WL#16081 - Native Vector Embeddings Support In HeatWave

BUG#36165262 - WL#16081: Table is allowed to be partitioned on vector column
BUG#36167088 - WL#16081: Generated columns allowed on vector columns
BUG#36168511 - WL#16081: Issues with Vector column constraints
BUG#36195637 - Wl16081: Alter table not giving error when new dimension less than existing data
BUG#36194832 - WL#16081: STRING_TO_VECTOR function requires a vector type column
BUG#36168535 - [Wl16081] Select Hangs whe VECTOR_TO_STRING called on vector column
BUG#36206068 - [Wl16081] : Read-Replica broken with Vector data type
BUG#36214076 - WL#16081: Rpdserver crash - sig11 at from_string_to_vector
BUG#36225693 - WL#16081: Functions SHA1, MD5, SHA2 return error for vector data loaded to rapid
BUG#36241312 - WL#16081: Error All plans were rejected by HeatWave secondary engine
BUG#36255628 - WL#16081:setting secondary_engine of table having vector column taking long time
BUG#36265079 - WL#16081: wrong result when IS NULL applied to distance() function output
BUG#36272178 - WL#16081:virtual bool Item_func_get_user_var::propagate_type(THD*, const Type_properties&): Assertion `false' failed.
BUG#36239717 - to_base64() on vector column of miracl dataset crashes rapid
BUG#36255777 - WL#16081: Mysqld crash - Assertion `!std::isnan(nr)' failed.
BUG#36281463 - STRING_TO_VECTOR() returning error "Data cannot be converted"
BUG#36285521 - WL#16081: Mysqld crash - Assertion `!thd->is_error()' failed
BUG#36287504 - WL#16081: mysqld crash at Item_func_to_vector::val_str for ASAN
BUG#36267410 - WL#16081: mysqld crash at ParseBlob () in change_prop/rpd_binlog_parser.cc

This worklog will implement vector support in MySQL HeatWave.

- VECTOR column type: This will be an addition to the CREATE TABLE
statement as a new column data_type. Under the hood, it is mainly a
syntactic change; the VECTOR columns will be a wrapper around BLOB.
All the limitations that apply to BLOB apply for VECTOR.
More restrictions for VECTOR will also be in place, discussed below.

- DISTANCE function: This will be implemented as a component/UDF.
The function is responsible for computing the distance between two
VECTOR entries, of exactly the same dimension. Exact semantics will be
discussed as part of functional requirements.

- UTILITY functions:
-- VECTOR_DIM: Returns the dimensionality of each vector entry.
-- VECTOR_TO_STRING/FROM_VECTOR: Converts vector to human readable format.
-- STRING_TO_VECTOR/TO_VECTOR: Converts human readable format to vector.

Detailed Change Description
===========================
** VECTOR type support **
- sql/lex.h: The support for VECTOR keyword, along with VECTOR_SYM.
- sql/sql_yacc.yy: Change in directing VECTOR_SYM to PT_vector_type.
    - VECTOR(N) -> N is **optional** field_length.
    - If N is not provided, it will be assumed 2048 by default.
- sql/parse_tree_column_attrs.h: PT_vector_type, inherits PT_char_type.
    - It has additional vector_length to store given length.
    - The vector_length is computed by multiplying the given length
    value N by sizeof(float), since the entry precision is always
    a single-precision floating-point value.
    - It overrides is_vector to return true.
- sql/field.h:
    - Field_vector inherits from Field_blob
    - As its real_type, it will return MYSQL_TYPE_VECTOR: a newly
    introduced field type.
- sql/dd/types/column.h:
    - New virtual is_vector() function.
- sql/dd/impl/types/column_impl.h:
    - Column_impl overrides is_vector() -> It will return if its
    m_column_type_utf8 has "vector" string in its beginning.

** Restrictions on VECTOR type **
- share/messages_to_clients.txt: Introduced following errors:
    - ER_VECTOR_USED_AS_KEY
    - ER_UNABLE_TO_BUILD_HISTOGRAM_VECTOR
- sql/sql_table.cc:
    - At prepare_key_column, if a VECTOR typed column is found,
    throw ER_VECTOR_USED_AS_KEY. This ensures the restriction on
    VECTOR tyoe as PRIMARY KEY, FOREIGN KEY, UNIQUE, etc.
- sql/histograms/histogram.cc: If a histogram is being build on VECTOR
typed column, throw ER_UNABLE_TO_BUILD_HISTOGRAM_VECTOR.

** DISTANCE function support **
- new component: vector
- vector.cc:
    - Implements component_vector
    - Implements DISTANCE(arg_0, arg_1, <optional>distance_metric)
        - It will calculate the distance of each row at arg_0 and arg_1
        - If the length of either argument is not matching (evaluated
	for each row), the output for that row will be NULL.
        - The output precision of DISTANCE is double-precision (8 Bytes)
- vector.hpp: Standalone vectorized implementation of vector functions.

** Utility functions support **
- VECTOR_DIM
- VECTOR_TO_STRING
- STRING_TO_VECTOR
Functional requirements as described in the WL page.

** Restrictions on DISTANCE function **
- Expects 2 or 3 arguments.
- All arguments must be of type STRING_RESULT.
- If there is a 3rd argument, it must be one of these:
    - "DOT": dot/inner product
    - "COSINE": cosine distance
    - "EUCLIDIAN": euclidian distance
- If there is no 3rd argument, "DOT" will be used by default.

** HeatWave side support **
- VECTOR type:
    - For HeatWave, a vector("N") typed column is regarded as
    BLOB("sizeof(float)*N"), column.
    - One difference: For VECTOR typed columns, compression at load will
    be disabled by default.
- DISTANCE functions:
    - These will be interpreted as other Item_funcs. Same restrictions
    as mysql side apply.
    - Additionally, we can check if there is at least one vector typed
    column in the Item_func arguments.
    - The new primitive introduced uses the same distance function
    implementations as in vector.hpp

** 36165262 **
Vector column as partitioning key is blocked.

** 36167088 **
Vector column as part of generated column expression is blocked.

** 36168511 **
Addressed HeatWave load constraints: DICT encoding, ZONEMAP

** 36168535 **
This was a QCOMP bug at rapid: Mismatch between vbsize setting
and the consideration of the nullbv for chunkv size setting.
No need to consider nullbv for VARLEN columns.

** 36206068 **
In rpl_utility, enable conversion between BLOB and VECTOR

** 36214076 **
Handle zero length strings at STRING_TO_VECTOR

** 36225693 **
This was an issue on the trunk (charset handling in QKRN)

** 36241312 **
QKRN stats being set conservatively for conversion functions.

** 36255628 **
Field_vector was using is_equal of parent Field_blob, which
was leading to unnecessary not equal behavior. Overridden.

** 36265079 **
Issue was in rapid primitive: Instead of comparing dims, we
were comparing varlen max lengths.

** 36272178 **
Handle session var type propagation.

** 36239717 **
This is a bug on trunk, related to QKRN stats setting for
TO_BASE64 at QKRN.

** 36255777 **
Handle NaN values in distance calculation

** 36281463 **
reset errno before starting the conversion

** 36285521 **
VECTOR type check at Item_func_equal::resolve_type

** 36287504 **
Limit the number of chars printed at ER_TO_VECTOR_CONVERSION

** 36267410 **
Vector fields are marked as memcpy-able, which is not correct.
This was causing a binlog corruption.

Change-Id: I68d562193acd0cb00823908b0398f8215345521f

Author: kaankara

client/mysql.cc

@@ -1161,6 +1161,11 @@ static COMMANDS commands[] = {
     {"TO_DAYS", 0, nullptr, false, ""},
     {"TOUCHES", 0, nullptr, false, ""},
     {"TRIM", 0, nullptr, false, ""},
+    {"TO_VECTOR", 0, nullptr, false, ""},
+    {"STRING_TO_VECTOR", 0, nullptr, false, ""},
+    {"FROM_VECTOR", 0, nullptr, false, ""},
+    {"VECTOR_TO_STRING", 0, nullptr, false, ""},
+    {"VECTOR_DIM", 0, nullptr, false, ""},
     {"UCASE", 0, nullptr, false, ""},
     {"UNCOMPRESS", 0, nullptr, false, ""},
     {"UNCOMPRESSED_LENGTH", 0, nullptr, false, ""},
@@ -3909,15 +3914,15 @@ static void print_field_types(MYSQL_RES *result) {
 /* Used to determine if we should invoke print_as_hex for this field */
 
 static bool is_binary_field(MYSQL_FIELD *field) {
-  return ((field->charsetnr == 63) &&
-          (field->type == MYSQL_TYPE_BIT || field->type == MYSQL_TYPE_BLOB ||
-           field->type == MYSQL_TYPE_LONG_BLOB ||
-           field->type == MYSQL_TYPE_MEDIUM_BLOB ||
-           field->type == MYSQL_TYPE_TINY_BLOB ||
-           field->type == MYSQL_TYPE_VAR_STRING ||
-           field->type == MYSQL_TYPE_STRING ||
-           field->type == MYSQL_TYPE_VARCHAR ||
-           field->type == MYSQL_TYPE_GEOMETRY));
+  return (
+      (field->charsetnr == 63) &&
+      (field->type == MYSQL_TYPE_BIT || field->type == MYSQL_TYPE_BLOB ||
+       field->type == MYSQL_TYPE_LONG_BLOB ||
+       field->type == MYSQL_TYPE_MEDIUM_BLOB ||
+       field->type == MYSQL_TYPE_TINY_BLOB ||
+       field->type == MYSQL_TYPE_VAR_STRING ||
+       field->type == MYSQL_TYPE_STRING || field->type == MYSQL_TYPE_VARCHAR ||
+       field->type == MYSQL_TYPE_VECTOR || field->type == MYSQL_TYPE_GEOMETRY));
 }
 
 /* Print binary value as hex literal (0x ...) */

client/mysqldump.cc

@@ -4084,6 +4084,7 @@ static void dump_table(char *table, char *db) {
                                         field->type == MYSQL_TYPE_VAR_STRING ||
                                         field->type == MYSQL_TYPE_VARCHAR ||
                                         field->type == MYSQL_TYPE_BLOB ||
+                                        field->type == MYSQL_TYPE_VECTOR ||
                                         field->type == MYSQL_TYPE_LONG_BLOB ||
                                         field->type == MYSQL_TYPE_MEDIUM_BLOB ||
                                         field->type == MYSQL_TYPE_TINY_BLOB ||

client/mysqltest.cc

@@ -8287,6 +8287,30 @@ static void append_field(DYNAMIC_STRING *ds, uint col_idx, MYSQL_FIELD *field,
   }
 #endif
 
+  const size_t temp_val_max_width = (1 << 14);
+  char temp_val[temp_val_max_width];
+  DYNAMIC_STRING ds_temp = {.str = nullptr, .length = 0, .max_length = 0};
+  if (field->type == MYSQL_TYPE_VECTOR && !is_null) {
+    /* Do a binary to hex conversion for vector type */
+    size_t orig_len = len;
+    len = 2 + orig_len * 2;
+    char *destination = temp_val;
+    if (len > temp_val_max_width) {
+      init_dynamic_string(&ds_temp, "", len + 1);
+      destination = ds_temp.str;
+    }
+    const char *ptr = val;
+    const char *end = ptr + orig_len;
+    val = destination;
+    int written = sprintf(destination, "0x");
+    destination += written;
+    for (; ptr < end; ptr++, destination += written) {
+      written = sprintf(
+          destination, "%02X",
+          *(static_cast<const uchar *>(static_cast<const void *>(ptr))));
+    }
+  }
+
   if (!display_result_vertically) {
     if (col_idx) dynstr_append_mem(ds, "\t", 1);
     replace_dynstr_append_mem(ds, val, len);
@@ -8296,6 +8320,10 @@ static void append_field(DYNAMIC_STRING *ds, uint col_idx, MYSQL_FIELD *field,
     replace_dynstr_append_mem(ds, val, len);
     dynstr_append_mem(ds, "\n", 1);
   }
+
+  if (ds_temp.str != nullptr) {
+    dynstr_free(&ds_temp);
+  }
 }
 
 /*

cmake/mysql_version.cmake

@@ -26,7 +26,7 @@
 #
 
 SET(SHARED_LIB_MAJOR_VERSION "24")
-SET(SHARED_LIB_MINOR_VERSION "0")
+SET(SHARED_LIB_MINOR_VERSION "1")
 SET(PROTOCOL_VERSION "10")
 
 # Generate "something" to trigger cmake rerun when MYSQL_VERSION changes

include/field_types.h

@@ -74,6 +74,7 @@ enum enum_field_types {
   MYSQL_TYPE_DATETIME2,   /**< Internal to MySQL. Not used in protocol */
   MYSQL_TYPE_TIME2,       /**< Internal to MySQL. Not used in protocol */
   MYSQL_TYPE_TYPED_ARRAY, /**< Used for replication only */
+  MYSQL_TYPE_VECTOR = 242,
   MYSQL_TYPE_INVALID = 243,
   MYSQL_TYPE_BOOL = 244, /**< Currently just a placeholder */
   MYSQL_TYPE_JSON = 245,

include/mysql.h.pp

@@ -23,6 +23,7 @@
   MYSQL_TYPE_DATETIME2,
   MYSQL_TYPE_TIME2,
   MYSQL_TYPE_TYPED_ARRAY,
+  MYSQL_TYPE_VECTOR = 242,
   MYSQL_TYPE_INVALID = 243,
   MYSQL_TYPE_BOOL = 244,
   MYSQL_TYPE_JSON = 245,

include/mysql/components/services/bits/stored_program_bits.h

@@ -59,5 +59,6 @@
 #define MYSQL_SP_ARG_TYPE_VAR_STRING (1ULL << 31)
 #define MYSQL_SP_ARG_TYPE_STRING (1ULL << 32)
 #define MYSQL_SP_ARG_TYPE_GEOMETRY (1ULL << 33)
+#define MYSQL_SP_ARG_TYPE_VECTOR (1ULL << 34)
 
 #endif

libmysql/libmysql.cc

@@ -2751,6 +2751,7 @@ static void fetch_string_with_conversion(MYSQL_BIND *param, char *value,
     case MYSQL_TYPE_MEDIUM_BLOB:
     case MYSQL_TYPE_LONG_BLOB:
     case MYSQL_TYPE_BLOB:
+    case MYSQL_TYPE_VECTOR:
     case MYSQL_TYPE_DECIMAL:
     case MYSQL_TYPE_NEWDECIMAL:
     default: {
@@ -3360,10 +3361,10 @@ static bool is_binary_compatible(enum enum_field_types type1,
       range2[] = {MYSQL_TYPE_INT24, MYSQL_TYPE_LONG, MYSQL_TYPE_NULL},
       range3[] = {MYSQL_TYPE_DATETIME, MYSQL_TYPE_TIMESTAMP, MYSQL_TYPE_NULL},
       range4[] = {
-          MYSQL_TYPE_ENUM,        MYSQL_TYPE_SET,       MYSQL_TYPE_TINY_BLOB,
-          MYSQL_TYPE_MEDIUM_BLOB, MYSQL_TYPE_LONG_BLOB, MYSQL_TYPE_BLOB,
-          MYSQL_TYPE_VAR_STRING,  MYSQL_TYPE_STRING,    MYSQL_TYPE_GEOMETRY,
-          MYSQL_TYPE_DECIMAL,     MYSQL_TYPE_NULL};
+          MYSQL_TYPE_ENUM,        MYSQL_TYPE_SET,        MYSQL_TYPE_TINY_BLOB,
+          MYSQL_TYPE_MEDIUM_BLOB, MYSQL_TYPE_LONG_BLOB,  MYSQL_TYPE_BLOB,
+          MYSQL_TYPE_VECTOR,      MYSQL_TYPE_VAR_STRING, MYSQL_TYPE_STRING,
+          MYSQL_TYPE_GEOMETRY,    MYSQL_TYPE_DECIMAL,    MYSQL_TYPE_NULL};
   static const enum enum_field_types *range_list[] = {range1, range2, range3,
                                                       range4},
                                      **range_list_end =
@@ -3464,6 +3465,7 @@ static bool setup_one_fetch_function(MYSQL_BIND *param, MYSQL_FIELD *field) {
     case MYSQL_TYPE_MEDIUM_BLOB:
     case MYSQL_TYPE_LONG_BLOB:
     case MYSQL_TYPE_BLOB:
+    case MYSQL_TYPE_VECTOR:
     case MYSQL_TYPE_BIT:
       assert(param->buffer_length != 0);
       param->fetch_result = fetch_result_bin;
@@ -3543,6 +3545,7 @@ static bool setup_one_fetch_function(MYSQL_BIND *param, MYSQL_FIELD *field) {
     case MYSQL_TYPE_MEDIUM_BLOB:
     case MYSQL_TYPE_LONG_BLOB:
     case MYSQL_TYPE_BLOB:
+    case MYSQL_TYPE_VECTOR:
     case MYSQL_TYPE_VAR_STRING:
     case MYSQL_TYPE_STRING:
     case MYSQL_TYPE_BIT:

libs/mysql/binlog/event/binary_log_funcs.cpp

@@ -189,6 +189,7 @@ unsigned int max_display_length_for_field(enum_field_types sql_type,
        */
       return uint_max(metadata * 8);
 
+    case MYSQL_TYPE_VECTOR:
     case MYSQL_TYPE_LONG_BLOB:
     case MYSQL_TYPE_GEOMETRY:
     case MYSQL_TYPE_JSON:
@@ -347,6 +348,7 @@ uint32_t calc_field_size(unsigned char col, const unsigned char *master_data,
     case MYSQL_TYPE_MEDIUM_BLOB:
     case MYSQL_TYPE_LONG_BLOB:
     case MYSQL_TYPE_BLOB:
+    case MYSQL_TYPE_VECTOR:
     case MYSQL_TYPE_GEOMETRY:
     case MYSQL_TYPE_JSON: {
       /*

libs/mysql/binlog/event/rows_event.cpp

@@ -253,6 +253,25 @@ static void parse_geometry_type(std::vector<unsigned int> &vec,
   }
 }
 
+/**
+   Parses VECTOR_DIMENSIONALITY field.
+
+   @param[out] vec        stores vector columns's dimensionality extracted from
+   field.
+   @param[in]  reader_obj the Event_reader object containing the serialized
+                          field.
+   @param[in]  length     length of the field
+ */
+static void parse_vector_dimensionality(std::vector<unsigned int> &vec,
+                                        Event_reader &reader_obj,
+                                        unsigned int length) {
+  const char *field = reader_obj.ptr();
+  while (reader_obj.ptr() < field + length) {
+    vec.push_back(reader_obj.net_field_length_ll());
+    if (reader_obj.has_error()) return;
+  }
+}
+
 /**
    Parses SIMPLE_PRIMARY_KEY field.
 
@@ -373,6 +392,9 @@ Table_map_event::Optional_metadata_fields::Optional_metadata_fields(
       case COLUMN_VISIBILITY:
         parse_column_visibility(&m_column_visibility, reader_obj, len);
         break;
+      case VECTOR_DIMENSIONALITY:
+        parse_vector_dimensionality(m_vector_dimensionality, reader_obj, len);
+        break;
       default:
         BAPI_ASSERT(0);
     }

libs/mysql/binlog/event/rows_event.h

@@ -567,8 +567,9 @@ class Table_map_event : public Binary_log_event {
                                      columns, optimized to minimize
                                      space when many columns have the
                                      same charset. */
-    COLUMN_VISIBILITY             /* Flag to indicate column visibility
+    COLUMN_VISIBILITY,            /* Flag to indicate column visibility
                                      attribute. */
+    VECTOR_DIMENSIONALITY         /* Vector column dimensionality */
   };
 
   /**
@@ -605,6 +606,7 @@ class Table_map_event : public Binary_log_event {
     std::vector<str_vector> m_enum_str_value;
     std::vector<str_vector> m_set_str_value;
     std::vector<unsigned int> m_geometry_type;
+    std::vector<unsigned int> m_vector_dimensionality;
     /*
       The uint_pair means <column index, prefix length>.  Prefix length is 0 if
       whole column value is used.

mysql-test/r/archive.result

@@ -12975,5 +12975,14 @@ id	x
 5	6
 DROP VIEW v;
 DROP TABLE archive_table, innodb_table;
+CREATE TABLE ta (pk INT, embedding VECTOR(4)) ENGINE=ARCHIVE;
+INSERT INTO ta VALUES
+(0, TO_VECTOR("[1,2,3,4]")),
+(1, TO_VECTOR("[4,5,6,7]"));
+SELECT FROM_VECTOR(embedding) FROM ta;
+FROM_VECTOR(embedding)
+[1.00000e+00,2.00000e+00,3.00000e+00,4.00000e+00]
+[4.00000e+00,5.00000e+00,6.00000e+00,7.00000e+00]
+DROP TABLE ta;
 Warnings:
 Warning	1287	'@@binlog_format' is deprecated and will be removed in a future release.

mysql-test/r/blackhole.result

@@ -28,3 +28,10 @@ score
 SELECT 1 FROM t WHERE MATCH (a) AGAINST ('abc');
 1
 DROP TABLE t;
+CREATE TABLE ta (pk INT, embedding VECTOR(4), PRIMARY KEY (pk)) ENGINE=BLACKHOLE;
+INSERT INTO ta VALUES
+(0, TO_VECTOR("[1,2,3,4]")),
+(1, TO_VECTOR("[4,5,6,7]"));
+SELECT FROM_VECTOR(embedding) FROM ta;
+FROM_VECTOR(embedding)
+DROP TABLE ta;

mysql-test/r/csv.result

@@ -5496,3 +5496,12 @@ SELECT * FROM t;
 j
 {"a": 1, "b": 2}
 DROP TABLE t;
+CREATE TABLE ta (pk INT NOT NULL, embedding VECTOR(4) NOT NULL) ENGINE=csv;
+INSERT INTO ta VALUES
+(0, TO_VECTOR("[1,2,3,4]")),
+(1, TO_VECTOR("[4,5,6,7]"));
+SELECT FROM_VECTOR(embedding) FROM ta;
+FROM_VECTOR(embedding)
+[1.00000e+00,2.00000e+00,3.00000e+00,4.00000e+00]
+[4.00000e+00,5.00000e+00,6.00000e+00,7.00000e+00]
+DROP TABLE ta;

mysql-test/r/dd_schema_dd_properties_debug.result

@@ -603,7 +603,7 @@ check_constraints=
     'MYSQL_TYPE_MEDIUM_BLOB', 'MYSQL_TYPE_LONG_BLOB',
     'MYSQL_TYPE_BLOB', 'MYSQL_TYPE_VAR_STRING',
     'MYSQL_TYPE_STRING', 'MYSQL_TYPE_GEOMETRY',
-    'MYSQL_TYPE_JSON'
+    'MYSQL_TYPE_JSON', 'MYSQL_TYPE_VECTOR' 
   ) NOT NULL
             lbl=FIELD_TYPE
             pos=4
@@ -1835,7 +1835,7 @@ parameters=
     'MYSQL_TYPE_MEDIUM_BLOB',     'MYSQL_TYPE_LONG_BLOB', 'MYSQL_TYPE_BLOB',
     'MYSQL_TYPE_VAR_STRING',
     'MYSQL_TYPE_STRING', 'MYSQL_TYPE_GEOMETRY',
-    'MYSQL_TYPE_JSON'
+    'MYSQL_TYPE_JSON', 'MYSQL_TYPE_VECTOR'
   ) NOT NULL
             lbl=FIELD_DATA_TYPE
             pos=5
@@ -2065,7 +2065,7 @@ routines=
     'MYSQL_TYPE_MEDIUM_BLOB',     'MYSQL_TYPE_LONG_BLOB', 'MYSQL_TYPE_BLOB',
     'MYSQL_TYPE_VAR_STRING',
     'MYSQL_TYPE_STRING', 'MYSQL_TYPE_GEOMETRY',
-    'MYSQL_TYPE_JSON'
+    'MYSQL_TYPE_JSON', 'MYSQL_TYPE_VECTOR'
   ) DEFAULT NULL
             lbl=FIELD_RESULT_DATA_TYPE
             pos=4

mysql-test/r/dd_schema_definition_debug.result

@@ -165,7 +165,7 @@ SET debug = '+d,skip_dd_table_access_check';
 ########################################################################
 # The actual DD version stored on disk.
 ########################################################################
-DD_VERSION=80300
+DD_VERSION=90000
 ########################################################################
 # List the CREATE TABLE statements for the DD tables.
 # Mask the AUTO INCREMENT counter, which is not
@@ -232,7 +232,7 @@ columns	CREATE TABLE `columns` (
   `table_id` bigint unsigned NOT NULL,
   `name` varchar(64) CHARACTER SET utf8mb3 COLLATE utf8mb3_tolower_ci NOT NULL,
   `ordinal_position` int unsigned NOT NULL,
-  `type` enum('MYSQL_TYPE_DECIMAL','MYSQL_TYPE_TINY','MYSQL_TYPE_SHORT','MYSQL_TYPE_LONG','MYSQL_TYPE_FLOAT','MYSQL_TYPE_DOUBLE','MYSQL_TYPE_NULL','MYSQL_TYPE_TIMESTAMP','MYSQL_TYPE_LONGLONG','MYSQL_TYPE_INT24','MYSQL_TYPE_DATE','MYSQL_TYPE_TIME','MYSQL_TYPE_DATETIME','MYSQL_TYPE_YEAR','MYSQL_TYPE_NEWDATE','MYSQL_TYPE_VARCHAR','MYSQL_TYPE_BIT','MYSQL_TYPE_TIMESTAMP2','MYSQL_TYPE_DATETIME2','MYSQL_TYPE_TIME2','MYSQL_TYPE_NEWDECIMAL','MYSQL_TYPE_ENUM','MYSQL_TYPE_SET','MYSQL_TYPE_TINY_BLOB','MYSQL_TYPE_MEDIUM_BLOB','MYSQL_TYPE_LONG_BLOB','MYSQL_TYPE_BLOB','MYSQL_TYPE_VAR_STRING','MYSQL_TYPE_STRING','MYSQL_TYPE_GEOMETRY','MYSQL_TYPE_JSON') COLLATE utf8mb3_bin NOT NULL,
+  `type` enum('MYSQL_TYPE_DECIMAL','MYSQL_TYPE_TINY','MYSQL_TYPE_SHORT','MYSQL_TYPE_LONG','MYSQL_TYPE_FLOAT','MYSQL_TYPE_DOUBLE','MYSQL_TYPE_NULL','MYSQL_TYPE_TIMESTAMP','MYSQL_TYPE_LONGLONG','MYSQL_TYPE_INT24','MYSQL_TYPE_DATE','MYSQL_TYPE_TIME','MYSQL_TYPE_DATETIME','MYSQL_TYPE_YEAR','MYSQL_TYPE_NEWDATE','MYSQL_TYPE_VARCHAR','MYSQL_TYPE_BIT','MYSQL_TYPE_TIMESTAMP2','MYSQL_TYPE_DATETIME2','MYSQL_TYPE_TIME2','MYSQL_TYPE_NEWDECIMAL','MYSQL_TYPE_ENUM','MYSQL_TYPE_SET','MYSQL_TYPE_TINY_BLOB','MYSQL_TYPE_MEDIUM_BLOB','MYSQL_TYPE_LONG_BLOB','MYSQL_TYPE_BLOB','MYSQL_TYPE_VAR_STRING','MYSQL_TYPE_STRING','MYSQL_TYPE_GEOMETRY','MYSQL_TYPE_JSON','MYSQL_TYPE_VECTOR') COLLATE utf8mb3_bin NOT NULL,
   `is_nullable` tinyint(1) NOT NULL,
   `is_zerofill` tinyint(1) DEFAULT NULL,
   `is_unsigned` tinyint(1) DEFAULT NULL,
@@ -449,7 +449,7 @@ parameters	CREATE TABLE `parameters` (
   `ordinal_position` int unsigned NOT NULL,
   `mode` enum('IN','OUT','INOUT') COLLATE utf8mb3_bin DEFAULT NULL,
   `name` varchar(64) CHARACTER SET utf8mb3 COLLATE utf8mb3_general_ci DEFAULT NULL,
-  `data_type` enum('MYSQL_TYPE_DECIMAL','MYSQL_TYPE_TINY','MYSQL_TYPE_SHORT','MYSQL_TYPE_LONG','MYSQL_TYPE_FLOAT','MYSQL_TYPE_DOUBLE','MYSQL_TYPE_NULL','MYSQL_TYPE_TIMESTAMP','MYSQL_TYPE_LONGLONG','MYSQL_TYPE_INT24','MYSQL_TYPE_DATE','MYSQL_TYPE_TIME','MYSQL_TYPE_DATETIME','MYSQL_TYPE_YEAR','MYSQL_TYPE_NEWDATE','MYSQL_TYPE_VARCHAR','MYSQL_TYPE_BIT','MYSQL_TYPE_TIMESTAMP2','MYSQL_TYPE_DATETIME2','MYSQL_TYPE_TIME2','MYSQL_TYPE_NEWDECIMAL','MYSQL_TYPE_ENUM','MYSQL_TYPE_SET','MYSQL_TYPE_TINY_BLOB','MYSQL_TYPE_MEDIUM_BLOB','MYSQL_TYPE_LONG_BLOB','MYSQL_TYPE_BLOB','MYSQL_TYPE_VAR_STRING','MYSQL_TYPE_STRING','MYSQL_TYPE_GEOMETRY','MYSQL_TYPE_JSON') COLLATE utf8mb3_bin NOT NULL,
+  `data_type` enum('MYSQL_TYPE_DECIMAL','MYSQL_TYPE_TINY','MYSQL_TYPE_SHORT','MYSQL_TYPE_LONG','MYSQL_TYPE_FLOAT','MYSQL_TYPE_DOUBLE','MYSQL_TYPE_NULL','MYSQL_TYPE_TIMESTAMP','MYSQL_TYPE_LONGLONG','MYSQL_TYPE_INT24','MYSQL_TYPE_DATE','MYSQL_TYPE_TIME','MYSQL_TYPE_DATETIME','MYSQL_TYPE_YEAR','MYSQL_TYPE_NEWDATE','MYSQL_TYPE_VARCHAR','MYSQL_TYPE_BIT','MYSQL_TYPE_TIMESTAMP2','MYSQL_TYPE_DATETIME2','MYSQL_TYPE_TIME2','MYSQL_TYPE_NEWDECIMAL','MYSQL_TYPE_ENUM','MYSQL_TYPE_SET','MYSQL_TYPE_TINY_BLOB','MYSQL_TYPE_MEDIUM_BLOB','MYSQL_TYPE_LONG_BLOB','MYSQL_TYPE_BLOB','MYSQL_TYPE_VAR_STRING','MYSQL_TYPE_STRING','MYSQL_TYPE_GEOMETRY','MYSQL_TYPE_JSON','MYSQL_TYPE_VECTOR') COLLATE utf8mb3_bin NOT NULL,
   `data_type_utf8` mediumtext COLLATE utf8mb3_bin NOT NULL,
   `is_zerofill` tinyint(1) DEFAULT NULL,
   `is_unsigned` tinyint(1) DEFAULT NULL,
@@ -481,7 +481,7 @@ routines	CREATE TABLE `routines` (
   `schema_id` bigint unsigned NOT NULL,
   `name` varchar(64) CHARACTER SET utf8mb3 COLLATE utf8mb3_general_ci NOT NULL,
   `type` enum('FUNCTION','PROCEDURE') COLLATE utf8mb3_bin NOT NULL,
-  `result_data_type` enum('MYSQL_TYPE_DECIMAL','MYSQL_TYPE_TINY','MYSQL_TYPE_SHORT','MYSQL_TYPE_LONG','MYSQL_TYPE_FLOAT','MYSQL_TYPE_DOUBLE','MYSQL_TYPE_NULL','MYSQL_TYPE_TIMESTAMP','MYSQL_TYPE_LONGLONG','MYSQL_TYPE_INT24','MYSQL_TYPE_DATE','MYSQL_TYPE_TIME','MYSQL_TYPE_DATETIME','MYSQL_TYPE_YEAR','MYSQL_TYPE_NEWDATE','MYSQL_TYPE_VARCHAR','MYSQL_TYPE_BIT','MYSQL_TYPE_TIMESTAMP2','MYSQL_TYPE_DATETIME2','MYSQL_TYPE_TIME2','MYSQL_TYPE_NEWDECIMAL','MYSQL_TYPE_ENUM','MYSQL_TYPE_SET','MYSQL_TYPE_TINY_BLOB','MYSQL_TYPE_MEDIUM_BLOB','MYSQL_TYPE_LONG_BLOB','MYSQL_TYPE_BLOB','MYSQL_TYPE_VAR_STRING','MYSQL_TYPE_STRING','MYSQL_TYPE_GEOMETRY','MYSQL_TYPE_JSON') COLLATE utf8mb3_bin DEFAULT NULL,
+  `result_data_type` enum('MYSQL_TYPE_DECIMAL','MYSQL_TYPE_TINY','MYSQL_TYPE_SHORT','MYSQL_TYPE_LONG','MYSQL_TYPE_FLOAT','MYSQL_TYPE_DOUBLE','MYSQL_TYPE_NULL','MYSQL_TYPE_TIMESTAMP','MYSQL_TYPE_LONGLONG','MYSQL_TYPE_INT24','MYSQL_TYPE_DATE','MYSQL_TYPE_TIME','MYSQL_TYPE_DATETIME','MYSQL_TYPE_YEAR','MYSQL_TYPE_NEWDATE','MYSQL_TYPE_VARCHAR','MYSQL_TYPE_BIT','MYSQL_TYPE_TIMESTAMP2','MYSQL_TYPE_DATETIME2','MYSQL_TYPE_TIME2','MYSQL_TYPE_NEWDECIMAL','MYSQL_TYPE_ENUM','MYSQL_TYPE_SET','MYSQL_TYPE_TINY_BLOB','MYSQL_TYPE_MEDIUM_BLOB','MYSQL_TYPE_LONG_BLOB','MYSQL_TYPE_BLOB','MYSQL_TYPE_VAR_STRING','MYSQL_TYPE_STRING','MYSQL_TYPE_GEOMETRY','MYSQL_TYPE_JSON','MYSQL_TYPE_VECTOR') COLLATE utf8mb3_bin DEFAULT NULL,
   `result_data_type_utf8` mediumtext COLLATE utf8mb3_bin NOT NULL,
   `result_is_zerofill` tinyint(1) DEFAULT NULL,
   `result_is_unsigned` tinyint(1) DEFAULT NULL,
@@ -734,5 +734,5 @@ Warnings:
 Warning	1681	Integer display width is deprecated and will be removed in a future release.
 include/assert.inc [The group concat max length is sufficient.]
 CHECK_STATUS
-The schema checksum corresponds to DD version 80200.
+The schema checksum corresponds to DD version 90000.
 include/assert.inc [The schema checksum corresponds to a known DD version.]

mysql-test/r/information_schema_keywords.result

@@ -712,6 +712,7 @@ VARCHARACTER	1
 VARIABLES	0
 VARYING	1
 VCPU	0
+VECTOR	0
 VIEW	0
 VIRTUAL	1
 VISIBLE	0