多参表达式Hive UDF

zjk 发布于 1 天前 9 次阅读


使用说明

支持的操作符

  1. :跳过,即无条件筛选
  2. =:等于
  3. !=:不等于
  4. range:区间内,range[n,m]表示 between n and m
  5. nrange:区间外,即not between and 
  6. in:集合内,in(n,m,j,k)表示 in ('n','m','j','k')
  7. nin:集合外,即not in
  8. regex:正则表达式,regex(^a.*)表示以a字母头

输入与输出

  1. 输入:参数个数成双。每一对参数中,参数1为任意类型的筛选列,参数2为字符串类型的条件列。
  2. 输出:true or false

demo

select
    -- 等于空串
    multi_exp('', '=') as t1,
    -- 空条件key 跳过比较
    multi_exp('abc', '') as t2,
    -- 数字会自动转为string比较
    multi_exp(123, '=123') as t3,
    -- string类型
    multi_exp('abc', '=abc') as t4,
    multi_exp('abc', '!=aaa') as t5,
    multi_exp('23', 'range[1,100]') as t6,
    multi_exp('23', 'nrange[1,100]') as t7,
    multi_exp('abc', 'in(abc,aaa,bbb,ccc)') as t8,
    multi_exp('abc', 'nin(abc,aaa,bbb,ccc)') as t9,
    multi_exp('abc', 'regex(^a.+)') as t10,
    multi_exp(
                'abc', '=abc'
                , 'abc', '!=aaa'
                , '23', 'range[1,100]'
                , 'abc', 'in(abc,aaa,bbb,ccc)'
                , 'abc', 'regex(^a.+)'
        ) t11
    -- map类型
    ,
    multi_exp(str_to_map('k1:v1,k2:v2,k3:v3', ',', ':'), '{"k1":"=v1","k2":"=v2"}') as t12, -- k1=v1 and k2=v2, true
    multi_exp(str_to_map('k1:v1,k2:v2,k3:v3', ',', ':'), '{"k1":"=v1","k2":"=v3"}') as t13  -- k1=v1 但是k2!=v3, false
    -- json格式字符串
    ,
    multi_exp('{"k1":"v1", "k2":"v2", "k3":"v3"}', '{"k1":"=v1","k2":"=v2"}') as t14,  -- k1=v1 and k2=v2, true
    multi_exp('{"k1":"v1", "k2":"v2", "k3":"v3"}', '{"k1":"=v1","k3":"=v2"}') as t15  -- k1=v1 但是k3!=v2, false
;

源码

import com.alibaba.fastjson.JSONObject;
import org.apache.hadoop.hive.ql.exec.UDF;
import java.util.*;
import java.util.regex.Pattern;
public class MultiExpUDF extends UDF {
private static final String IGNORE = "";
private static final String OPERATOR_EQ = "=";
private static final String OPERATOR_NE = "!=";
private static final String OPERATOR_RANGE = "range";
private static final String OPERATOR_NRANGE = "nrange";
private static final String OPERATOR_IN = "in";
private static final String OPERATOR_NIN = "nin";
private static final String OPERATOR_REGEX = "regex";
public boolean evaluate(Object... args) {
if (args.length % 2 != 0) {
return false;
}
for (int i = 0; i < args.length; i += 2) {
Object originalValue = args[i];
Object conditionValue = args[i + 1];
// 明细列处理 空值转空串
originalValue = originalValue == null ? "" : originalValue;
// 跳过指定条件key
if (IGNORE.equals(conditionValue)) {
continue;
}
try {
if (originalValue instanceof Map) {
@SuppressWarnings("unchecked")
Map<String, String> originalMap = (Map<String, String>) originalValue;
String conditionStr = conditionValue.toString();
if (!conditionStr.startsWith("{") || !conditionStr.endsWith("}")) {
return false;
}
JSONObject conditionJson = JSONObject.parseObject(conditionStr);
Set<String> keys = conditionJson.keySet();
for (String key : keys) {
String condition = conditionJson.getString(key);
String original = originalMap.get(key);
// 处理map中key为null或空的情况
original = original == null ? "" : original;
if (compares(original, condition)) {
return false;
}
}
} else {
String originalStr = originalValue.toString();
try {
JSONObject originalJson = JSONObject.parseObject(originalStr);
String conditionStr = conditionValue.toString();
if (conditionStr.startsWith("{") && conditionStr.endsWith("}")) {
JSONObject conditionJson = JSONObject.parseObject(conditionStr);
Set<String> keys = conditionJson.keySet();
for (String key : keys) {
String condition = conditionJson.getString(key);
String original = originalJson.getString(key);
// 处理json中key为null或空的情况
original = original == null ? "" : original;
if (compares(original, condition)) {
return false;
}
}
} else {
if (compares(originalStr, conditionStr)) {
return false;
}
}
} catch (Exception e) {
if (compares(originalStr, conditionValue.toString())) {
return false;
}
}
}
} catch (Exception e) {
return false;
}
}
return true;
}
private boolean compares(String original, String condition) {
if (condition.startsWith(OPERATOR_EQ)) {
String expected = condition.substring(1);
if (expected.isEmpty()) {
return !original.isEmpty();
}
return !original.equals(expected);
} else if (condition.startsWith(OPERATOR_NE)) {
String expected = condition.substring(2);
if (expected.isEmpty()) {
return original.isEmpty();
}
return original.equals(expected);
} else if (condition.startsWith(OPERATOR_RANGE)) {
String rangeStr = condition.substring(OPERATOR_RANGE.length() + 1, condition.length() - 1);
String[] range = rangeStr.split(",");
if (range.length == 2) {
try {
Comparable<?> originalVal = getComparable(original);
Comparable<?> lower = getComparable(range[0]);
Comparable<?> upper = getComparable(range[1]);
if (originalVal.getClass().equals(lower.getClass()) && originalVal.getClass().equals(upper.getClass())) {
@SuppressWarnings("unchecked")
int lowerCompare = ((Comparable<Object>) originalVal).compareTo(lower);
@SuppressWarnings("unchecked")
int upperCompare = ((Comparable<Object>) originalVal).compareTo(upper);
return lowerCompare < 0 || upperCompare > 0;
}
} catch (Exception e) {
return true;
}
}
return true;
} else if (condition.startsWith(OPERATOR_NRANGE)) {
String rangeStr = condition.substring(OPERATOR_NRANGE.length() + 1, condition.length() - 1);
String[] range = rangeStr.split(",");
if (range.length == 2) {
try {
Comparable<?> originalVal = getComparable(original);
Comparable<?> lower = getComparable(range[0]);
Comparable<?> upper = getComparable(range[1]);
if (originalVal.getClass().equals(lower.getClass()) && originalVal.getClass().equals(upper.getClass())) {
@SuppressWarnings("unchecked")
int lowerCompare = ((Comparable<Object>) originalVal).compareTo(lower);
@SuppressWarnings("unchecked")
int upperCompare = ((Comparable<Object>) originalVal).compareTo(upper);
return lowerCompare >= 0 && upperCompare <= 0;
}
} catch (Exception e) {
return true;
}
}
return true;
} else if (condition.startsWith(OPERATOR_IN)) {
String inStr = condition.substring(OPERATOR_IN.length() + 1, condition.length() - 1);
List<String> values = Arrays.asList(inStr.split(","));
return !values.contains(original);
} else if (condition.startsWith(OPERATOR_NIN)) {
String ninStr = condition.substring(OPERATOR_NIN.length() + 1, condition.length() - 1);
List<String> values = Arrays.asList(ninStr.split(","));
return values.contains(original);
} else if (condition.startsWith(OPERATOR_REGEX)) {
String patternStr = condition.substring(OPERATOR_REGEX.length() + 1, condition.length() - 1);
Pattern pattern = Pattern.compile(patternStr);
return !pattern.matcher(original).find();
}
return true;
}
private Comparable<?> getComparable(String value) {
try {
return Integer.parseInt(value);
} catch (NumberFormatException e) {
try {
return Double.parseDouble(value);
} catch (NumberFormatException ex) {
return value;
}
}
}
//
public static void main(String[] args) {
MultiExpUDF udf = new MultiExpUDF();
// 测试 普通字符串
boolean result1 = udf.evaluate( "1", null);
System.out.println("非 map 或 json 字符串测试结果: " + result1);
}
//        // 测试 含正则
//        boolean result2 = udf.evaluate("abc", "=abc", "aabcdef", "regex(cd.*)");
//        System.out.println("含正则输入测试结果: " + result2);
//
//        // 测试 map 类型
//        Map<String, String> map = new HashMap<>();
//        map.put("key1", "abc");
//        map.put("key2", "efg");
//        map.put("key3", "ab");
//        boolean result3 = udf.evaluate(map, "{\"key1\":\"=abc\",\"key2\":\"!=bcd\",\"key3\":\"in(ab,b)\"}");
//        System.out.println("map 类型测试结果: " + result3);
//
//        // 测试 json 格式字符串测试
//        String jsonStr = "{\"key1\":\"abc\",\"key2\":\"efg\",\"key3\":\"5\"}";
//        boolean result4 = udf.evaluate(jsonStr, "{\"key1\":\"=abc\",\"key2\":\"!=bcd\",\"key3\":\"range[5,10]\"}");
//        System.out.println("json 格式字符串测试结果: " + result4);
//
//        // 测试 map 中 key 为空的情况
//        Map<String, String> map2 = new HashMap<>();
//        map2.put("key1", "abc");
//        map2.put("key2", null);
//        boolean result6 = udf.evaluate(map2, "{\"key1\":\"=abc\",\"key2\":\"=\"}");
//        System.out.println("测试 map 中 key 为空的情况: " + result6);
//
//        // 测试 json 中 key 为空的情况
//        String jsonStr2 = "{\"key1\":\"abc\",\"key2\":null}";
//        boolean result7 = udf.evaluate(jsonStr2, "{\"key1\":\"=abc\",\"key3\":\"=\"}");
//        System.out.println("测试 json 中 key 为空的情况: " + result7);
//
//        // 测试 not in 功能
//        boolean result8 = udf.evaluate("abc1", "nin(abc,def,ghi)");
//        System.out.println("测试 not in 功能结果: " + result8);
//
//        // 测试 not between 功能
//        boolean result9 = udf.evaluate("3", "nrange[4,6]");
//        System.out.println("测试 not between 功能结果: " + result9);
//    }
}