矢量函数

向量函数

名称 说明
COSINE_DISTANCE 返回两个向量之间的余弦距离
DOT_PRODUCT 返回两个向量之间的点积
EUCLIDEAN_DISTANCE 返回两个向量之间的欧几里得距离
MANHATTAN_DISTANCE 返回两个向量之间的曼哈顿距离
VECTOR_LENGTH 返回向量中的元素数量

COSINE_DISTANCE

语法：

cosine_distance(x: VECTOR, y: VECTOR) -> FLOAT64

说明：

返回 x 和 y 之间的余弦距离。

Web

const sampleVector = [0.0, 1, 2, 3, 4, 5];
const result = await execute(db.pipeline()
  .collection("books")
  .select(
    field("embedding").cosineDistance(sampleVector).as("cosineDistance")));
test.firestore.js
Swift
let sampleVector = [0.0, 1, 2, 3, 4, 5]
let result = try await db.pipeline()
  .collection("books")
  .select([
    Field("embedding").cosineDistance(sampleVector).as("cosineDistance")
  ])
  .execute()
PipelineSnippets.swift

Kotlin

val sampleVector = doubleArrayOf(0.0, 1.0, 2.0, 3.0, 4.0, 5.0)
val result = db.pipeline()
    .collection("books")
    .select(
        field("embedding").cosineDistance(sampleVector).alias("cosineDistance")
    )
    .execute()
DocSnippets.kt

Java

double[] sampleVector = {0.0, 1.0, 2.0, 3.0, 4.0, 5.0};
Task<Pipeline.Snapshot> result = db.pipeline()
    .collection("books")
    .select(
        field("embedding").cosineDistance(sampleVector).alias("cosineDistance")
    )
    .execute();
DocSnippets.java
Python
from google.cloud.firestore_v1.pipeline_expressions import Field
from google.cloud.firestore_v1.vector import Vector

sample_vector = Vector([0.0, 1.0, 2.0, 3.0, 4.0, 5.0])
result = (
    client.pipeline()
    .collection("books")
    .select(
        Field.of("embedding").cosine_distance(sample_vector).as_("cosineDistance")
    )
    .execute()
)
firestore_pipelines.py

DOT_PRODUCT

语法：

dot_product(x: VECTOR, y: VECTOR) -> FLOAT64

说明：

返回 x 和 y 的点积。

Web

const sampleVector = [0.0, 1, 2, 3, 4, 5];
const result = await execute(db.pipeline()
  .collection("books")
  .select(
    field("embedding").dotProduct(sampleVector).as("dotProduct")
  )
);
test.firestore.js
Swift
let sampleVector = [0.0, 1, 2, 3, 4, 5]
let result = try await db.pipeline()
  .collection("books")
  .select([
    Field("embedding").dotProduct(sampleVector).as("dotProduct")
  ])
  .execute()
PipelineSnippets.swift

Kotlin

val sampleVector = doubleArrayOf(0.0, 1.0, 2.0, 3.0, 4.0, 5.0)
val result = db.pipeline()
    .collection("books")
    .select(
        field("embedding").dotProduct(sampleVector).alias("dotProduct")
    )
    .execute()
DocSnippets.kt

Java

double[] sampleVector = {0.0, 1.0, 2.0, 3.0, 4.0, 5.0};
Task<Pipeline.Snapshot> result = db.pipeline()
    .collection("books")
    .select(
        field("embedding").dotProduct(sampleVector).alias("dotProduct")
    )
    .execute();
DocSnippets.java
Python
from google.cloud.firestore_v1.pipeline_expressions import Field
from google.cloud.firestore_v1.vector import Vector

sample_vector = Vector([0.0, 1.0, 2.0, 3.0, 4.0, 5.0])
result = (
    client.pipeline()
    .collection("books")
    .select(Field.of("embedding").dot_product(sample_vector).as_("dotProduct"))
    .execute()
)
firestore_pipelines.py

EUCLIDEAN_DISTANCE

语法：

euclidean_distance(x: VECTOR, y: VECTOR) -> FLOAT64

说明：

计算 x 和 y 之间的欧几里得距离。

Web

const sampleVector = [0.0, 1, 2, 3, 4, 5];
const result = await execute(db.pipeline()
  .collection("books")
  .select(
    field("embedding").euclideanDistance(sampleVector).as("euclideanDistance")
  )
);
test.firestore.js
Swift
let sampleVector = [0.0, 1, 2, 3, 4, 5]
let result = try await db.pipeline()
  .collection("books")
  .select([
    Field("embedding").euclideanDistance(sampleVector).as("euclideanDistance")
  ])
  .execute()
PipelineSnippets.swift

Kotlin

val sampleVector = doubleArrayOf(0.0, 1.0, 2.0, 3.0, 4.0, 5.0)
val result = db.pipeline()
    .collection("books")
    .select(
        field("embedding").euclideanDistance(sampleVector).alias("euclideanDistance")
    )
    .execute()
DocSnippets.kt

Java

double[] sampleVector = {0.0, 1.0, 2.0, 3.0, 4.0, 5.0};
Task<Pipeline.Snapshot> result = db.pipeline()
    .collection("books")
    .select(
        field("embedding").euclideanDistance(sampleVector).alias("euclideanDistance")
    )
    .execute();
DocSnippets.java
Python
from google.cloud.firestore_v1.pipeline_expressions import Field
from google.cloud.firestore_v1.vector import Vector

sample_vector = Vector([0.0, 1.0, 2.0, 3.0, 4.0, 5.0])
result = (
    client.pipeline()
    .collection("books")
    .select(
        Field.of("embedding")
        .euclidean_distance(sample_vector)
        .as_("euclideanDistance")
    )
    .execute()
)
firestore_pipelines.py

MANHATTAN_DISTANCE

语法：

manhattan_distance(x: VECTOR, y: VECTOR) -> FLOAT64

说明：

计算 x 和 y 之间的曼哈顿距离。

VECTOR_LENGTH

语法：

vector_length(vector: VECTOR) -> INT64

说明：

返回 VECTOR 中的元素数量。

Web

const result = await execute(db.pipeline()
  .collection("books")
  .select(
    field("embedding").vectorLength().as("vectorLength")
  )
);
test.firestore.js
Swift
let result = try await db.pipeline()
  .collection("books")
  .select([
    Field("embedding").vectorLength().as("vectorLength")
  ])
  .execute()
PipelineSnippets.swift

Kotlin

val result = db.pipeline()
    .collection("books")
    .select(
        field("embedding").vectorLength().alias("vectorLength")
    )
    .execute()
DocSnippets.kt

Java

Task<Pipeline.Snapshot> result = db.pipeline()
    .collection("books")
    .select(
        field("embedding").vectorLength().alias("vectorLength")
    )
    .execute();
DocSnippets.java
Python
from google.cloud.firestore_v1.pipeline_expressions import Field

result = (
    client.pipeline()
    .collection("books")
    .select(Field.of("embedding").vector_length().as_("vectorLength"))
    .execute()
)
firestore_pipelines.py