from hdfs import Client
import pandas as pd
 
HDFSHOST = "http://xxx:50070"
FILENAME = "/tmp/preprocess/part-00000" #hdfs文件路徑
COLUMNNAMES = [xx']
 
def readHDFS():
'''
讀取hdfs文件
 
Returns：
df:dataframe hdfs數(shù)據(jù)
'''
client = Client(HDFSHOST)
# 目前讀取hdfs文件采用方式：
# 1. 先從hdfs讀取二進制數(shù)據(jù)流文件
# 2. 將二進制文件另存為.csv
# 3. 使用pandas讀取csv文件
with client.read(FILENAME) as fs:
content = fs.read()
s = str(content, 'utf-8')
file = open("data/tmp/data.csv", "w")
file.write(s)
df = pd.read_csv("data/tmp/data.csv", names=COLUMNNAMES)
return df

from pyhdfs import HdfsClient
client=HdfsClient(hosts='ghym:50070')#hdfs地址
res=client.open('/sy.txt')#hdfs文件路徑,根目錄/
for r in res:
  line=str(r,encoding='utf8')#open后是二進制,str()轉(zhuǎn)換為字符串并轉(zhuǎn)碼
  print(line)

from pyhdfs import HdfsClient
client=HdfsClient(hosts='ghym:50070',user_name='hadoop')#只有hadoop用戶擁有寫權(quán)限
str='hello world'
client.create('/py.txt',str)#創(chuàng)建新文件并寫入字符串

from pyhdfs import HdfsClient
client = HdfsClient(hosts='ghym:50070', user_name='hadoop')
client.copy_from_local('d:/pydemo.txt', '/pydemo')#本地文件絕對路徑,HDFS目錄必須不存在

# pyhdfs讀取文本文件,分隔符為逗號,
from pyhdfs import HdfsClient
client = HdfsClient(hosts='ghym:50070', user_name='hadoop')
inputfile=client.open('/int.txt')
# pandas調(diào)用讀取方法read_table
import pandas as pd
df=pd.read_table(inputfile,encoding='gbk',sep=',')#參數(shù)為源文件,編碼,分隔符
# 數(shù)據(jù)集to_csv方法轉(zhuǎn)換為csv
df.to_csv('demo.csv',encoding='gbk',index=None)#參數(shù)為目標(biāo)文件,編碼,是否要索引