Hash deeply nested values without recursing per nesting level

std::hash<basic_json> hashed an array or object by hashing each element,
which called detail::hash again once per nesting level. A value nested
deeply enough - 50,000 levels of objects on an 8 MiB stack - exhausted
the call stack and terminated the process. parse() accepts such values
without complaint, since the parser is iterative, and a parsed value is
hashed wherever it is used as a key in an unordered container.

Bound the descent the same way dump() does: detail::hash takes the
nesting level, and once hash_depth_limit() (128) levels have been entered,
hash_iteratively() hashes what is left on an explicit stack. It combines
the seeds in exactly the same order, so hash values are unchanged. A value
nested less deeply than the bound is hashed by the same code as before,
without allocating, and is as fast as before.

Tests check that every depth up to twice the bound hashes exactly like
the recursive definition of the hash, and that values nested 100,000
levels deep hash without crashing.

Fixes #5545 for std::hash.

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

tests/src/unit-hash.cpp

@@ -13,6 +13,78 @@ using json = nlohmann::json;
 using ordered_json = nlohmann::ordered_json;
 
 #include <set>
+#include <string>
+
+namespace
+{
+// how detail::hash defines the hash of an array or object: the seeds of the
+// elements, combined in order. Recursive, so only usable on values nested a
+// few hundred levels deep - which is exactly what is needed to check that the
+// iterative path taken below detail::hash_depth_limit() computes the same.
+template<typename BasicJsonType>
+std::size_t reference_hash(const BasicJsonType& j)
+{
+    using nlohmann::detail::combine;
+    using string_t = typename BasicJsonType::string_t;
+
+    if (!j.is_structured())
+    {
+        return std::hash<BasicJsonType> {}(j);
+    }
+
+    auto seed = combine(static_cast<std::size_t>(j.type()), j.size());
+    for (const auto& element : j.items())
+    {
+        if (j.is_object())
+        {
+            seed = combine(seed, std::hash<string_t> {}(element.key()));
+        }
+        seed = combine(seed, reference_hash(element.value()));
+    }
+    return seed;
+}
+
+// a value nested `depth` levels deep, with siblings on every level
+template<typename BasicJsonType>
+BasicJsonType nested(const std::size_t depth, const bool objects)
+{
+    BasicJsonType value = "leaf";
+    for (std::size_t i = 0; i < depth; ++i)
+    {
+        if (objects)
+        {
+            value = BasicJsonType{{"before", i}, {"nested", std::move(value)}, {"after", {i, "x"}}};
+        }
+        else
+        {
+            value = BasicJsonType::array({i, std::move(value), BasicJsonType::object({{"k", i}})});
+        }
+    }
+    return value;
+}
+
+std::string nested_text(const std::size_t depth, const bool objects)
+{
+    std::string text;
+    if (objects)
+    {
+        text.reserve(6 * depth + 1);
+        for (std::size_t i = 0; i < depth; ++i)
+        {
+            text += "{\"a\":";
+        }
+        text += "1";
+        text.append(depth, '}');
+    }
+    else
+    {
+        text.assign(depth, '[');
+        text += "1";
+        text.append(depth, ']');
+    }
+    return text;
+}
+} // namespace
 
 TEST_CASE("hash<nlohmann::json>")
 {
@@ -111,3 +183,44 @@ TEST_CASE("hash<nlohmann::ordered_json>")
 
     CHECK(hashes.size() == 21);
 }
+
+TEST_CASE("hash of deeply nested values")
+{
+    SECTION("hashing past the descent bound computes the same values")
+    {
+        // every depth on either side of where the iterative path takes over
+        for (std::size_t depth = 0; depth <= 2 * nlohmann::detail::hash_depth_limit() + 10; ++depth)
+        {
+            CAPTURE(depth);
+            const auto arrays = nested<json>(depth, false);
+            const auto objects = nested<json>(depth, true);
+            const auto ordered = nested<ordered_json>(depth, true);
+            CHECK(std::hash<json> {}(arrays) == reference_hash(arrays));
+            CHECK(std::hash<json> {}(objects) == reference_hash(objects));
+            CHECK(std::hash<ordered_json> {}(ordered) == reference_hash(ordered));
+        }
+    }
+
+    SECTION("values nested too deeply for the call stack (#5545)")
+    {
+        // recursing once per level used to exhaust the call stack here; the
+        // values are only parsed and hashed, never copied or compared, since
+        // those recurse as well
+        const std::size_t depth = 100000;
+        for (const bool objects :
+                {
+                    false, true
+                })
+        {
+            CAPTURE(objects);
+            const auto text = nested_text(depth, objects);
+            const auto a = json::parse(text);
+            const auto b = json::parse(text);
+            CHECK(std::hash<json> {}(a) == std::hash<json> {}(b));
+
+            const auto c = ordered_json::parse(text);
+            const auto d = ordered_json::parse(text);
+            CHECK(std::hash<ordered_json> {}(c) == std::hash<ordered_json> {}(d));
+        }
+    }
+}