\n", {"b": null}], "s": "x\ty", "n": -1.5})";
+        const json j = json::parse(text);
+        const json_document d = json_document::parse(text);
+        CHECK(same_positions(d.root()["a"].materialize(), j["a"]));
+        CHECK(same_positions(d.root()["a"][2].materialize(), j["a"][2]));
+        CHECK(same_positions(d.root()["a"][1].materialize(), j["a"][1]));
+        CHECK(same_positions(d.root()["s"].materialize(), j["s"]));
+        CHECK(same_positions(d.root()["n"].materialize(), j["n"]));
+        CHECK(range(d.root()["a"][1].materialize()) == "10-14");
+        CHECK(range(d.root()["s"].materialize()) == "35-41");
+    }
+
+#if !defined(JSON_NOEXCEPTION)
+    SECTION("exception messages")
+    {
+        const std::string text = "{\n  \"address\": {\"street\": \"Fake\\tStreet\", \"housenumber\": \"1\"},\n  \"list\": [1, [true]]\n}";
+        const json j = json::parse(text);
+        const json_document d = json_document::parse(text);
+        const json m = d.root().materialize();
+        const std::string expected = json_view_test::exception_of_call([&]
+        {
+            static_cast<void>(j.at("address").at("housenumber").get<int>());
+        });
+        CHECK(expected.find("(bytes ") != std::string::npos);
+        CHECK(json_view_test::exception_of_call([&]
+        {
+            static_cast<void>(m.at("address").at("housenumber").get<int>());
+        }) == expected);
+        CHECK(json_view_test::exception_of_call([&]
+        {
+            static_cast<void>(m.at("address").at("street").get<int>());
+        }) == json_view_test::exception_of_call([&]
+        {
+            static_cast<void>(j.at("address").at("street").get<int>());
+        }));
+        CHECK(json_view_test::exception_of_call([&]
+        {
+            static_cast<void>(m.at("list").at(1).get<int>());
+        }) == json_view_test::exception_of_call([&]
+        {
+            static_cast<void>(j.at("list").at(1).get<int>());
+        }));
+    }
+#endif
+
+    SECTION("edited documents")
+    {
+        const std::string text = R"({"a": [1, 2], "b": {"c": "\n", "d": true}, "e": [3, [4]], "f": "x"})";
+        const json j = json::parse(text);
+
+        SECTION("new scalars have none")
+        {
+            json_editable_document d = json_editable_document::parse(text);
+            d.set(d.root()["f"], 5);
+            d.set(d.root()["b"]["d"], "s");
+            const json m = d.root().materialize();
+            CHECK(range(m["f"]) == "none");
+            CHECK(range(m["b"]["d"]) == "none");
+            // the object whose last member is new has none; the others keep theirs
+            CHECK(range(m) == "none");
+            CHECK(range(m["b"]) == "none");
+            CHECK(same_positions(m["a"], j["a"]));
+            CHECK(same_positions(m["e"], j["e"]));
+            CHECK(same_positions(m["b"]["c"], j["b"]["c"]));
+        }
+
+        SECTION("containers whose elements change have none")
+        {
+            json_editable_document d = json_editable_document::parse(text);
+            d.push_back(d.root()["a"], 7);
+            d.erase(d.root()["e"], 0);
+            const json m = d.root().materialize();
+            CHECK(range(m["a"]) == "none");
+            CHECK(range(m["a"][2]) == "none");
+            CHECK(range(m["e"]) == "none");
+            // the elements kept keep theirs
+            CHECK(range(m["a"][0]) == range(j["a"][0]));
+            CHECK(range(m["a"][1]) == range(j["a"][1]));
+            CHECK(same_positions(m["e"][0], j["e"][1]));
+            CHECK(same_positions(m["b"], j["b"]));
+            // the root's elements did not change
+            CHECK(range(m) == range(j));
+        }
+
+        SECTION("new containers have none")
+        {
+            json_editable_document d = json_editable_document::parse(text);
+            d.set(d.root()["b"], json::parse(R"({"x": [1]})"));
+            const json m = d.root().materialize();
+            CHECK(range(m["b"]) == "none");
+            CHECK(range(m["b"]["x"]) == "none");
+            CHECK(range(m["b"]["x"][0]) == "none");
+            CHECK(range(m) == range(j));
+        }
+
+        SECTION("copies of scalars of the document keep theirs")
+        {
+            // (a copy of an array or object is a new one; its scalars are the
+            // parsed ones)
+            json_editable_document d = json_editable_document::parse(text);
+            d.set(d.root()["f"], d.root()["e"]);
+            d.set(d.root()["a"], d.root()["b"]["c"]);
+            const json m = d.root().materialize();
+            CHECK(range(m["a"]) == range(j["b"]["c"]));
+            CHECK(range(m["f"]) == "none");
+            CHECK(range(m["f"][1]) == "none");
+            CHECK(range(m["f"][0]) == range(j["e"][0]));
+            CHECK(range(m["f"][1][0]) == range(j["e"][1][0]));
+        }
+    }
+
+#if !(defined(__BYTE_ORDER__) && defined(__ORDER_BIG_ENDIAN__) && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
+    SECTION("images")
+    {
+        const std::string text = R"({"a\n": ["\u00e9", 1.5, {"b": [true, null]}], "c": "d"})";
+        const json j = json::parse(text);
+
+        SECTION("round trip")
+        {
+            const json_document d = json_document::load(json_document::parse(text).save());
+            CHECK(same_positions(d.root().materialize(), j));
+        }
+
+        SECTION("edited")
+        {
+            json_editable_document e = json_editable_document::parse(text);
+            e.set(e.root()["c"], 1);
+            e.push_back(e.root()[std::string("a") + '\n'], false);
+            e.set(e.root()[std::string("a") + '\n'][2]["b"][0], "\"");
+            const json_document d = json_document::load(e.save());
+            const json m = d.root().materialize();
+            CHECK(range(m["c"]) == "none");
+            CHECK(range(m) == "none");
+            const json& a = m[std::string("a") + '\n'];
+            const json& ja = j[std::string("a") + '\n'];
+            CHECK(range(a) == "none");
+            CHECK(range(a[3]) == "none");
+            CHECK(range(a[0]) == range(ja[0]));
+            CHECK(range(a[1]) == range(ja[1]));
+            CHECK(range(a[2]["b"][0]) == "none");
+            CHECK(range(a[2]["b"][1]) == range(ja[2]["b"][1]));
+            // the last element of b is still the parsed one
+            CHECK(range(a[2]["b"]) == range(ja[2]["b"]));
+            CHECK(range(a[2]) == range(ja[2]));
+        }
+
+        SECTION("without positions")
+        {
+            // an image saved without JSON_DIAGNOSTIC_POSITIONS: next is 0 for
+            // the strings with escapes
+            std::vector<std::uint8_t> image = json_document::parse(text).save();
+            std::uint64_t count = 0;
+            std::memcpy(&count, image.data() + 8, sizeof(count));
+            for (std::size_t i = 0; i < count; ++i)
+            {
+                nlohmann::detail::view::node n{};
+                std::memcpy(&n, image.data() + 64 + (i * sizeof(n)), sizeof(n));
+                if (n.kind == nlohmann::detail::view::node_kind::string)
+                {
+                    n.next = 0;
+                }
+                std::memcpy(image.data() + 64 + (i * sizeof(n)), &n, sizeof(n));
+            }
+            const json_document d = json_document::load(image);
+            const json m = d.root().materialize();
+            CHECK(range(m[std::string("a") + '\n'][0]) == "none");
+            CHECK(range(m["c"]) == range(j["c"]));
+            CHECK(same_positions(m[std::string("a") + '\n'][2], j[std::string("a") + '\n'][2]));
+        }
+
+#if !defined(JSON_NOEXCEPTION)
+        SECTION("a decoded string's position must be within the text")
+        {
+            std::vector<std::uint8_t> image = json_document::parse(R"(["\n"])").save();
+            nlohmann::detail::view::node n{};
+            std::memcpy(&n, image.data() + 64 + sizeof(n), sizeof(n));
+            REQUIRE(n.next == 2);
+            n.next = 7;
+            std::memcpy(image.data() + 64 + sizeof(n), &n, sizeof(n));
+            CHECK_THROWS_AS(json_document::load(image), json::parse_error&);
+            // without the check, the position is out of bounds and ignored
+            const json_document d = json_document::load(image, json_document::image_check::none);
+            CHECK(range(d.root()[0].materialize()) == "none");
+        }
+#endif
+    }
+#endif
+}