item_json_func.cc File Reference

#include "sql/item_json_func.h"
#include <assert.h>
#include <algorithm>
#include <cstring>
#include <limits>
#include <memory>
#include <new>
#include <string>
#include <utility>
#include <vector>
#include "decimal.h"
#include "field_types.h"
#include "lex_string.h"
#include "my_alloc.h"
#include "my_dbug.h"
#include "my_sys.h"
#include "mysql/mysql_lex_string.h"
#include "mysql/strings/m_ctype.h"
#include "mysqld_error.h"
#include "prealloced_array.h"
#include "scope_guard.h"
#include "sql-common/json_diff.h"
#include "sql-common/json_dom.h"
#include "sql-common/json_path.h"
#include "sql-common/json_schema.h"
#include "sql-common/json_syntax_check.h"
#include "sql-common/my_decimal.h"
#include "sql/current_thd.h"
#include "sql/debug_sync.h"
#include "sql/error_handler.h"
#include "sql/field.h"
#include "sql/field_common_properties.h"
#include "sql/item_cmpfunc.h"
#include "sql/item_create.h"
#include "sql/parser_yystype.h"
#include "sql/psi_memory_key.h"
#include "sql/sql_class.h"
#include "sql/sql_const.h"
#include "sql/sql_error.h"
#include "sql/sql_exception_handler.h"
#include "sql/sql_time.h"
#include "sql/system_variables.h"
#include "sql/table.h"
#include "sql/table_function.h"
#include "sql/thd_raii.h"
#include "sql/thr_malloc.h"
#include "sql_string.h"
#include "string_with_len.h"
#include "template_utils.h"

Classes
struct	Item_func_json_value::Default_value

Typedefs
typedef Prealloced_array< size_t, 16 >	Sorted_index_array
typedef Prealloced_array< std::string, 16 >	String_set

Functions
bool	ensure_utf8mb4 (const String &val, String *buf, const char **resptr, size_t *reslength, bool require_string)
	Helper routines. More...
bool	parse_json (const String &res, Json_dom_ptr *dom, bool require_str_or_json, const JsonParseErrorHandler &error_handler, const JsonErrorHandler &depth_handler)
	Parse a JSON dom out of an argument to a JSON function. More...
enum_json_diff_status	apply_json_diffs (Field_json *field, const Json_diff_vector *diffs)
	Apply a sequence of JSON diffs to the value stored in a JSON column. More...
static enum_field_types	get_real_blob_type (const Field *arg)
	Get correct blob type of given Field. More...
static enum_field_types	get_real_blob_type (const Item *arg)
	Get correct blob type of given Item. More...
static enum_field_types	get_normalized_field_type (const Item *arg)
	Get the field type of an item. More...
bool	get_json_object_member_name (const THD *thd, Item *arg_item, String *value, String *utf8_res, const char **safep, size_t *safe_length)
	Gets a JSON object member name from an Item. More...
static bool	is_convertible_to_json (const Item *item)
	A helper method that checks whether or not the given argument can be converted to JSON. More...
static bool	check_convertible_to_json (const Item *item, int argument_number, const char *function_name)
	Checks if an Item is of a type that is convertible to JSON. More...
static bool	json_is_valid (Item **args, uint arg_idx, String *value, const char *func_name, Json_dom_ptr *dom, bool require_str_or_json, bool *valid)
	Helper method for Item_func_json_* methods. More...
bool	parse_path (const String &path_value, bool forbid_wildcards, Json_path *json_path)
	A helper function that uses the above one as workhorse. More...
static enum_one_or_all_type	parse_one_or_all (const String *candidate, const char *func_name)
	Parse a oneOrAll argument. More...
static enum_one_or_all_type	parse_and_cache_ooa (const THD *thd, Item *arg, enum_one_or_all_type *cached_ooa, const char *func_name)
	Parse and cache a (possibly constant) oneOrAll argument. More...
static bool	evaluate_constant_json_schema (THD *thd, Item *json_schema, Json_schema_validator *cached_schema_validator, Item **ref)
static bool	do_json_schema_validation (const THD *thd, Item *json_schema, Item *json_document, const char *func_name, const Json_schema_validator &cached_schema_validator, bool *null_value, bool *validation_result, Json_schema_validation_report *validation_report)
bool	json_value (Item *arg, Json_wrapper *result, bool *has_value)
	Return the JSON value of the argument in a wrapper. More...
bool	get_json_wrapper (Item **args, uint arg_idx, String *str, const char *func_name, Json_wrapper *wrapper)
	Return the JSON value of the argument in a wrapper. More...
static String *	val_string_from_json (Item_func *item, String *buffer)
static bool	get_date_from_json (Item_func *item, MYSQL_TIME *ltime, my_time_flags_t)
static bool	get_time_from_json (Item_func *item, MYSQL_TIME *ltime)
longlong	val_int_from_json (Item_func *item)
static double	val_real_from_json (Item_func *item)
static my_decimal *	val_decimal_from_json (Item_func *item, my_decimal *decimal_value)
template<typename T , typename... Args>
static bool	create_scalar (Json_scalar_holder *scalar, Json_dom_ptr *dom, Args &&...args)
	Create a new Json_scalar object, either in memory owned by a Json_scalar_holder object or on the heap. More...
bool	sql_scalar_to_json (Item *arg, const char *calling_function, String *value, String *tmp, Json_wrapper *wr, Json_scalar_holder *scalar, bool scalar_string)
	Get a JSON value from an SQL scalar value. More...
bool	get_json_atom_wrapper (Item **args, uint arg_idx, const char *calling_function, String *value, String *tmp, Json_wrapper *wr, Json_scalar_holder *scalar, bool accept_string)
	Convert Json values or MySQL values to JSON. More...
bool	get_atom_null_as_null (Item **args, uint arg_idx, const char *calling_function, String *value, String *tmp, Json_wrapper *wr)
	Convert JSON values or MySQL values to JSON. More...
static bool	possible_root_path (const Json_path_iterator &begin, const Json_path_iterator &end)
	Is this a path that could possibly return the root node of a JSON document? More...
static void	disable_logical_diffs (const Field_json *field)
static void	disable_binary_diffs (const Field_json *field)
static bool	find_matches (const Json_wrapper &wrapper, String *path, Json_dom_vector *matches, String_set *duplicates, bool one_match, Item *like_node, Item_string *source_string)
	Recursive function to find the string values, nested inside a json document, which satisfy the LIKE condition. More...
static void	set_data_type_from_cast_type (Item *item, Cast_target cast_type, unsigned length, unsigned decimals, const CHARSET_INFO *charset)
	Sets the data type of an Item_func_array_cast or Item_func_json_value based on the Cast_type. More...
static void	print_cast_type (Cast_target cast_type, const Item *item, String *str)
	Prints the target type of a cast operation (either CAST or JSON_VALUE). More...
static enum Item_result	json_cast_result_type (Cast_target cast_type)
static enum_field_types	data_type_to_real_type (enum_field_types data_type)
	Converts the "data type" used by Item to a "real type" used by Field. More...
static bool	can_store_json_value_unencoded (const Field *field_to_store_in, const Json_wrapper *json_data)
	Check if a JSON value is a JSON OPAQUE, and if it can be printed in the field as a non base64 value. More...
bool	save_json_to_field (THD *thd, Field *field, const Json_wrapper *w, bool no_error)
	Save JSON to a given field. More...
static bool	decimal_within_range (const Item *item, const my_decimal *decimal)
	Checks if a decimal value is within the range of the data type of an Item. More...
static bool	same_response_type (Json_on_response_type type1, Json_on_response_type type2)
	Checks if two Json_on_response_type values represent the same response. More...
static bool	handle_json_value_conversion_error (Json_on_response_type on_error, const char *type, Item_func_json_value *item)
	Handles conversion errors for JSON_VALUE according to the ON ERROR clause. More...

Typedef Documentation

Sorted_index_array

typedef Prealloced_array<size_t, 16> Sorted_index_array

String_set

typedef Prealloced_array<std::string, 16> String_set

Function Documentation

apply_json_diffs()

enum_json_diff_status apply_json_diffs	(	Field_json *	field,
		const Json_diff_vector *	diffs
	)

Apply a sequence of JSON diffs to the value stored in a JSON column.

Parameters

field	the column to update
diffs	the diffs to apply

Returns

an enum_json_diff_status value that tells if the diffs were applied successfully

can_store_json_value_unencoded()

static bool can_store_json_value_unencoded ( const Field *  field_to_store_in, const Json_wrapper *  json_data  )

static

Check if a JSON value is a JSON OPAQUE, and if it can be printed in the field as a non base64 value.

This is currently used by JSON_TABLE to see if we can print the JSON value in a field without having to encode it in base64.

Parameters

field_to_store_in	The field we want to store the JSON value in
json_data	The JSON value we want to store.

Returns

true The JSON value can be stored without encoding it in base64 false The JSON value can not be stored without encoding it, or it is not a JSON OPAQUE value.

check_convertible_to_json()

static bool check_convertible_to_json ( const Item *  item, int  argument_number, const char *  function_name  )

static

Checks if an Item is of a type that is convertible to JSON.

An error is raised if it is not convertible.

create_scalar()

template<typename T , typename... Args>

static bool create_scalar ( Json_scalar_holder *  scalar, Json_dom_ptr *  dom, Args &&...  args  )

static

Create a new Json_scalar object, either in memory owned by a Json_scalar_holder object or on the heap.

Parameters

[in,out]	scalar	the Json_scalar_holder in which to create the new Json_scalar, or nullptr if it should be created on the heap
[in,out]	dom	pointer to the Json_scalar if it's created on the heap
[in]	args	the arguments to pass to T's constructor

Template Parameters

T	the type of object to create; a subclass of Json_scalar
Args	type of the arguments to pass to T's constructor

Return values

false	if successful
true	if memory could not be allocated

data_type_to_real_type()

static enum_field_types data_type_to_real_type ( enum_field_types  data_type )

static

Converts the "data type" used by Item to a "real type" used by Field.

decimal_within_range()

static bool decimal_within_range ( const Item *  item, const my_decimal *  decimal  )

static

Checks if a decimal value is within the range of the data type of an Item.

It is considered within range if it can be converted to the data type without losing any leading significant digits.

disable_binary_diffs()

static void disable_binary_diffs ( const Field_json *  field )

static

disable_logical_diffs()

static void disable_logical_diffs ( const Field_json *  field )

static

do_json_schema_validation()

static bool do_json_schema_validation ( const THD *  thd, Item *  json_schema, Item *  json_document, const char *  func_name, const Json_schema_validator &  cached_schema_validator, bool *  null_value, bool *  validation_result, Json_schema_validation_report *  validation_report  )

static

ensure_utf8mb4()

bool ensure_utf8mb4	(	const String &	val,
		String *	buf,
		const char **	resptr,
		size_t *	reslength,
		bool	require_string
	)

Helper routines.

Check a non-empty val for character set.

evaluate_constant_json_schema()

static bool evaluate_constant_json_schema ( THD *  thd, Item *  json_schema, Json_schema_validator *  cached_schema_validator, Item **  ref  )

static

find_matches()

static bool find_matches ( const Json_wrapper &  wrapper, String *  path, Json_dom_vector *  matches, String_set *  duplicates, bool  one_match, Item *  like_node, Item_string *  source_string  )

static

Recursive function to find the string values, nested inside a json document, which satisfy the LIKE condition.

As matches are found, their path locations are added to an evolving vector of matches.

Parameters

[in]	wrapper	A subdocument of the original document.
[in]	path	The path location of the subdocument
[in,out]	matches	The evolving vector of matches.
[in,out]	duplicates	Sorted set of paths found already, which is used to avoid inserting duplicates into matches.
[in]	one_match	If true, then terminate search after first match.
[in]	like_node	The LIKE node that's evaluated on the string values.
[in]	source_string	The input string item of the LIKE node.

Return values

false	on success
true	on failure

get_atom_null_as_null()

bool get_atom_null_as_null	(	Item **	args,
		uint	arg_idx,
		const char *	calling_function,
		String *	value,
		String *	tmp,
		Json_wrapper *	wr
	)

Convert JSON values or MySQL values to JSON.

Converts SQL NULL to the JSON null literal.

Parameters

[in]	args	arguments to function
[in]	arg_idx	the index of the argument to process
[in]	calling_function	name of the calling function
[in,out]	value	working area (if the returned Json_wrapper points to a binary value rather than a DOM, this string will end up holding the binary representation, and it must stay alive until the wrapper is destroyed or converted from binary to DOM)
[in,out]	tmp	temporary scratch space for converting strings to the correct charset; only used if accept_string is true and conversion is needed
[in,out]	wr	the result wrapper

Returns

false if we found a value or NULL, true otherwise

get_date_from_json()

static bool get_date_from_json ( Item_func *  item, MYSQL_TIME *  ltime, my_time_flags_t    )

static

get_json_atom_wrapper()

bool get_json_atom_wrapper	(	Item **	args,
		uint	arg_idx,
		const char *	calling_function,
		String *	value,
		String *	tmp,
		Json_wrapper *	wr,
		Json_scalar_holder *	scalar,
		bool	accept_string
	)

Convert Json values or MySQL values to JSON.

Parameters

[in]	args	arguments to function
[in]	arg_idx	the index of the argument to process
[in]	calling_function	name of the calling function
[in,out]	value	working area (if the returned Json_wrapper points to a binary value rather than a DOM, this string will end up holding the binary representation, and it must stay alive until the wrapper is destroyed or converted from binary to DOM)
[in,out]	tmp	temporary scratch space for converting strings to the correct charset; only used if accept_string is true and conversion is needed
[in,out]	wr	the result wrapper
[in,out]	scalar	pointer to pre-allocated memory that can be borrowed by the result wrapper if the result is a scalar. If the pointer is NULL, memory for a scalar result will be allocated on the heap.
[in]	accept_string	if true, accept MySQL strings as JSON strings by converting them to UTF8, else emit an error

Returns

false if we found a value or NULL, true otherwise

get_json_object_member_name()

bool get_json_object_member_name	(	const THD *	thd,
		Item *	arg_item,
		String *	value,
		String *	utf8_res,
		const char **	safep,
		size_t *	safe_length
	)

Gets a JSON object member name from an Item.

An error is raised if the Item evaluates to NULL, or if it cannot be converted to a utf8mb4 string.

Parameters

[in]	thd	THD handle
[in]	arg_item	An argument Item
[out]	value	Where to materialize the arg_item's string value
[out]	utf8_res	Buffer for use by ensure_utf8mb4.
[out]	safep	String pointer after any relevant conversion
[out]	safe_length	Corresponding string length

Returns

true if the Item is not a utf8mb4 string

get_json_wrapper()

bool get_json_wrapper	(	Item **	args,
		uint	arg_idx,
		String *	str,
		const char *	func_name,
		Json_wrapper *	wrapper
	)

Return the JSON value of the argument in a wrapper.

Abstracts whether the value comes from a field or a function or a valid JSON text.

Parameters

[in]	args	the arguments
[in]	arg_idx	the argument index
[out]	str	the string buffer
[in]	func_name	the name of the function we are executing
[out]	wrapper	the JSON value wrapper

Returns

false if we found a value or NULL, true if not.

get_normalized_field_type()

static enum_field_types get_normalized_field_type ( const Item *  arg )

static

Get the field type of an item.

This function returns the same value as arg->data_type() in most cases, but in some cases it may return another field type in order to ensure that the item gets handled the same way as items of a different type.

get_real_blob_type() [1/2]

static enum_field_types get_real_blob_type ( const Field *  arg )

static

Get correct blob type of given Field.

A helper function for get_normalized_field_type().

Parameters

arg	the field to get blob type of

Returns

correct blob type

get_real_blob_type() [2/2]

static enum_field_types get_real_blob_type ( const Item *  arg )

static

Get correct blob type of given Item.

A helper function for get_normalized_field_type().

Parameters

arg	the item to get blob type of

Returns

correct blob type

get_time_from_json()

static bool get_time_from_json ( Item_func *  item, MYSQL_TIME *  ltime  )

static

handle_json_value_conversion_error()

static bool handle_json_value_conversion_error ( Json_on_response_type  on_error, const char *  type, Item_func_json_value *  item  )

static

Handles conversion errors for JSON_VALUE according to the ON ERROR clause.

Called when the conversion of the extracted JSON value cannot be converted to the target type without truncation or data loss.

If ERROR ON ERROR is specified, an error is raised, and true is returned.

If NULL ON ERROR is specified (explicitly or implicitly), the item's null_value is set to true, and false is returned.

If DEFAULT ... ON ERROR is specified, the item's null_value is set to false, and false is returned. It is up to the caller to return the correct default value.

Parameters

	on_error	the type of response to give to the error
	type	the data type returned by the JSON_VALUE expression
[in,out]	item	the Item representing the JSON_VALUE expression

Return values

true	for ERROR ON ERROR (my_error() is called before returning)
false	if DEFAULT .. ON ERROR or NULL ON ERROR was given

is_convertible_to_json()

static bool is_convertible_to_json ( const Item *  item )

static

A helper method that checks whether or not the given argument can be converted to JSON.

The function only checks the type of the given item, and doesn't do any parsing or further checking of the item.

Parameters

item	The item to be checked

Return values

true	The item is possibly convertible to JSON
false	The item is not convertible to JSON

json_cast_result_type()

static enum Item_result json_cast_result_type ( Cast_target  cast_type )

static

json_is_valid()

static bool json_is_valid ( Item **  args, uint  arg_idx, String *  value, const char *  func_name, Json_dom_ptr *  dom, bool  require_str_or_json, bool *  valid  )

static

Helper method for Item_func_json_* methods.

Check if a JSON item or JSON text is valid and, for the latter, optionally construct a DOM tree (i.e. only if valid).

Parameters

[in]	args	Item_func::args alias
[in]	arg_idx	Index (0-based) of argument into the args array
[out]	value	Alias for Item_func_json_*::m_value
[in]	func_name	Name of the user-invoked JSON_ function
[in,out]	dom	If non-null, we want any text parsed DOM returned at the location pointed to
[in]	require_str_or_json	If true, generate an error if other types used as input
[out]	valid	true if a valid JSON value was found (or NULL), else false

Returns

true iff syntax error and dom != null, else false

json_value()

bool json_value	(	Item *	arg,
		Json_wrapper *	result,
		bool *	has_value
	)

Return the JSON value of the argument in a wrapper.

Handles arguments with type JSON, including array objects (which do not report type JSON but rather the type of individual elements).

Does not handle literals. See also get_json_wrapper.

Parameters

[in]	arg	the argument
[in,out]	result	the JSON value wrapper
[out]	has_value	true if argument was handled, false otherwise undefined when error

parse_and_cache_ooa()

static enum_one_or_all_type parse_and_cache_ooa ( const THD *  thd, Item *  arg, enum_one_or_all_type *  cached_ooa, const char *  func_name  )

static

Parse and cache a (possibly constant) oneOrAll argument.

Parameters

[in]	thd	THD handle.
[in]	arg	The oneOrAll arg passed to the JSON function.
[in]	cached_ooa	Previous result of parsing this arg.
[in]	func_name	The name of the calling JSON function.

Returns

ooa_one, ooa_all, ooa_null or ooa_error, based on the match

parse_json()

bool parse_json	(	const String &	res,
		Json_dom_ptr *	dom,
		bool	require_str_or_json,
		const JsonParseErrorHandler &	error_handler,
		const JsonErrorHandler &	depth_handler
	)

Parse a JSON dom out of an argument to a JSON function.

Parameters

[in]	res	Pointer to string value of arg.
[in,out]	dom	If non-null, we want any text parsed DOM returned at the location pointed to
[in]	require_str_or_json	If true, generate an error if other types used as input
[in]	error_handler	Pointer to a function that should handle reporting of parsing error.
[in]	depth_handler	Pointer to a function that should handle error occurred when depth is exceeded.

Returns

false if the arg parsed as valid JSON, true otherwise

parse_one_or_all()

static enum_one_or_all_type parse_one_or_all ( const String *  candidate, const char *  func_name  )

static

Parse a oneOrAll argument.

Parameters

[in]	candidate	The string to compare to "one" or "all"
[in]	func_name	The name of the calling function

Returns

ooa_one, ooa_all, or ooa_error, based on the match

parse_path()

bool parse_path	(	const String &	path_value,
		bool	forbid_wildcards,
		Json_path *	json_path
	)

A helper function that uses the above one as workhorse.

Entry point for for JSON_TABLE (Table_function_json class) and Json_path_cache. Raises an error if the path expression is syntactically incorrect. Raises an error if the path expression contains wildcard tokens but is not supposed to. Otherwise updates the supplied Json_path object with the parsed path.

Parameters

[in]	path_value	A String to be interpreted as a path.
[in]	forbid_wildcards	True if the path shouldn't contain * or **
[out]	json_path	The object that will hold the parsed path

Returns

false on success (valid path or NULL), true on error

possible_root_path()

static bool possible_root_path ( const Json_path_iterator &  begin, const Json_path_iterator &  end  )

static

Is this a path that could possibly return the root node of a JSON document?

A path that returns the root node must be on one of the following forms:

• the root ('$'), or
• a sequence of array cells at index 0 or last that any non-array element at the top level could have been autowrapped to, i.e. '$[0]' or '$[0][0]...[0]'.

See also

Json_path_leg::is_autowrap

Parameters

begin	the beginning of the path
end	the end of the path (exclusive)

Returns

true if the path may match the root, false otherwise

print_cast_type()

static void print_cast_type ( Cast_target  cast_type, const Item *  item, String *  str  )

static

Prints the target type of a cast operation (either CAST or JSON_VALUE).

Parameters

	cast_type	the cast type
	item	the Item in which the cast operation is performed
[out]	str	the string to print to

same_response_type()

static bool same_response_type ( Json_on_response_type  type1, Json_on_response_type  type2  )

static

Checks if two Json_on_response_type values represent the same response.

Implicit responses are equal to NULL ON EMPTY/ERROR.

save_json_to_field()

bool save_json_to_field	(	THD *	thd,
		Field *	field,
		const Json_wrapper *	w,
		bool	no_error
	)

Save JSON to a given field.

Value is saved in type-aware manner. Into a JSON-typed column any JSON data could be saved. Into an SQL scalar field only a scalar could be saved. If data being saved isn't scalar or can't be coerced to the target type, an error is returned.

Parameters

thd	Thread handler
field	Field to save data to
w	JSON data to save
no_error	If true, don't raise an error when the value cannot be converted to the target type

Returns

false ok true coercion error occur

set_data_type_from_cast_type()

static void set_data_type_from_cast_type ( Item *  item, Cast_target  cast_type, unsigned  length, unsigned  decimals, const CHARSET_INFO *  charset  )

static

Sets the data type of an Item_func_array_cast or Item_func_json_value based on the Cast_type.

Parameters

item	the Item whose data type to set
cast_type	the type of cast
length	the declared length of the target type
decimals	the declared precision of the target type
charset	the character set of the target type (nullptr if not specified)

sql_scalar_to_json()

bool sql_scalar_to_json	(	Item *	arg,
		const char *	calling_function,
		String *	value,
		String *	tmp,
		Json_wrapper *	wr,
		Json_scalar_holder *	scalar,
		bool	scalar_string
	)

Get a JSON value from an SQL scalar value.

Parameters

[in]	arg	the function argument
[in]	calling_function	the name of the calling function
[in,out]	value	a scratch area
[in,out]	tmp	temporary scratch space for converting strings to the correct charset; only used if accept_string is true and conversion is needed
[out]	wr	the retrieved JSON value
[in,out]	scalar	pointer to pre-allocated memory that can be borrowed by the result wrapper to hold the scalar result. If the pointer is NULL, memory will be allocated on the heap.
[in]	scalar_string	if true, interpret SQL strings as scalar JSON strings. if false, interpret SQL strings as JSON objects. If conversion fails, return the string as a scalar JSON string instead.

Returns

false if we could get a value or NULL, otherwise true

Scalar processing is irrelevant. Geometry types are converted to JSON objects.

val_decimal_from_json()

static my_decimal * val_decimal_from_json ( Item_func *  item, my_decimal *  decimal_value  )

static

val_int_from_json()

longlong val_int_from_json

(

Item_func *

item

)

val_real_from_json()

static double val_real_from_json ( Item_func *  item )

static

val_string_from_json()

static String * val_string_from_json ( Item_func *  item, String *  buffer  )

static