json_schema.cc File Reference

#include "sql/json_schema.h"
#include <my_rapidjson_size_t.h>
#include <assert.h>
#include <rapidjson/document.h>
#include <rapidjson/error/error.h>
#include <rapidjson/memorystream.h>
#include <rapidjson/pointer.h>
#include <rapidjson/reader.h>
#include <rapidjson/schema.h>
#include <rapidjson/stringbuffer.h>
#include <algorithm>
#include <cassert>
#include <cstdint>
#include <string>
#include <unordered_set>
#include <utility>
#include "my_alloc.h"
#include "my_inttypes.h"
#include "my_sys.h"
#include "mysqld_error.h"
#include "sql-common/json_syntax_check.h"
#include "sql/sql_exception_handler.h"

Functions
static bool	is_local_json_pointer_ref (const rapidjson::Value &ref_value)
	Check whether a $ref value is a local JSON Pointer reference. More...
static const rapidjson::Value *	resolve_local_json_pointer_ref (const rapidjson::Document &schema_document, const rapidjson::Value &ref_value)
	Resolve a local JSON Pointer $ref against the schema document. More...
static bool	json_schema_has_unsafe_expansion (const rapidjson::Document &schema_document, const rapidjson::Value *node, uint32_t depth_so_far, uint32_t limit, std::unordered_set< const rapidjson::Value * > *visiting)
	Check whether any path from a schema node is unsafe for rapidjson schema compilation. More...
static bool	check_json_schema_expansion_depth (const rapidjson::Document &schema_document)
	Validate that a JSON Schema can be safely compiled by rapidjson. More...
static bool	parse_json_schema (const char *json_schema_str, size_t json_schema_length, const char *function_name, rapidjson::Document *schema_document)
	parse_json_schema will parse a JSON input into a JSON Schema. More...
bool	is_valid_json_schema (const char *document_str, size_t document_length, const char *json_schema_str, size_t json_schema_length, const char *function_name, bool *is_valid, Json_schema_validation_report *validation_report)
	This function will validate a JSON document against a JSON Schema using the validation provided by rapidjson. More...
unique_ptr_destroy_only< const Json_schema_validator >	create_json_schema_validator (MEM_ROOT *mem_root, const char *json_schema_str, size_t json_schema_length, const char *function_name)
	Create a Json_schema_validator, allocated on a given MEM_ROOT. More...

Variables
static constexpr uint32_t	JSON_SCHEMA_MAX_EXPANSION_DEPTH = 100
	Maximum "expansion depth" of a JSON Schema during rapidjson schema compilation. More...

Function Documentation

check_json_schema_expansion_depth()

static bool check_json_schema_expansion_depth ( const rapidjson::Document &  schema_document )

static

Validate that a JSON Schema can be safely compiled by rapidjson.

rapidjson schema compilation expands local $ref by recursively compiling the referenced subschemas. The recursion depth is therefore related to the number of local reference "hops" along the deepest expansion path. This depth can be large even if no referred object has a $ref at its own top level, as long as each referred object contains a nested object with a $ref.

This function checks whether any expansion path exceeds the limit or contains a local $ref cycle. Each step into a nested object/array and each local $ref expansion counts as 1.

It does this by walking JSON-tree edges (object members / array elements) and local $ref edges (JSON pointer references starting with #), with early exit as soon as the limit is exceeded.

If the limit is exceeded (JSON_SCHEMA_MAX_EXPANSION_DEPTH) or a cycle is detected, ER_JSON_DOCUMENT_TOO_DEEP is reported.

Parameters

schema_document

Parsed JSON schema document.

Return values

true	The schema is unsafe and an error was reported.
false	No unsafe expansion path was found.

create_json_schema_validator()

unique_ptr_destroy_only< const Json_schema_validator > create_json_schema_validator	(	MEM_ROOT *	mem_root,
		const char *	json_schema_str,
		size_t	json_schema_length,
		const char *	function_name
	)

Create a Json_schema_validator, allocated on a given MEM_ROOT.

Parameters

mem_root	The MEM_ROOT to allocate the validator on
json_schema_str	A pointer to the JSON Schema
json_schema_length	The length of the JSON Schema input
function_name	The function name of the caller (to be used in error reporting)

Return values

nullptr

on error (my_error has been called)

is_local_json_pointer_ref()

static bool is_local_json_pointer_ref ( const rapidjson::Value &  ref_value )

static

Check whether a $ref value is a local JSON Pointer reference.

We only guard references of the form "#" or "#/...".

Parameters

ref_value

The value of the $ref member.

Return values

true	The $ref is a local JSON pointer reference.
false	Otherwise.

is_valid_json_schema()

bool is_valid_json_schema	(	const char *	document_str,
		size_t	document_length,
		const char *	json_schema_str,
		size_t	json_schema_length,
		const char *	function_name,
		bool *	is_valid,
		Json_schema_validation_report *	report
	)

This function will validate a JSON document against a JSON Schema using the validation provided by rapidjson.

Parameters

	document_str	A pointer to the JSON document to be validated.
	document_length	The length of the JSON document to be validated.
	json_schema_str	A pointer to the JSON Schema.
	json_schema_length	The length of the JSON Schema.
	function_name	The name of the SQL function calling this function. Used in error reporting.
[out]	is_valid	A variable containing the result of the validation. If true, the JSON document is valid according to the given JSON Schema.
[out]	report	A structure containing a detailed report from the validation. Is only populated if is_valid is set to "false". Can be nullptr if a detailed report isn't needed.

Return values

true	if anything went wrong (like parsing the JSON inputs). my_error has been called with an appropriate error message.
false	if the validation succeeded. The result of the validation can be found in the output variable "is_valid".

json_schema_has_unsafe_expansion()

static bool json_schema_has_unsafe_expansion ( const rapidjson::Document &  schema_document, const rapidjson::Value *  node, uint32_t  depth_so_far, uint32_t  limit, std::unordered_set< const rapidjson::Value * > *  visiting  )

static

Check whether any path from a schema node is unsafe for rapidjson schema compilation.

The expansion graph consists of:

• JSON-tree edges (object members / array elements), and
• local $ref edges (local JSON Pointer references).

A path is unsafe if it exceeds the maximum expansion depth, or if it contains a cycle through local $ref edges. Cyclic local references are rejected before constructing rapidjson::SchemaDocument because rapidjson can recurse while compiling such schemas.

Parameters

	schema_document	Parsed JSON schema document (used for resolving $ref).
	node	The node to start the search from.
	depth_so_far	Current depth from the schema root.
	limit	Maximum allowed expansion depth.
[in,out]	visiting	Nodes currently on the expansion path.

Return values

true	The limit is exceeded along some path, or a cycle was detected.
false	No unsafe path was found.

parse_json_schema()

static bool parse_json_schema ( const char *  json_schema_str, size_t  json_schema_length, const char *  function_name, rapidjson::Document *  schema_document  )

static

parse_json_schema will parse a JSON input into a JSON Schema.

If the input isn't a valid JSON, or if the JSON is too deeply nested, an error will be returned to the user.

Parameters

	json_schema_str	A pointer to the JSON Schema input
	json_schema_length	The length of the JSON Schema input
	function_name	The function name of the caller (to be used in error reporting)
[out]	schema_document	An object where the JSON Schema will be put. This variable MUST be initialized.

Return values

true	on error (my_error has been called)
false	on success. The JSON Schema can be found in the output parameter schema_document.

resolve_local_json_pointer_ref()

static const rapidjson::Value * resolve_local_json_pointer_ref ( const rapidjson::Document &  schema_document, const rapidjson::Value &  ref_value  )

static

Resolve a local JSON Pointer $ref against the schema document.

Parameters

schema_document	The full JSON schema document.
ref_value	The $ref member value (must be a local JSON pointer).

Returns

Pointer to the referenced schema node on success, otherwise nullptr.

Variable Documentation

JSON_SCHEMA_MAX_EXPANSION_DEPTH

constexpr uint32_t JSON_SCHEMA_MAX_EXPANSION_DEPTH = 100

staticconstexpr

Maximum "expansion depth" of a JSON Schema during rapidjson schema compilation.

rapidjson schema compilation is recursive. The call stack depth is driven by a combination of:

• JSON nesting within the schema (object members / array elements), and
• local $ref expansion (compiling referenced subschemas recursively).

This cap bounds the maximum number of edges along any path in the schema's "expansion graph", where both JSON-tree edges and local $ref edges count as 1.

Since MySQL rejects JSON documents deeper than 100 levels (see sql-common/json_syntax_check.cc), schemas requiring more than 100 levels of expansion cannot describe JSON documents supported by MySQL. We therefore use 100 as the single guard limit here as well.