Skip to content
Percona Software for MySQL Documentation
Rate this page
☆ ☆ ☆ ☆ ☆
Thanks for your feedback
Thank you! The feedback has been submitted.

For help, click the link below to get free database assistance or contact our experts for personalized support.

Get help from Percona

The VECTOR Type

The VECTOR data type stores an ordered list of numeric values, called a vector, in a table column. For example, [0.1, 0.3, 0.2] represents a vector with three elements.

Vectors can represent numerical features, such as embeddings used for similarity search and machine learning. Each element in a VECTOR value uses a single-precision floating-point number.

Syntax

VECTOR(N)

N specifies the dimension of the VECTOR column and defines the maximum number of elements that a stored vector can contain.

For example, the following statement creates an embedding column with a dimension of three:

CREATE TABLE documents (
    id INT PRIMARY KEY,
    title VARCHAR(255),
    embedding VECTOR(3)
);

A VECTOR(3) column can store a vector with up to three elements. For example, it accepts vectors with two or three elements but rejects a vector with four elements.

Using the same number of elements as the declared dimension is strongly recommended.

Store vector values

Use TO_VECTOR() to convert the string representation of a vector to a VECTOR value.

The following statement inserts a three-element vector into the embedding column:

INSERT INTO documents (id, title, embedding)
VALUES (
    1,
    'Example document',
    TO_VECTOR('[0.1, 0.3, 0.2]')
);

The number of elements in the vector must not exceed the dimension specified for the VECTOR column.

Retrieve vector values

Use a SELECT statement to retrieve vector values:

SELECT id, embedding
FROM documents;

Compare vectors

Vector applications often compare vectors to determine how similar they are. The result of a comparison is a numeric distance between the vectors.

Use DISTANCE() to calculate the distance between two vectors with a supported distance metric. The two vector values passed to DISTANCE() must contain the same number of elements. A mismatch causes an error.

For example:

SELECT DISTANCE(
    embedding,
    TO_VECTOR('[0.1, 0.3, 0.2]'),
    'COSINE'
) AS distance
FROM documents;

Different distance metrics compare vectors in different ways. The DISTANCE() quickstart runs every supported metric against the same stored vector and the same query vector. For arguments, return values, and metric definitions, see DISTANCE() Function.

See also