Implementation:Alibaba MNN RapidJSON Schema

Knowledge Sources	Alibaba_MNN RapidJSON
Domains	JSON_Parsing, Schema_Validation
Last Updated	2026-02-10 12:00 GMT

Overview

3rd_party/rapidjson/schema.h (2496 lines) implements JSON Schema validation (Draft v4) for the vendored RapidJSON library within Alibaba MNN. It provides compile-time schema document construction and runtime validation of JSON values against a schema. The validator follows the SAX handler pattern, enabling both DOM-based and streaming validation.

Usage note: Vendored dependency used internally by MNN for JSON configuration parsing (model configs, LLM configs). Not directly imported by end users.

Key Classes

GenericSchemaDocument

Compiles a JSON document into an internal schema representation:

template <typename ValueT, typename Allocator = CrtAllocator>
class GenericSchemaDocument {
public:
    typedef ValueT ValueType;
    typedef IGenericRemoteSchemaDocumentProvider<GenericSchemaDocument> IRemoteSchemaDocumentProviderType;

    GenericSchemaDocument(const ValueType& document,
                          IRemoteSchemaDocumentProviderType* remoteProvider = 0,
                          Allocator* allocator = 0);
    ~GenericSchemaDocument();

    const SchemaType& GetRoot() const;
};

GenericSchemaValidator

Validates JSON data against a compiled schema. It implements the SAX Handler interface, so it can validate during parsing or accept events from DOM traversal:

template <typename SchemaDocumentType,
          typename OutputHandler = BaseReaderHandler<typename SchemaDocumentType::SchemaType::EncodingType>,
          typename StateAllocator = CrtAllocator>
class GenericSchemaValidator :
    public internal::ISchemaStateFactory<typename SchemaDocumentType::SchemaType>,
    public internal::ISchemaValidator,
    public internal::IValidationErrorHandler<typename SchemaDocumentType::SchemaType>
{
public:
    GenericSchemaValidator(const SchemaDocumentType& schemaDocument,
                           StateAllocator* allocator = 0,
                           size_t schemaStackCapacity = kDefaultSchemaStackCapacity,
                           size_t documentStackCapacity = kDefaultDocumentStackCapacity);

    // SAX handler methods
    bool Null();
    bool Bool(bool b);
    bool Int(int i);
    bool Uint(unsigned u);
    bool Int64(int64_t i);
    bool Uint64(uint64_t u);
    bool Double(double d);
    bool String(const Ch* str, SizeType length, bool copy);
    bool StartObject();
    bool Key(const Ch* str, SizeType length, bool copy);
    bool EndObject(SizeType memberCount);
    bool StartArray();
    bool EndArray(SizeType elementCount);

    // Validation result
    bool IsValid() const;
    void Reset();
    StringRefType GetInvalidSchemaKeyword() const;
};

The default typedef is:

typedef GenericSchemaDocument<Value> SchemaDocument;
typedef GenericSchemaValidator<SchemaDocument> SchemaValidator;

Supported Schema Keywords

The validator supports the following JSON Schema Draft v4 keywords:

Category	Keywords
Type validation	type, enum
Numeric constraints	multipleOf, minimum, maximum, exclusiveMinimum, exclusiveMaximum
String constraints	minLength, maxLength, pattern
Array constraints	items, additionalItems, minItems, maxItems, uniqueItems
Object constraints	properties, additionalProperties, required, minProperties, maxProperties, patternProperties, dependencies
Composition	allOf, anyOf, oneOf, not
References	$ref (with IRemoteSchemaDocumentProvider for remote references)

Regex Support

The pattern keyword requires regex matching. The header automatically selects an engine:

1. Internal regex (default) -- uses internal/regex.h when RAPIDJSON_SCHEMA_USE_INTERNALREGEX is enabled
2. std::regex -- uses <regex> when RAPIDJSON_SCHEMA_USE_STDREGEX is enabled and C++11 is available

#if !defined(RAPIDJSON_SCHEMA_USE_INTERNALREGEX)
#define RAPIDJSON_SCHEMA_USE_INTERNALREGEX 1
#else
#define RAPIDJSON_SCHEMA_USE_INTERNALREGEX 0
#endif

Dependencies

• document.h -- DOM value types for schema document input
• pointer.h -- JSON Pointer for schema $ref resolution and error path reporting
• stringbuffer.h -- String buffer for internal operations
• internal/regex.h -- Lightweight regex engine (when using internal regex)
• <cmath> -- abs(), floor() for numeric validation

License

MIT License. Copyright (C) 2015 THL A29 Limited (Tencent) and Milo Yip.

See Also

• Alibaba_MNN_RapidJSON_Document -- DOM document validated by schema
• Alibaba_MNN_RapidJSON_Pointer -- JSON Pointer used for $ref resolution
• Alibaba_MNN_RapidJSON_Regex -- Internal regex engine for pattern validation
• Alibaba_MNN_RapidJSON_Reader -- SAX reader that can validate during parsing