Apache nifi 使用ApacheNIFI的复杂转换和过滤器
我有一个JSON数组:Apache nifi 使用ApacheNIFI的复杂转换和过滤器,apache-nifi,Apache Nifi,我有一个JSON数组: [ { "account_login" : "some_mail@gmail.com", "view_id" : 11313231, "join_id" : "utm_campaign=toyota&utm_content=multiformat_sites&utm_medium=cpc&utm_source=mytarget", &
[ {
"account_login" : "some_mail@gmail.com",
"view_id" : 11313231,
"join_id" : "utm_campaign=toyota&utm_content=multiformat_sites&utm_medium=cpc&utm_source=mytarget",
"start_date" : "2020-08-01",
"end_date" : "2020-08-31"
}, {
"account_login" : "another_mail@lab.net",
"view_id" : 19556319183,
"join_id" : "utm_campaign=mazda&utm_content=keywords_social-networks&utm_medium=cpc&utm_source=facebook",
"start_date" : "2020-12-22",
"end_date" : "2020-12-23"
}, {
...
} ]
对于每个join\u id
我应该做下一件事:
utm_活动,丰田;utm_内容、多格式网站;etc
public class GaUtmFactoryService {
private static final String INVALID_MACRO_FOOTPRINTS = "^.*[{\\[%]+.+[}\\]%].*$";
public Map<String, String> extractUtmMarks(String utmMarks) {
if (utmMarks == null || utmMarks.isBlank()) {
return Collections.emptyMap();
}
return Arrays.stream(utmMarks.split("\\s*&\\s*"))
.map(s -> s.trim().split("\\s*=\\s*"))
.filter(this::isUtmMarksValid)
.collect(Collectors.toMap(
key -> convertCsUtmMarkToGa(key[0]),
value -> value[1],
(val1, val2) -> val2)
);
}
private boolean isUtmMarksValid(String[] utmMarks) {
return utmMarks.length == 2
&& !convertCsUtmMarkToGa(utmMarks[0]).isBlank()
&& !utmMarks[1].isBlank()
&& Arrays.stream(utmMarks).noneMatch(this::isUtmMarkContainsInvalidChars);
}
private boolean isUtmMarkContainsInvalidChars(String utmMark) {
return utmMark.matches(INVALID_MACRO_FOOTPRINTS)
|| !StandardCharsets.US_ASCII.newEncoder().canEncode(utmMark);
}
private String convertCsUtmMarkToGa(String utmMark) {
switch (utmMark) {
case "utm_medium":
return "ga:medium";
case "utm_campaign":
return "ga:campaign";
case "utm_source":
return "ga:source";
case "utm_content":
return "ga:adContent";
case "utm_term":
return "ga:keyword";
case "utm_target":
case "utm_a":
return "";
default:
return rowUtmMarks;
}
}
}
公共类GaUtmFactoryService{
私有静态最终字符串无效的_宏_封装=“^.*[{\\[%]+.+[}\\\]%.*$”;
公共地图提取utmMarks(字符串utmMarks){
if(utmMarks==null | | utmMarks.isBlank()){
return Collections.emptyMap();
}
返回Arrays.stream(utmMarks.split(“\\s*&\\s*”))
.map(s->s.trim().split(“\\s*=\\s*”)
.filter(此::IsUtmarkValid)
.collect(collector.toMap)(
键->ConvertCsutmarktoga(键[0]),
值->值[1],
(val1,val2)->val2)
);
}
私有布尔值IsUtmarks有效(字符串[]Utmarks){
返回utmMarks.length==2
&&!convertcsutmarktoga(utmMarks[0]).isBlank()
&&!utmMarks[1]。isBlank()
&&Arrays.stream(utmMarks.noneMatch)(this::isUtmMarkContainsInvalidChars);
}
私有布尔值IsUtmarkContainsInValidChars(字符串Utmark){
返回utmMark.matches(无效的\u宏\u封装外形)
||!StandardCharsets.US_ASCII.newEncoder().canEncode(utmMark);
}
私有字符串转换器UTMMarktoga(字符串utmMark){
交换机(utmMark){
案例“utm_中等”:
返回“ga:中等”;
案例“utm_活动”:
返回“ga:活动”;
案例“utm_来源”:
返回“ga:source”;
案例“utm_内容”:
返回“ga:adContent”;
案例“utm_术语”:
返回“ga:关键字”;
案例“utm_目标”:
案例“utm_a”:
返回“”;
违约:
返回rowUtmMarks;
}
}
}
来自外部的用法:
public Map<String, String> getConvertedMarks() {
GaUtmFactoryService gaUtmFactoryService = new GaUtmFactoryService();
String utmMarks = "utm_campaign=toyota&utm_content=multiformat_sites&utm_medium=cpc&utm_source=facebook";
Map<String, String> converted = gaUtmFactoryService.extractUtmMarks(utmMarks);
//should be:
////{ga:campaign=toyota, ga:adContent=multiformat_sites, ga:medium=cpc, ga:source=facebook}
return converted;
}
publicmap getConvertedMarks(){
GaUtmFactoryService GaUtmFactoryService=新的GaUtmFactoryService();
String utmMarks=“utm_活动=丰田&utm_内容=多格式网站&utm_媒体=cpc&utm_来源=facebook”;
映射转换=gaUtmFactoryService.ExtractUtmarks(Utmarks);
//应该是:
////{ga:campaign=toyota,ga:adContent=multiformat_sites,ga:medium=cpc,ga:source=facebook}
转换收益;
}
NiFi有可能吗?或者,如果这很难,也许我应该为这个任务创建带有一些端点的REST microservice
更新
我做了EvaluateJsonPath
和SplitJson
。现在每个json文件都有一个属性:utm.marks=utm\u campaign=toyota&utm\u content=multiformat\u sites&utm\u medium=cpc&utm\u source=mytarget
我需要拆分这些属性并获得如下smth:
campaign.key=ga:campaign
campaign.value=toyota
content.key=ga:content
content.value=多格式站点
等等。对于此转换,ExecuteGroovyScript可能如下所示:
import groovy.json*
//从会话获取文件
def ff=session.get()
如果(!ff)返回
//读取流、转换为读取器、解析为列表/对象
def data=ff.read().withReader(“UTF-8”){r->new JsonSlurper().parse(r)}
//转换json
每个{i->
i、 join\u id=i.join\u id
.split(“\\s*&\\s*”/#到数组
.1条{
//#将每个项目转换为映射项
字符串[]kv=it.split(“\\s*=\\s*”)
千伏[0]=[
“utm_中等”:“ga:中等”,
“utm_活动”:“ga:活动”,
“utm_源”:“ga:源”,
“utm_内容”:“ga:adContent”,
“utm_术语”:“ga:关键字”,
].get(千伏[0])
千伏
}
.findAll{k,v->k}/#过滤掉空/空键
}
//写回文件
ff.write(“UTF-8”){w->newjsonbuilder(data.writeTo(w)}
//走向成功
REL_SUCCESS基于一个JSON(非数组)答案的解决方案:
import groovy.json*
//从会话获取文件
def ff=session.get()
如果(!ff)返回
//读取流、转换为读取器、解析为列表/对象
def data=ff.read().withReader(“UTF-8”){r->new JsonSlurper().parse(r)}
def builder=新JsonBuilder(数据)
builder.content.join_id=builder.content.join_id.split(“\\s*&\\s*”)/#到数组
.1条{
//#将每个项目转换为映射项
字符串[]kv=it.split(“\\s*=\\s*”)
千伏[0]=[
“utm_中等”:“ga:中等”,
“utm_活动”:“ga:活动”,
“utm_源”:“ga:源”,
“utm_内容”:“ga:adContent”,
“utm_术语”:“ga:关键字”,
].get(千伏[0])
千伏
}
.findAll{k,v->k}/#过滤掉空/空键
ff.write(“UTF-8”){w->builder.writeTo(w)}
//走向成功
如果你有java代码,使用executegroovyscript会更容易。是的,我有同样的想法。但问题是。。我不知道groovy。你的案例应该输出什么?例如字符串utm_campaign=toyota&utm_content=multiformat_sites&utm_source=facebook应该被转移到这个地图:{key:ga:campaign,value:toyota,key:ga:adContent,value:multiformat_sites,key:ga:source,value:facebook}
你的代码不完整,没有入口点,有未初始化的变量,。。。请编辑您的问题,并提供输入和相应的输出。谢谢!我试过这个代码,然后
import groovy.json.*
//get file from session
def ff=session.get()
if(!ff)return
//read stream, convert to reader, parse to list/objects
def data=ff.read().withReader("UTF-8"){r-> new JsonSlurper().parse(r) }
def builder = new JsonBuilder(data)
builder.content.join_id = builder.content.join_id.split("\\s*&\\s*") //# to array
.collectEntries{
//# convert each item to map entry
String[] kv = it.split("\\s*=\\s*")
kv[0] = [
"utm_medium" : "ga:medium",
"utm_campaign" : "ga:campaign",
"utm_source" : "ga:source",
"utm_content" : "ga:adContent",
"utm_term" : "ga:keyword",
].get( kv[0] )
kv
}
.findAll{ k,v-> k } //# filter out empty/null keys
ff.write("UTF-8"){w-> builder.writeTo(w)}
//transfer to success
REL_SUCCESS<<ff