FILL_VECTOR Function

Overview

The FILL_VECTOR function creates a vector of the specified dimension and fills all elements with a given value. This function is commonly used to initialize fixed-value vectors, such as creating zero vectors, unit vectors, or reference vectors for vector operations.

Syntax

fill_vector(dimension, fill_value)

Parameter Description

• dimension: The dimension of the vector, must be a positive integer constant (determined at compile time), type INT. Specifies the number of elements in the generated vector.
• fill_value: The fill value used to populate each element of the vector. Supports the following types:
  • TINYINT: Generates a vector(tinyint, N) type vector
  • INT and other integer types (SMALLINT, BIGINT): Generates a vector(int, N) type vector
  • FLOAT, DOUBLE, DECIMAL: Generates a vector(float, N) type vector

Return Result

Returns a VECTOR type value where the vector dimension equals the dimension parameter, and all elements have the value of fill_value. The element type of the vector is determined by the type of fill_value.

Examples

1. Create a 3-dimensional float vector with all elements filled to 1.0:

   SELECT fill_vector(3, 1.0) AS res;
   +-----------+
   |    res    |
   +-----------+
   | [1,1,1]   |
   +-----------+
   

2. Create a 4-dimensional zero vector and use it with cosine_distance:

   SELECT cosine_distance(vector(1.0, 2.0, 3.0, 4.0), fill_vector(4, 1.0)) AS cos_dist;
   +---------------------+
   |      cos_dist       |
   +---------------------+
   | 0.13397459621556141 |
   +---------------------+
   

Notes

• The dimension parameter must be a positive integer constant determinable at compile time; dynamic expressions are not supported.
• fill_value does not support NULL; passing NULL will cause a semantic analysis error.
• The type of the fill value determines the element type of the vector: TINYINT generates vector(tinyint, N), other integer types generate vector(int, N), and floating-point and DECIMAL types generate vector(float, N).
• fill_vector is commonly used to create reference vectors, for example when used with vector distance functions such as cosine_distance and l2_distance.
• The vector type is a specialized form of the array type and can be mutually converted and used interchangeably with the array type.