HBase Filter 过滤器之 DependentColumnFilter 详解

前言：本文详细介绍了 HBase DependentColumnFilter 过滤器 Java&Shell API 的使用，并贴出了相关示例代码以供参考。DependentColumnFilter 也称参考列过滤器，是一种允许用户指定一个参考列或引用列来过滤其他列的过滤器，过滤的原则是基于参考列的时间戳来进行筛选。

该过滤器尝试找到该列所在的每一行，并返回该行具有相同时间戳的全部键值对;如果某行不包含这个指定的列，则什么都不返回。参数dropDependentColumn 决定参考列被返回还是丢弃，为true时表示参考列被返回，为false时表示被丢弃。可以把DependentColumnFilter理解为一个valueFilter和一个时间戳过滤器的组合。如果想要获取同一时间线的数据可以考虑使用此过滤器。比较器细节及原理请参照之前的更文：HBase Filter 过滤器之比较器 Comparator 原理及源码学习。

一。Java Api

头部代码

public class DependentColumnFilterDemo {

    private static boolean isok = false;
    private static String tableName = "test";
    private static String[] cfs = new String[]{"f1", "f2"};
    private static String[] data1 = new String[]{"row-1:f2:c3:1234abc56", "row-3:f1:c3:1234321"};
    private static String[] data2 = new String[]{
            "row-1:f1:c1:abcdefg", "row-1:f2:c2:abc", "row-2:f1:c1:abc123456", "row-2:f2:c2:1234abc567"
    };

    public static void main(String[] args) throws IOException, InterruptedException {

        MyBase myBase = new MyBase();
        Connection connection = myBase.createConnection();
        if (isok) {
            myBase.deleteTable(connection, tableName);
            myBase.createTable(connection, tableName, cfs);
            // 造数据
            myBase.putRows(connection, tableName, data1);  // 第一批数据
            Thread.sleep(10);
            myBase.putRows(connection, tableName, data2);  // 第二批数据
        }
        Table table = connection.getTable(TableName.valueOf(tableName));
        Scan scan = new Scan();

中部代码
向右滑动滚动条可查看输出结果。

        // 构造方法一
        DependentColumnFilter filter = new DependentColumnFilter(Bytes.toBytes("f1"), Bytes.toBytes("c1"));  // [row-1:f1:c1:abcdefg, row-1:f2:c2:abc, row-2:f1:c1:abc123456, row-2:f2:c2:1234abc567]

        // 构造方法二 boolean dropDependentColumn=true
        DependentColumnFilter filter = new DependentColumnFilter(Bytes.toBytes("f1"), Bytes.toBytes("c1"), true);  // [row-1:f2:c2:abc, row-2:f2:c2:1234abc567]

        // 构造方法二 boolean dropDependentColumn=false  默认为false
        DependentColumnFilter filter = new DependentColumnFilter(Bytes.toBytes("f1"), Bytes.toBytes("c1"), false); // [row-1:f1:c1:abcdefg, row-1:f2:c2:abc, row-2:f1:c1:abc123456, row-2:f2:c2:1234abc567]

        // 构造方法三 + BinaryComparator 比较器过滤数据
        DependentColumnFilter filter = new DependentColumnFilter(Bytes.toBytes("f1"), Bytes.toBytes("c1"), false,
                CompareFilter.CompareOp.EQUAL, new BinaryComparator(Bytes.toBytes("abcdefg"))); // [row-1:f1:c1:abcdefg, row-1:f2:c2:abc]

        // 构造方法三 + BinaryPrefixComparator 比较器过滤数据
        DependentColumnFilter filter = new DependentColumnFilter(Bytes.toBytes("f1"), Bytes.toBytes("c1"), false,
                CompareFilter.CompareOp.EQUAL, new BinaryPrefixComparator(Bytes.toBytes("abc")));  // [row-1:f1:c1:abcdefg, row-1:f2:c2:abc, row-2:f1:c1:abc123456, row-2:f2:c2:1234abc567]

        // 构造方法三 + SubstringComparator 比较器过滤数据
        DependentColumnFilter filter = new DependentColumnFilter(Bytes.toBytes("f1"), Bytes.toBytes("c1"), false,
                CompareFilter.CompareOp.EQUAL, new SubstringComparator("1234"));  // [row-2:f1:c1:abc123456, row-2:f2:c2:1234abc567]

        // 构造方法三 + RegexStringComparator 比较器过滤数据
        DependentColumnFilter filter = new DependentColumnFilter(Bytes.toBytes("f1"), Bytes.toBytes("c1"), false,
                CompareFilter.CompareOp.EQUAL, new RegexStringComparator("[a-z]"));  // [row-1:f1:c1:abcdefg, row-1:f2:c2:abc, row-2:f1:c1:abc123456, row-2:f2:c2:1234abc567]

        // 构造方法三 + RegexStringComparator 比较器过滤数据
        DependentColumnFilter filter = new DependentColumnFilter(Bytes.toBytes("f1"), Bytes.toBytes("c1"), false,
                CompareFilter.CompareOp.EQUAL, new RegexStringComparator("1234[a-z]"));  // []  思考题：与上例对比，想想为什么为空？

该过滤器同时也支持各比较器的不同比较语法，同之前介绍的各种过滤器是一样的，这里不再一一举例了。

尾部代码

		scan.setFilter(filter);
        ResultScanner scanner = table.getScanner(scan);
        Iterator<Result> iterator = scanner.iterator();
        LinkedList<String> keys = new LinkedList<>();
        while (iterator.hasNext()) {
            String key = "";
            Result result = iterator.next();
            for (Cell cell : result.rawCells()) {
                byte[] rowkey = CellUtil.cloneRow(cell);
                byte[] family = CellUtil.cloneFamily(cell);
                byte[] column = CellUtil.cloneQualifier(cell);
                byte[] value = CellUtil.cloneValue(cell);
                key = Bytes.toString(rowkey) + ":" + Bytes.toString(family) + ":" + Bytes.toString(column) + ":" + Bytes.toString(value);
                keys.add(key);
            }
        }
        System.out.println(keys);
        scanner.close();
        table.close();
        connection.close();
    }
}

二。Shell Api

HBase test 表数据一览：

hbase(main):009:0> scan 'test'
ROW                                              COLUMN+CELL
 row-1                                           column=f1:c1, timestamp=1589794115268, value=abcdefg
 row-1                                           column=f2:c2, timestamp=1589794115268, value=abc
 row-1                                           column=f2:c3, timestamp=1589794115241, value=1234abc56
 row-2                                           column=f1:c1, timestamp=1589794115268, value=abc123456
 row-2                                           column=f2:c2, timestamp=1589794115268, value=1234abc567
 row-3                                           column=f1:c3, timestamp=1589794115241, value=1234321
3 row(s) in 0.0280 seconds

0. 简单构造方法

hbase(main):006:0> scan 'test',{FILTER=>"DependentColumnFilter('f1','c1')"}
ROW                                              COLUMN+CELL
 row-1                                           column=f1:c1, timestamp=1589794115268, value=abcdefg
 row-1                                           column=f2:c2, timestamp=1589794115268, value=abc
 row-2                                           column=f1:c1, timestamp=1589794115268, value=abc123456
 row-2                                           column=f2:c2, timestamp=1589794115268, value=1234abc567
2 row(s) in 0.0450 seconds

hbase(main):008:0> scan 'test',{FILTER=>"DependentColumnFilter('f1','c1',false)"}
ROW                                              COLUMN+CELL
 row-1                                           column=f1:c1, timestamp=1589794115268, value=abcdefg
 row-1                                           column=f2:c2, timestamp=1589794115268, value=abc
 row-2                                           column=f1:c1, timestamp=1589794115268, value=abc123456
 row-2                                           column=f2:c2, timestamp=1589794115268, value=1234abc567
2 row(s) in 0.0310 seconds

hbase(main):007:0> scan 'test',{FILTER=>"DependentColumnFilter('f1','c1',true)"}
ROW                                              COLUMN+CELL
 row-1                                           column=f2:c2, timestamp=1589794115268, value=abc
 row-2                                           column=f2:c2, timestamp=1589794115268, value=1234abc567
2 row(s) in 0.0250 seconds

1. BinaryComparator 构造过滤器

方式一：

hbase(main):004:0> scan 'test',{FILTER=>"DependentColumnFilter('f1','c1',false,=,'binary:abcdefg')"}
ROW                                              COLUMN+CELL
 row-1                                           column=f1:c1, timestamp=1589794115268, value=abcdefg
 row-1                                           column=f2:c2, timestamp=1589794115268, value=abc
1 row(s) in 0.0330 seconds

hbase(main):005:0> scan 'test',{FILTER=>"DependentColumnFilter('f1','c1',true,=,'binary:abcdefg')"}
ROW                                              COLUMN+CELL
 row-1                                           column=f2:c2, timestamp=1589794115268, value=abc
1 row(s) in 0.0120 seconds

支持的比较运算符：= != > >= < <=，不再一一举例。

方式二：

import org.apache.hadoop.hbase.filter.CompareFilter
import org.apache.hadoop.hbase.filter.BinaryComparator
import org.apache.hadoop.hbase.filter.DependentColumnFilter

hbase(main):016:0> scan 'test',{FILTER => DependentColumnFilter.new(Bytes.toBytes("f1"), Bytes.toBytes("c1"), false,CompareFilter::CompareOp.valueOf('EQUAL'), BinaryComparator.new(Bytes.toBytes('abcdefg')))}
ROW                                              COLUMN+CELL
 row-1                                           column=f1:c1, timestamp=1589794115268, value=abcdefg
 row-1                                           column=f2:c2, timestamp=1589794115268, value=abc
1 row(s) in 0.0170 seconds

hbase(main):017:0> scan 'test',{FILTER => DependentColumnFilter.new(Bytes.toBytes("f1"), Bytes.toBytes("c1"), true,CompareFilter::CompareOp.valueOf('EQUAL'), BinaryComparator.new(Bytes.toBytes('abcdefg')))}
ROW                                              COLUMN+CELL
 row-1                                           column=f2:c2, timestamp=1589794115268, value=abc
1 row(s) in 0.0140 seconds

支持的比较运算符：LESS、LESS_OR_EQUAL、EQUAL、NOT_EQUAL、GREATER、GREATER_OR_EQUAL，不再一一举例。

推荐使用方式一，更简洁方便。

2. BinaryPrefixComparator 构造过滤器

方式一：

hbase(main):019:0> scan 'test',{FILTER=>"DependentColumnFilter('f1','c1',false,=,'binaryprefix:abc')"}
ROW                                              COLUMN+CELL
 row-1                                           column=f1:c1, timestamp=1589794115268, value=abcdefg
 row-1                                           column=f2:c2, timestamp=1589794115268, value=abc
 row-2                                           column=f1:c1, timestamp=1589794115268, value=abc123456
 row-2                                           column=f2:c2, timestamp=1589794115268, value=1234abc567
2 row(s) in 0.0330 seconds

hbase(main):020:0> scan 'test',{FILTER=>"DependentColumnFilter('f1','c1',true,=,'binaryprefix:abc')"}
ROW                                              COLUMN+CELL
 row-1                                           column=f2:c2, timestamp=1589794115268, value=abc
 row-2                                           column=f2:c2, timestamp=1589794115268, value=1234abc567
2 row(s) in 0.0600 seconds

方式二：

import org.apache.hadoop.hbase.filter.CompareFilter
import org.apache.hadoop.hbase.filter.BinaryPrefixComparator
import org.apache.hadoop.hbase.filter.DependentColumnFilter

hbase(main):023:0> scan 'test',{FILTER => DependentColumnFilter.new(Bytes.toBytes("f1"), Bytes.toBytes("c1"), false,CompareFilter::CompareOp.valueOf('EQUAL'), BinaryPrefixComparator.new(Bytes.toBytes('abc')))}
ROW                                              COLUMN+CELL
 row-1                                           column=f1:c1, timestamp=1589794115268, value=abcdefg
 row-1                                           column=f2:c2, timestamp=1589794115268, value=abc
 row-2                                           column=f1:c1, timestamp=1589794115268, value=abc123456
 row-2                                           column=f2:c2, timestamp=1589794115268, value=1234abc567
2 row(s) in 0.0180 seconds

hbase(main):022:0> scan 'test',{FILTER => DependentColumnFilter.new(Bytes.toBytes("f1"), Bytes.toBytes("c1"), true,CompareFilter::CompareOp.valueOf('EQUAL'), BinaryPrefixComparator.new(Bytes.toBytes('abc')))}
ROW                                              COLUMN+CELL
 row-1                                           column=f2:c2, timestamp=1589794115268, value=abc
 row-2                                           column=f2:c2, timestamp=1589794115268, value=1234abc567
2 row(s) in 0.0190 seconds

其它同上。

3. SubstringComparator 构造过滤器

方式一：

hbase(main):025:0> scan 'test',{FILTER=>"DependentColumnFilter('f1','c1',false,=,'substring:abc')"}
ROW                                              COLUMN+CELL
 row-1                                           column=f1:c1, timestamp=1589794115268, value=abcdefg
 row-1                                           column=f2:c2, timestamp=1589794115268, value=abc
 row-2                                           column=f1:c1, timestamp=1589794115268, value=abc123456
 row-2                                           column=f2:c2, timestamp=1589794115268, value=1234abc567
2 row(s) in 0.0340 seconds

hbase(main):024:0> scan 'test',{FILTER=>"DependentColumnFilter('f1','c1',true,=,'substring:abc')"}
ROW                                              COLUMN+CELL
 row-1                                           column=f2:c2, timestamp=1589794115268, value=abc
 row-2                                           column=f2:c2, timestamp=1589794115268, value=1234abc567
2 row(s) in 0.0160 seconds

方式二：

import org.apache.hadoop.hbase.filter.CompareFilter
import org.apache.hadoop.hbase.filter.SubstringComparator
import org.apache.hadoop.hbase.filter.DependentColumnFilter

hbase(main):028:0> scan 'test',{FILTER => DependentColumnFilter.new(Bytes.toBytes("f1"), Bytes.toBytes("c1"), false,CompareFilter::CompareOp.valueOf('EQUAL'), SubstringComparator.new('abc'))}
ROW                                              COLUMN+CELL
 row-1                                           column=f1:c1, timestamp=1589794115268, value=abcdefg
 row-1                                           column=f2:c2, timestamp=1589794115268, value=abc
 row-2                                           column=f1:c1, timestamp=1589794115268, value=abc123456
 row-2                                           column=f2:c2, timestamp=1589794115268, value=1234abc567
2 row(s) in 0.0150 seconds

hbase(main):029:0> scan 'test',{FILTER => DependentColumnFilter.new(Bytes.toBytes("f1"), Bytes.toBytes("c1"), true,CompareFilter::CompareOp.valueOf('EQUAL'), SubstringComparator.new('abc'))}
ROW                                              COLUMN+CELL
 row-1                                           column=f2:c2, timestamp=1589794115268, value=abc
 row-2                                           column=f2:c2, timestamp=1589794115268, value=1234abc567
2 row(s) in 0.0170 seconds

区别于上的是这里直接传入字符串进行比较，且只支持EQUAL和NOT_EQUAL两种比较符。

4. RegexStringComparator 构造过滤器

import org.apache.hadoop.hbase.filter.CompareFilter
import org.apache.hadoop.hbase.filter.RegexStringComparator
import org.apache.hadoop.hbase.filter.DependentColumnFilter

hbase(main):035:0> scan 'test',{FILTER => DependentColumnFilter.new(Bytes.toBytes("f1"), Bytes.toBytes("c1"), false,CompareFilter::CompareOp.valueOf('EQUAL'), RegexStringComparator.new('[a-z]'))}
ROW                                              COLUMN+CELL
 row-1                                           column=f1:c1, timestamp=1589794115268, value=abcdefg
 row-1                                           column=f2:c2, timestamp=1589794115268, value=abc
 row-2                                           column=f1:c1, timestamp=1589794115268, value=abc123456
 row-2                                           column=f2:c2, timestamp=1589794115268, value=1234abc567
2 row(s) in 0.0170 seconds

hbase(main):034:0* scan 'test',{FILTER => DependentColumnFilter.new(Bytes.toBytes("f1"), Bytes.toBytes("c1"), true,CompareFilter::CompareOp.valueOf('EQUAL'), RegexStringComparator.new('[a-z]'))}
ROW                                              COLUMN+CELL
 row-1                                           column=f2:c2, timestamp=1589794115268, value=abc
 row-2                                           column=f2:c2, timestamp=1589794115268, value=1234abc567
2 row(s) in 0.0150 seconds

该比较器直接传入字符串进行比较，且只支持EQUAL和NOT_EQUAL两种比较符。若想使用第一种方式可以传入regexstring试一下，我的版本有点低暂时不支持，不再演示了。

注意这里的正则匹配指包含关系，对应底层find()方法。

DependentColumnFilter不支持使用LongComparator比较器，且BitComparator、NullComparator比较器用之甚少，也不再介绍。

到此为止，所有的比较过滤器就总结完毕了。

查看文章全部源代码请访以下GitHub地址：

https://github.com/zhoupengbo/demos-bigdata/blob/master/hbase/hbase-filters-demos/src/main/java/com/zpb/demos/DependentColumnFilterDemo.java

转载请注明出处！欢迎关注本人微信公众号【HBase工作笔记】