“林子雨大数据” 实验3 HBase操作与接口编程环境搭建
VM虚拟机和Ubuntu系统的安装
在Windows中使用VirtualBox安装Ubuntu虚拟机（2020年7月版本）_厦大数据库实验室博客 (xmu.edu.cn)
Hadoop安装（伪分布式）
Hadoop3.1.3安装教程_单机/伪分布式配置_Hadoop3.1.3/Ubuntu18.04(16.04)_厦大数据库实验室博客 (xmu.edu.cn)
HBase安装（伪分布式）
HBase2.2.2安装和编程实践指南_厦大数据库实验室博客 (xmu.edu.cn)
通过NAT转发使本地主机连接虚拟机
PowerShell SSH 连接 VirtualBox Ubuntu 虚拟机的具体步骤 – 小能日记 – 博客园 (cnblogs.com)
文件传输工具
FlashFXP – Secure FTP Client Software for Windows. Upload, Download, and Synchronize your files.
在VSCODE中使用SSH进行远程开发
在VScode中使用SSH进行远程开发_vscode ssh_Shipmaster_23的博客-CSDN博客
Linux中安装go环境并运行
【golang】linux中安装go环境并运行_linux 运行golang_沉默小管的博客-CSDN博客
理解Hbase
我终于看懂了HBase，太不容易了… – 知乎 (zhihu.com)
第三方软件包
tsuna/gohbase: Pure-Go HBase client (github.com)
hrpc package – github.com/tsuna/gohbase/hrpc – Go Packages
gin框架 · Go语言中文文档 (topgoer.com)
遇到过的问题
go build command fails with “$GOPATH/go.mod exists but should not” · Issue #38896 · golang/go (github.com)
Infinite Getting code actions from ”Go’, ‘Go’, ‘Go” on save · Issue #3105 · microsoft/vscode-go (github.com)
Linux 查看端口占用情况 | 菜鸟教程 (runoob.com)
接口编程
实验过程创建三个表
对应第2题答案
create 'student','S_No','S_Name','S_Sex','S_Age'put 'student','s001','S_No','2015001'put 'student','s001','S_Name','zhangsan'put 'student','s001','S_Sex','male'put 'student','s001','S_Age','23'put 'student','s002','S_No','2015002'put 'student','s002','S_Name','Mary'put 'student','s002','S_Sex','female'put 'student','s002','S_Age','22'put 'student','s003','S_No','2015003'put 'student','s003','S_Name','Lisi'put 'student','s003','S_Sex','male'put 'student','s003','S_Age','24'create 'course','C_No','C_Name','C_Credit'put 'course','c001','C_No','123001'put 'course','c001','C_Name','Math'put 'course','c001','C_Credit','2.0'put 'course','c002','C_No','123002'put 'course','c002','C_Name','Computer'put 'course','c002','C_Credit','5.0'put 'course','c003','C_No','123003'put 'course','c003','C_Name','English'put 'course','c003','C_Credit','3.0'create 'SC','SC_Sno','SC_Cno','SC_Score'put 'SC','sc001','SC_Sno','2015001'put 'SC','sc001','SC_Cno','123001'put 'SC','sc001','SC_Score','86'put 'SC','sc002','SC_Sno','2015001'put 'SC','sc002','SC_Cno','123002'put 'SC','sc002','SC_Score','77'put 'SC','sc003','SC_Sno','2015002'put 'SC','sc003','SC_Cno','123002'put 'SC','sc003','SC_Score','77'put 'SC','sc004','SC_Sno','2015002'put 'SC','sc004','SC_Cno','123003'put 'SC','sc004','SC_Score','99'put 'SC','sc005','SC_Sno','2015003'put 'SC','sc005','SC_Cno','123001'put 'SC','sc005','SC_Score','98'put 'SC','sc006','SC_Sno','2015003'put 'SC','sc006','SC_Cno','123002'put 'SC','sc006','SC_Score','95'后端编程
后端在启动时分别与本地HBase建立管理员客户端、普通客户端的连接。管理员客户端管理所有表，普通客户端管理表数据的增删改查。并启动HTTP服务提供一系列API接口。
package variableimport "github.com/tsuna/gohbase"var AdminClient gohbase.AdminClientvar Client gohbase.Clientpackage mainimport ("github.com/gin-gonic/gin""github.com/tsuna/gohbase""wolflong.com/hbase_gin/router""wolflong.com/hbase_gin/variable")func init() {variable.AdminClient = gohbase.NewAdminClient("127.0.0.1")variable.Client = gohbase.NewClient("127.0.0.1")}func main() {r := gin.Default()router.Router(r)r.Run(":1313")}给定一个处理错误的通用方法
package controllerimport "github.com/gin-gonic/gin"func checkError(err error, c *gin.Context, handlers ...gin.HandlerFunc) {if err != nil {c.JSON(500, gin.H{"error": "致命错误", "back": err.Error()})panic(err)}}1.1 列出所有表的相关信息
HBase Shell 对应代码
list
从管理客户端发出请求，获取所有表，遍历表将表的命名空间与表名存储至切片中以json格式返回
type table struct {Namespace string `json:"namespace"`Qualifier string `json:"qualifier"`}func ShowTableList(c *gin.Context) {var tables []tablet, err := hrpc.NewListTableNames(context.Background())checkError(err, c)res, err := variable.AdminClient.ListTableNames(t)checkError(err, c)for _, v := range res {tables = append(tables, table{string(v.GetNamespace()), string(v.GetQualifier())})}fmt.Println(tables)c.JSON(200, tables)}测试结果
[ { "namespace": "default", "qualifier": "SC" }, { "namespace": "default", "qualifier": "course" }, { "namespace": "default", "qualifier": "student" }, { "namespace": "default", "qualifier": "test" }]1.2 打印指定表的所有记录数据
HBase Shell 对应代码
scan "course"
获取请求参数table，从普通客户端发出请求，获取scanner，遍历scanner直到io.EOF。请注意，当查询的表没数据时，第一条拿到的是nil数据，需要进行判断。
func ShowTableRows(c *gin.Context) {var Cells [][]*hrpc.Cellt, err := hrpc.NewScan(context.Background(), []byte(c.Query("table")))checkError(err, c)res := variable.Client.Scan(t)row, err := res.Next()for err != io.EOF && row != nil {Cells = append(Cells, row.Cells)fmt.Println(row.Cells)row, err = res.Next()}c.JSON(200, Cells)}localhost:1313/TableRows?table=course 测试结果。
请注意，Gin框架在转换二进制数据时使用BASE64编码，使得控制字符在网络上正常传输。
[ [ { "row": "YzAwMQ==", "family": "Q19DcmVkaXQ=", "timestamp": 1680431640294, "cell_type": 4, "value": "Mi4w" }, { "row": "YzAwMQ==", "family": "Q19DcmVkaXQ=", "qualifier": "bmV3", "timestamp": 1680432352886, "cell_type": 4, "value": "NS4w" }, { "row": "YzAwMQ==", "family": "Q19OYW1l", "timestamp": 1680431640279, "cell_type": 4, "value": "TWF0aA==" }, { "row": "YzAwMQ==", "family": "Q19Obw==", "timestamp": 1680431640250, "cell_type": 4, "value": "MTIzMDAx" } ], [ { "row": "YzAwMg==", "family": "Q19DcmVkaXQ=", "timestamp": 1680431640328, "cell_type": 4, "value": "NS4w" }, { "row": "YzAwMg==", "family": "Q19OYW1l", "timestamp": 1680431640318, "cell_type": 4, "value": "Q29tcHV0ZXI=" }, { "row": "YzAwMg==", "family": "Q19Obw==", "timestamp": 1680431640305, "cell_type": 4, "value": "MTIzMDAy" } ], [ { "row": "YzAwMw==", "family": "Q19DcmVkaXQ=", "timestamp": 1680431640363, "cell_type": 4, "value": "My4w" }, { "row": "YzAwMw==", "family": "Q19OYW1l", "timestamp": 1680431640352, "cell_type": 4, "value": "RW5nbGlzaA==" }, { "row": "YzAwMw==", "family": "Q19Obw==", "timestamp": 1680431640343, "cell_type": 4, "value": "MTIzMDAz" } ]]1.3 向已创建好的表添加和删除指定的列族或列
HBase Shell 对应代码
put 'course','c001','C_Credit:new','5.0'delete 'course','c001','C_Credit:new'使用普通客户端进行put操作，需要准备一个item数据包含当前操作的列族或列以及对应的值。支持覆盖重写与新增。
func TableInsertRowCol(c *gin.Context) {table := c.PostForm("table")rowKey := c.PostForm("rowKey")colFamily := c.PostForm("colFamily")col := c.PostForm("col")val := c.PostForm("val")var item map[string]map[string][]byte = make(map[string]map[string][]byte)item[colFamily] = make(map[string][]byte)item[colFamily][col] = []byte(val)fmt.Println(item)t, err := hrpc.NewPutStr(context.Background(), table, rowKey, item)checkError(err, c)res, err := variable.Client.Put(t)checkError(err, c)c.JSON(200, res)}
测试结果
{ "Cells": null, "Stale": false, "Partial": false, "Exists": null}
删除列族或列，不需要val数据。列可以为空字符串，仅删除列族，不为空时则删除指定的列
func TableDeleteRowCol(c *gin.Context) {table := c.PostForm("table")rowKey := c.PostForm("rowKey")colFamily := c.PostForm("colFamily")col := c.PostForm("col")// val := c.PostForm("val")var item map[string]map[string][]byte = make(map[string]map[string][]byte)item[colFamily] = make(map[string][]byte)item[colFamily][col] = []byte{}fmt.Println(item)t, err := hrpc.NewPutStr(context.Background(), table, rowKey, item)checkError(err, c)res, err := variable.Client.Delete(t)checkError(err, c)c.JSON(200, res)}
1.4 清空指定表的所有数据
HBase Shell 对应代码
truncate 'course'先disable表，再删除表。
func TableDelete(c *gin.Context) {t := hrpc.NewDisableTable(context.Background(), []byte(c.Query("table")))err := variable.AdminClient.DisableTable(t)checkError(err, c)t2 := hrpc.NewDeleteTable(context.Background(), []byte(c.Query("table")))err = variable.AdminClient.DeleteTable(t2)checkError(err, c)c.JSON(200, gin.H{"result": "删除成功"})}1.5 统计表的行数
HBase Shell 对应代码
count 'course'修改1.2的代码
func ShowTableRowsCount(c *gin.Context) {var count intt, err := hrpc.NewScan(context.Background(), []byte(c.Query("table")))checkError(err, c)res := variable.Client.Scan(t)row, err := res.Next()for err != io.EOF && row != nil {count++row, err = res.Next()}c.JSON(200, count)}localhost:1313/TableRowsCount?table=course 测试结果
33.1 创建表
func TableCreate(c *gin.Context) {table := c.PostForm("table")fs := c.PostForm("fields")var fields []string// fmt.Println(table, fs)err := json.Unmarshal([]byte(fs), &fields)checkError(err, c)// 验证是否存在表flag := falset, err := hrpc.NewListTableNames(context.Background())checkError(err, c)res, err := variable.AdminClient.ListTableNames(t)checkError(err, c)for _, v := range res {if string(v.GetQualifier()) == table {flag = true}}// 如存在删除表if flag {t := hrpc.NewDisableTable(context.Background(), []byte(table))err := variable.AdminClient.DisableTable(t)checkError(err, c)t2 := hrpc.NewDeleteTable(context.Background(), []byte(table))err = variable.AdminClient.DeleteTable(t2)checkError(err, c)}// 插入新表var items map[string]map[string]string = make(map[string]map[string]string)for _, v := range fields {items[v] = make(map[string]string)}t2 := hrpc.NewCreateTable(context.Background(), []byte(table), items)err = variable.AdminClient.CreateTable(t2)checkError(err, c)c.JSON(200, gin.H{"result": "创建成功"})}通过1.1函数接口我们可知test表已经存在，现在我们使用3.1函数接口重新创建该表，并为接下来的3.2函数接口调用做准备。
{ "result": "创建成功"}3.2 新增记录
为了增强健壮性，我们需要判断传入的fields、values参数个数是否一致，否则应当主动报错。
func TableInsertRow(c *gin.Context) {table := c.PostForm("table")rowKey := c.PostForm("rowKey")fs := c.PostForm("fields")vs := c.PostForm("values")var fields []stringvar values []stringerr := json.Unmarshal([]byte(fs), &fields)checkError(err, c)err = json.Unmarshal([]byte(vs), &values)checkError(err, c)if len(fields) != len(values) {checkError(fmt.Errorf("数量不一致"), c)}var item map[string]map[string][]byte = make(map[string]map[string][]byte)for i, v := range fields {vs := strings.Split(v, ":")item[vs[0]] = make(map[string][]byte)if len(vs) > 1 {item[vs[0]][vs[1]] = []byte(values[i])} else {item[vs[0]][""] = []byte(values[i])}}fmt.Println(item)t, err := hrpc.NewPutStr(context.Background(), table, rowKey, item)checkError(err, c)res, err := variable.Client.Put(t)checkError(err, c)c.JSON(200, res)}
3.3 通过列过滤数据
使用scanner实现，也可以通过过滤器实现。
// TODO USE FILTERtype item struct {Row string `json:"row"`Family string `json:"family"`Qualifier string `json:"qualifier"`Timestamp *uint64 `json:"timestamp"`Cell_type *pb.CellType `json:"cell_type"`Value string `json:"value"`}func TableColumnScan(c *gin.Context) {table := c.Query("table")column := c.Query("column")vs := strings.Split(column, ":")var items []itemt, err := hrpc.NewScan(context.Background(), []byte(table))checkError(err, c)res := variable.Client.Scan(t)row, err := res.Next()for err != io.EOF && row != nil {for _, v := range row.Cells {if string(v.Family) != vs[0] {continue}if len(vs) > 1 {if string(v.Qualifier) != vs[1] {continue}}fmt.Println(row.Cells)items = append(items, item{Row: string(v.Row),Family: string(v.Family),Qualifier: string(v.Qualifier),Timestamp: v.Timestamp,Cell_type: v.CellType,Value: string(v.Value),})}row, err = res.Next()}c.JSON(200, items)}再执行一遍1.3添加列的函数，调用接口，执行结果如下。
localhost:1313/TableColumnScan?table=course&column=C_Credit
[ { "row": "c001", "family": "C_Credit", "qualifier": "", "timestamp": 1680431640294, "cell_type": 4, "value": "2.0" }, { "row": "c001", "family": "C_Credit", "qualifier": "new", "timestamp": 1680434951646, "cell_type": 4, "value": "5.0" }, { "row": "c002", "family": "C_Credit", "qualifier": "", "timestamp": 1680431640328, "cell_type": 4, "value": "5.0" }, { "row": "c003", "family": "C_Credit", "qualifier": "", "timestamp": 1680431640363, "cell_type": 4, "value": "3.0" }]localhost:1313/TableColumnScan?table=course&column=C_Credit:new
[ { "row": "c001", "family": "C_Credit", "qualifier": "new", "timestamp": 1680434951646, "cell_type": 4, "value": "5.0" }]3.4 修改行数据
与 1.3 函数代码一致
3.5 删除表指定记录
package controllerimport ("context""github.com/gin-gonic/gin""github.com/tsuna/gohbase/hrpc""wolflong.com/hbase_gin/variable")func TableDeleteRow(c *gin.Context) {table := c.PostForm("table")rowKey := c.PostForm("rowKey")t, err := hrpc.NewDelStr(context.Background(), table, rowKey, nil)checkError(err, c)res, err := variable.Client.Delete(t)checkError(err, c)c.JSON(200, res)}
{ "Cells": null, "Stale": false, "Partial": false, "Exists": null}再次调用1.5函数接口，执行结果符合预期。
2
总结
装环境2小时，代码和文档编写4.5小时，代码编写过程中查阅官方文档和解决问题的时间为3小时。
总共花费6.5个小时，共编写333行代码，56行表数据。
代码编写能力得到了提升。提高了自己对HBase的理解，作为一个典型的NoSQL数据库，其一大优点是可在廉价PC服务器上搭建起大规模结构化存储集群，并提供易使用的HBase Shell操作数据集，水平扩展方便。