使用milvus-sdk-go的迭代器导出数据

使用milvus-sdk-go的迭代器导出数据

迭代器是一种功能强大的工具，可帮助您使用主键值和布尔表达式迭代集合中的大量数据或所有数据。这可以显著改善您检索数据的方式。与传统的offset和limit参数用法不同，后者可能会随着时间的推移而变得效率低下，而迭代器提供了更具可扩展性的解决方案。

当表数据很大，需要全量导出，我们可以使用迭代器，例如每次只查询1000行数据，直到所有数据查询完成，同时也可以减少服务器压力。

需要注意的是迭代器是一个客户端实现。

下面列举一个例子：写入3000条数据，每次读取100条，直至完全读完完毕。

package main

import (
	"context"
	"fmt"
	"io"
	"log"
	"math/rand"
	"strconv"

	"github.com/milvus-io/milvus-sdk-go/v2/client"
	"github.com/milvus-io/milvus-sdk-go/v2/entity"
)

const (
	milvusAddr     = `192.168.230.71:19530`
	nEntities, dim = 3000, 128
	collectionName = "hello_iterator"

	msgFmt                                     = "==== %s ====\n"
	idCol, randomCol, addressCol, embeddingCol = "ID", "random", "address", "embeddings"
	topK                                       = 3
)

func main() {
	ctx := context.Background()

	log.Printf(msgFmt, "start connecting to Milvus")
	c, err := client.NewClient(ctx, client.Config{
		Address: milvusAddr,
	})
	if err != nil {
		log.Fatal("failed to connect to milvus, err: ", err.Error())
	}
	defer c.Close()

	// delete collection if exists
	has, err := c.HasCollection(ctx, collectionName)
	if err != nil {
		log.Fatalf("failed to check collection exists, err: %v", err)
	}
	if has {
		c.DropCollection(ctx, collectionName)
	}

	// create collection
	log.Printf(msgFmt, fmt.Sprintf("create collection, `%s`", collectionName))
	schema := entity.NewSchema().WithName(collectionName).WithDescription("hello_milvus is the simplest demo to introduce the APIs").
		WithField(entity.NewField().WithName(idCol).WithDataType(entity.FieldTypeInt64).WithIsPrimaryKey(true).WithIsAutoID(false)).
		WithField(entity.NewField().WithName(randomCol).WithDataType(entity.FieldTypeDouble)).
		WithField(entity.NewField().WithName(addressCol).WithDataType(entity.FieldTypeVarChar).WithTypeParams(entity.TypeParamMaxLength, "50")).
		WithField(entity.NewField().WithName(embeddingCol).WithDataType(entity.FieldTypeFloatVector).WithDim(dim))

	if err := c.CreateCollection(ctx, schema, entity.DefaultShardNumber); err != nil { // use default shard number
		log.Fatalf("create collection failed, err: %v", err)
	}

	// build index
	log.Printf(msgFmt, "start creating index IVF_FLAT")
	idx, err := entity.NewIndexIvfFlat(entity.L2, 128)
	if err != nil {
		log.Fatalf("failed to create ivf flat index, err: %v", err)
	}
	if err := c.CreateIndex(ctx, collectionName, embeddingCol, idx, false); err != nil {
		log.Fatalf("failed to create index, err: %v", err)
	}

	log.Printf(msgFmt, "start loading collection")
	err = c.LoadCollection(ctx, collectionName, false)
	if err != nil {
		log.Fatalf("failed to load collection, err: %v", err)
	}

	// insert data
	log.Printf(msgFmt, "start inserting random entities")
	idList, randomList := make([]int64, 0, nEntities), make([]float64, 0, nEntities)
	addressList := make([]string, 0, nEntities)
	embeddingList := make([][]float32, 0, nEntities)

	// generate data
	for i := 0; i < nEntities; i++ {
		idList = append(idList, int64(i))
	}
	for i := 0; i < nEntities; i++ {
		randomList = append(randomList, rand.Float64())
	}
	for i := 0; i < nEntities; i++ {
		addressList = append(addressList, "wuhan"+strconv.Itoa(i))
	}
	for i := 0; i < nEntities; i++ {
		vec := make([]float32, 0, dim)
		for j := 0; j < dim; j++ {
			vec = append(vec, rand.Float32())
		}
		embeddingList = append(embeddingList, vec)
	}
	idColData := entity.NewColumnInt64(idCol, idList)
	randomColData := entity.NewColumnDouble(randomCol, randomList)
	addressColData := entity.NewColumnVarChar(addressCol, addressList)
	embeddingColData := entity.NewColumnFloatVector(embeddingCol, dim, embeddingList)

	if _, err := c.Insert(ctx, collectionName, "", idColData, randomColData, addressColData, embeddingColData); err != nil {
		log.Fatalf("failed to insert random data into `hello_milvus, err: %v", err)
	}

	if err := c.Flush(ctx, collectionName, false); err != nil {
		log.Fatalf("failed to flush data, err: %v", err)
	}
    // 使用迭代器,每次读取100行数据
	itr, err := c.QueryIterator(ctx, client.NewQueryIteratorOption(collectionName).WithOutputFields(idCol, randomCol, embeddingCol).WithBatchSize(100))
	if err != nil {
		log.Fatal("failed to query iterator: ", err.Error())
	}
	for {
		rs, err := itr.Next(ctx)
		if err != nil {
			if err == io.EOF {
				log.Println("iterator reach EOF")
				break
			}
			log.Fatal("failed to query iterator. next: ", err.Error())
		}
		var idlist []int64
		var randomlist []float64
		for _, col := range rs {
			if col.Name() == idCol {
				idColumn := col.(*entity.ColumnInt64)
				for i := 0; i < col.Len(); i++ {
					val, err := idColumn.ValueByIdx(i)
					if err != nil {
						log.Fatal(err)
					}
					idlist = append(idlist, val)
				}
			}
			if col.Name() == randomCol {
				randomColumn := col.(*entity.ColumnDouble)
				for i := 0; i < col.Len(); i++ {
					val, err := randomColumn.ValueByIdx(i)
					if err != nil {
						log.Fatal(err)
					}
					randomlist = append(randomlist, val)
				}
			}
		}
		log.Printf("\tids: %#v\n", idlist)
		log.Printf("\trandoms: %#v\n", randomlist)
	}

	// drop collection
	log.Printf(msgFmt, "drop collection `hello_milvus`")
	if err := c.DropCollection(ctx, collectionName); err != nil {
		log.Fatalf("failed to drop collection, err: %v", err)
	}
}