MemSQL supports the following data types. The sizes given are for in-memory rowstore tables. Columnstore (on-disk) tables are stored differently, and often compressed.

Integer Numbers

Remarks

BOOL and BOOLEAN are synonymous with TINYINT. A value of 0 is considered FALSE, non-zero values are considered TRUE.

The format: INT(x) (for example, INT(5)) is used to specify display width and not the size of the integer. Display width is not directly used within MemSQL but may be used by some clients.

Real Numbers

Working with Floats and Doubles

FLOAT and DOUBLE represent approximate values and MemSQL follows IEEE 754 standards (see FLOAT | see DOUBLE) for floating point numbers.

The precision listed in the Real Numbers table refers to the number of bits used to store the significand. For FLOAT, 23 bits is about 6 digits. For larger numbers (up to 53 bits) use DOUBLE, which stores up to 14 digits.

Numbers exceeding the precision of the data type are rounded according to the IEEE 754 standard when entered. For example:

CREATE TABLE t (id float);
INSERT INTO t VALUES (123456);
INSERT INTO t VALUES (1234567);
SELECT * FROM t;

****
+---------+
| id      |
+---------+
| 123456  |
| 1234570 |
+---------+

Time and Date

Note: It is recommended to use DATETIME or DATETIME(6) column types instead of TIMESTAMP or TIMESTAMP(6) because TIMESTAMP types only support dates through 2038-01-19 as opposed to 9999-12-31 for DATETIME types. This can avoid the need to do potentially time-consuming application and database maintenance and conversion in the future.

Inserting or Updating a Column With the Current Timestamp

To allow inserts or updates of a TIMESTAMP or TIMESTAMP(6) column with the current timestamp, define the column using the DEFAULT CURRENT_TIMESTAMP clause or ON UPDATE CURRENT_TIMESTAMP clause. Use the former clause to allow inserts of the current timestamp and the latter clause to allow updates of the current timestamp.

The following example demonstrates the use of DEFAULT CURRENT_TIMESTAMP and ON UPDATE CURRENT_TIMESTAMP in column definitions:

CREATE TABLE s(id INT, ts TIMESTAMP DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
ts2 TIMESTAMP(6) DEFAULT CURRENT_TIMESTAMP(6) ON UPDATE CURRENT_TIMESTAMP(6),
ts3 TIMESTAMP);

INSERT INTO s(id) VALUES(1);

INSERT INTO s(id) VALUES(2);

SELECT * FROM s;
****
+------+---------------------+----------------------------+------+
| id   | ts                  | ts2                        | ts3  |
+------+---------------------+----------------------------+------+
|    1 | 2020-02-20 11:18:02 | 2020-02-20 11:18:02.822309 | NULL |
|    2 | 2020-02-20 11:18:13 | 2020-02-20 11:18:13.767739 | NULL |
+------+---------------------+----------------------------+------+

CREATE TABLE r(d DATETIME(6), ts TIMESTAMP(6), id INT);

INSERT r(d, id) VALUES("2017-07-01", 1);

SELECT * FROM r;
****
+----------------------------+----------------------------+------+
| d                          | ts                         | id   |
+----------------------------+----------------------------+------+
| 2017-07-01 00:00:00.000000 | NULL                       |    1 |
+----------------------------+----------------------------+------+

UPDATE r SET id = 2 WHERE id = 1;

SELECT * FROM r;
****
+----------------------------+----------------------------+------+
| d                          | ts                         | id   |
+----------------------------+----------------------------+------+
| 2017-07-01 00:00:00.000000 | NULL                       |    2 |
+----------------------------+----------------------------+------+

If you don’t need to insert or update a column with the current timestamp, it is recommended that you use the DATETIME or DATETIME(6) types instead.

The explicit_defaults_for_timestamp Variable

The variable explicit_defaults_for_timestamp determines if certain nonstandard behaviors for default values and NULL-value are disabled for the TIMESTAMP column. By default, the variable is enabled, which disables the nonstandard behaviors.

If the explicit_defaults_for_timestamp variable is enabled, MemSQL handles the TIMESTAMP column behavior as follows:

• Assigning a TIMESTAMP column value of NULL does not set it to the current timestamp. To set to the current timestamp, set the TIMESTAMP column to CURRENT_TIMESTAMP or NOW().

• If the TIMESTAMP column is not declared as NOT NULL, then it is automatically declared with the NULL attribute and permits NULL values. Assigning a value of NULL to the column sets it to NULL, not the current timestamp.

• TIMESTAMP column declared with the NOT NULL attribute do not permit NULL values.

• TIMESTAMP columns explicitly declared with the NOT NULL attribute and without an explicit DEFAULT attribute are treated as having no default value.

• To declare TIMESTAMP column with the DEFAULT CURRENT_TIMESTAMP or ON UPDATE CURRENT_TIMESTAMP attributes, these must be explicitly specified.

• The first TIMESTAMP column in a table is handled in the same way as the TIMESTAMP columns following the first one.

If the explicit_defaults_for_timestamp variable is disabled, MemSQL handles the TIMESTAMP column behavior as follows:

• If the TIMESTAMP column is not declared as NULL, then it is automatically declared with the NOT NULL attribute. Assigning a value of NULL to the column is permitted and sets the column to the current timestamp.

• The first TIMESTAMP column in a table, if not declared as NULL, DEFAULT CURRENT_TIMESTAMP or ON UPDATE CURRENT_TIMESTAMP, is automatically declared with the DEFAULT CURRENT_TIMESTAMP and ON UPDATE CURRENT_TIMESTAMP attributes.

• TIMESTAMP columns following the first one, if not declared with the NULL attribute or a DEFAULT attribute, are automatically declared as DEFAULT ‘0000-00-00 00:00:00’ (the “zero” timestamp).

String Types

Fixed-Length String

There are three fixed-size string types in MemSQL. The space consumed is directly proportional to their length definitions. Values of type CHAR require 3 x LENGTH bytes of memory, as MemSQL supports only the UTF8 character set. Values of type BINARY require LENGTH bytes of memory. char(0) not null and binary(0) not null still allocate 1 byte.

A column declared as CHAR() stores values as per the defined length. When CHAR values are stored, they are right-padded with spaces to the specified length but when the CHAR values are retrieved, trailing spaces are removed.

CREATE TABLE t1 (c1 CHAR(3));

INSERT INTO t1 VALUES ('  ');
INSERT INTO t1 VALUES ('ab ');
INSERT INTO t1 VALUES ('abcd');

SELECT c1, LENGTH (c1) FROM t1;
****
+------+-------------+
| c1   | LENGTH (c1) |
+------+-------------+
| abc  |           3 |
| ab   |           2 |
|      |           0 |
+------+-------------+

Variable-Length Strings

There are several variable-length string types. The data for these variable types are stored in a memory allocation designed for them, and pointers to their locations in memory are stored in the main table structure. This means that the space consumed by a particular value is the byte length of the string, plus an 8-byte overhead for the pointer, and a 4-byte overhead for storing the length. (There is an additional 4-byte overhead if the field is nullable.) One exception is values less than 7 bytes long. Those small values are stored directly in the pointer area.

The three main types are VARCHAR, VARBINARY, and LONGTEXT. While they all operate similarly for most English-language alphanumeric strings, there are several important differences in their behavior. The length limit of the VARCHAR is measured in Unicode characters, not in bytes. VARBINARY fields store data verbatim, regardless of whether they are valid Unicode strings. Sorting and comparing a VARBINARY uses the exact binary representation.

To reduce fragmentation, allocations for variable sized data use sizes from a fixed list in multiples of eight (16, 32, 40, etc.). Variable data of other lengths is rounded up to the next nearest, so a ten (10) byte string would allocate sixteen (16) bytes.

Comparison operators treat trailing spaces differently when used with variable length TEXT type columns such as VARCHAR(), TEXT(), MEDIUMTEXT(), and TINYTEXT(). This behavior does not apply to BLOB type columns. Comparison operators ignore the trailing spaces at the end of the string being compared, except for the LIKE operator. The LIKE operator does not ignore the trailing spaces and considers them while comparing the strings.

The following example demonstrates how comparison operators treat trailing spaces for a VARCHAR() column. Consider a table Employee as follows:

DESCRIBE Employee;
****
+----------+-------------+------+------+---------+-------+
| Field    | Type        | Null | Key  | Default | Extra |
+----------+-------------+------+------+---------+-------+
| Location | varchar(10) | YES  |      | NULL    |       |
+----------+-------------+------+------+---------+-------+

Insert a few values in the Location column, with and without trailing spaces:

INSERT INTO Employee VALUES ('BERLIN '),('BERLIN  '),('BERLIN'),('BERLIN ');

SELECT LENGTH(Location),CONCAT('[', Location, ']') FROM Employee;
****
+------------------+----------------------------+
| LENGTH(Location) | CONCAT('[', Location, ']') |
+------------------+----------------------------+
|                6 | [Berlin]                   |
|                8 | [Berlin  ]                 |
|                7 | [Berlin ]                  |
|                7 | [Berlin ]                  |
+------------------+----------------------------+

The following SELECT statements ignore the trailing spaces and return the same output:

SELECT LENGTH(Location),CONCAT('[', Location, ']') FROM Employee WHERE Location = 'Berlin ';
SELECT LENGTH(Location),CONCAT('[', Location, ']') FROM Employee WHERE Location = 'Berlin';
SELECT LENGTH(Location),CONCAT('[', Location, ']') FROM Employee WHERE Location = 'Berlin  ';
****
+------------------+----------------------------+
| LENGTH(Location) | CONCAT('[', Location, ']') |
+------------------+----------------------------+
|                6 | [Berlin]                   |
|                8 | [Berlin  ]                 |
|                7 | [Berlin ]                  |
|                7 | [Berlin ]                  |
+------------------+----------------------------+

An exception is the LIKE comparison operator, where trailing spaces are significant while comparison.

SELECT LENGTH(Location), CONCAT('[', Location, ']') FROM Employee WHERE Location LIKE 'Berlin';
****
+------------------+----------------------------+
| LENGTH(Location) | CONCAT('[', Location, ']') |
+------------------+----------------------------+
|                6 | [Berlin]                   |
+------------------+----------------------------+

JSON Type

The JSON type can contain any valid Javascript Object Notation data, including maps, arrays, and nested types. For more information see the JSON Guide .

Geospatial Types

Geospatial types contain information about points, paths, and polygons on the surface of the Earth. For more information, see the Geospatial Guide.

MemSQL Procedural SQL (MPSQL) Types

Two data types are only supported in MemSQL Procedural SQL (MPSQL) scalar-valued functions and stored procedures: ARRAY and RECORD. The QUERY data type is only supported in stored procedures. For more information about these types, see the ARRAY, RECORD, and QUERY topics.

Other Types

Using the ENUM Data Type

An ENUM is a string value chosen from a user-defined list of values that are specified when creating a table. The values in an ENUM list must be quoted strings, which are distinct from others in the list. A column can be defined as ENUM using the following syntax:

CREATE TABLE table_name (
  ---
  col_name ENUM('val1','val2',...,'valN')
)

The values in an ENUM list are indexed, starting from 1. Consider a column defined as mode ENUM('Primary','Secondary','Tertiary'). The following table displays the index for each possible value that the column can have.

The index of a NULL value is NULL. An empty string denotes an invalid value, and it has an index of 0. You can use the following query to identify the rows with invalid ENUM value assignments:

SELECT ...
WHERE enum_column = 0;

ENUM values cannot be expressions or user-defined variables. In non-strict mode, an invalid value is inserted as an empty string. In strict mode, trying to insert an invalid values returns an error. ENUM values are sorted based on their index values. The NULL values are sorted first, then the empty strings (with index 0), followed by the values in the ENUM list.

The following example shows how ENUM values are defined and used.

CREATE TABLE delStat
  ( ID VARCHAR(3),
    Status ENUM('NOT_DISPATCHED','IN_TRANSIT','DELIVERED'));

INSERT INTO delStat VALUES('xdy',1);
INSERT INTO delStat VALUES('tsg','IN_TRANSIT');
INSERT INTO delStat(ID) VALUES('qfc');

SELECT * FROM delStat ORDER BY Status ASC;
****
+------+----------------+
| ID   | Status         |
+------+----------------+
| qfc  | NULL           |
| xdy  | NOT_DISPATCHED |
| tsg  | IN_TRANSIT     |
+------+----------------+

This example shows that if an integer values is specified, the ENUM value at that index is inserted. Because the NOT NULL constraint was not applied on the column, the default value inserted is NULL.

Using the SET Data Type

The SET data type restricts the values that can be inserted for a table column, and can have zero or more string values. However, the string value must exist in the list of values specified for the column at the time of table creation.

A SET column value can contain multiple set members, separated by commas. In such a case,

• the SET member values should not individually contain commas.
• the SET member values can be inserted in any order. The results are retrieved in the same order as listed at the time of table creation.
• even if a SET member value is repeated, the results display the value only once.

The following example displays the usage of the SET data type:

Create a table column with SET data type.

CREATE TABLE setexample (col SET('v1', 'v2', 'v3', 'v4'));

Insert column values as ‘v1,v2’, ‘v2,v1’, ‘v1,v4,v4’, and ‘v3,v4,v3’.

INSERT INTO setexample (col) VALUES ('v1,v2'), ('v2,v1'), ('v1,v4,v4'), ('v3,v4,v3');

The result is retrieved in the same order in which the SET member values were listed during table creation and duplicate values are ignored:

SELECT col FROM setexample;
****
+------------------------------+
| col                          |
+------------------------------+
| v1,v2                        |
| v1,v2                        |
| v1,v4                        |
| v3,v4                        |
+------------------------------+

Comparing mismatched datatypes

Using comparisons between mismatched datatypes may degrade query performance and may use unsafe type conversions which can yield undesirable query results. MemSQL will display a warning for queries with potentially problematic comparisons between mismatched datatypes in the EXPLAIN and information_schema records for the query.

These warnings do not necessarily indicate a problem, and you may have valid reasons for comparing different datatypes. But these type mismatches are flagged to help you identify potential problems that you may not be aware of.

When you see these warnings, you should check whether the datatypes are expected to be different. You may wish to consider changing the datatypes of the fields or values involved. You may also wish to consider adding explicit type conversion operations, such as the cast operators or functions like STR_TO_DATE which convert between different types.

These warnings can be seen in EXPLAIN <query>, information_schema.plancache.plan_warnings, information_schema.mv_queries.plan_warnings, and the PlanWarnings column of show plancache in textual form, as well as in EXPLAIN JSON <query>, information_schema.plancache.plan_info, and information_schema.mv_queries.plan_info in JSON form.

Example

An example of these warnings is the following:

CREATE TABLE t (id VARCHAR(50), PRIMARY KEY (id));

EXPLAIN SELECT * FROM t WHERE id = 123;
+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| EXPLAIN                                                                                                                                                                                                                                                                |
+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| WARNING: Comparisons between mismatched datatypes which may involve unsafe datatype conversions and/or degrade performance. Consider changing the datatypes, or adding explicit typecasts. See https://docs.memsql.com/docs/mismatched-datatypes for more information. |
|                                                                                                                                                                                                                                                                        |
| WARNING: Comparison between mismatched datatypes: (`t`.`id` = 123). Types 'varchar(50) CHARACTER SET utf8 COLLATE utf8_general_ci NOT NULL' vs 'bigint(20) NOT NULL'.                                                                                                  |
|                                                                                                                                                                                                                                                                        |
| Gather partitions:all alias:remote_0                                                                                                                                                                                                                                   |
| Project [t.id]                                                                                                                                                                                                                                                         |
| Filter [t.id = 123]                                                                                                                                                                                                                                                    |
| TableScan db.t, PRIMARY KEY (id)                                                                                                                                                                                                                                       |
+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+

In this example, the query involves the expression t.i = 123, where t.i is a string field. When a numeric value is compared to a string, the string is converted to a numeric value, e.g. the string '123' is converted to the number 123.

This comparison may be problematic for multiple reasons:

• The mismatched datatypes may indicate a mistake in how the query is written or how the table is defined. The query may behave differently than intended and yield undesired results. For example, all of the strings '123', '123.0', '0123', and '123abc' compare equal to the integer 123, so while the query may be intended to retrieve a single row with the specified id, this equality comparison may actually yield multiple rows whose id values all compare equal to 123.

• The mismatched datatypes negatively impact performance of the query. If the field and constant were either both strings or both integers, the query plan would be able to use the index to efficiently seek to lookup the matching id. But because there are actually many possible string ids that match the number 123, which do not come in any particular order in terms of string lexicographic order, the query cannot seek into the index and instead must scan all the rows of the table.

data_conversion_compatibility_level

Changing the value of data_conversion_compatibility_level sync variable can change the behavior of expressions in computed columns. This variable can have the following possible values: '6.0', '6.5', '7.0'.

If a computed column value changes due to a change in data_conversion_compatibility_level, columnstore sorting, indexes, and sharding can become logically corrupted. MemSQL does not recompile an existing plan when data_conversion_compatibility_level or sql_mode changes.

Best practices to avoid data corruption

• Review the computed column expressions when changing data_conversion_compatibility_level.
• Change data_conversion_compatibility_level only as a part of an application upgrade process.
• Perform application upgrade tests.

For example, if a computed column contains corrupted data and you have to switch to a higher data_conversion_compatibility_level, you may need to drop the computed column before changing the level. Once the level is increased, add the computed column to the table. Alternatively, if a persisted computed column is used in a columnstore key, you may have to create a new table and select data into this new table. After copying the data, drop the old table and rename the new table.

The following examples demonstrate how the behavior of expressions may change if data_conversion_compatibility_level is changed.

Example 1

SET GLOBAL data_conversion_compatibility_level = '6.0';

SELECT DATE('2019-04-20') + 2;
****
+------------------------+
| DATE('2019-04-20') + 2 |
+------------------------+
|               20190422 |
+------------------------+

Example 2

SET GLOBAL data_conversion_compatibility_level = '7.0';

SELECT DATE('2019-04-20') + 2;
****
+------------------------+
| DATE('2019-04-20') + 2 |
+------------------------+
| 2019-04-22             |
+------------------------+

Was this Article Helpful?

Thanks For Your Feedback!