Function build_vector_selector 

pub fn build_vector_selector(schema: &Schema) -> VectorSelectorSpec

Build an optimal vector selector based on the projected schema.

This function analyzes the DataFusion schema (after projection) to determine which vector fields are actually needed, enabling efficient queries that only fetch required data from Qdrant.

Arguments

• schema - The Arrow schema (potentially projected) defining which fields are needed

Returns

A VectorSelectorSpec that tells Qdrant exactly which vectors to include in the response.

Examples

use datafusion::arrow::datatypes::{Schema, Field, DataType};
use qdrant_datafusion::utils::{build_vector_selector, VectorSelectorSpec};
use std::sync::Arc;

// Schema with only metadata - no vectors needed
let metadata_schema = Schema::new(vec![
    Field::new("id", DataType::Utf8, false),
    Field::new("payload", DataType::Utf8, true),
]);
assert!(matches!(build_vector_selector(&metadata_schema), VectorSelectorSpec::None));

// Schema with unnamed vector - fetch all
let unnamed_schema = Schema::new(vec![
    Field::new("id", DataType::Utf8, false),
    Field::new("vector", DataType::List(Arc::new(Field::new("item", DataType::Float32, true))), true),
]);
assert!(matches!(build_vector_selector(&unnamed_schema), VectorSelectorSpec::All));