def dataframe_write_to_hdfs(hdfs_path, dataframe):
"""
:param client:
:param hdfs_path:
:param dataframe:
:return:
"""
HDFSUrl = "http://192.168.0.201:50070"
client = Client(HDFSUrl, root='/')
client.write(hdfs_path, dataframe.to_csv(header=False,index=False,sep="\t"), encoding='utf-8',overwrite=True)
def save_page_hdfs(ipPort, file_path, contents):
"""保存网页源码到hdfs
:param ipPort: hdfs连接地址
:param file_path: 文件路径
:param contents: 网页内容
:return: None
"""
client = Client(ipPort)
with client.write(file_path) as writer:
writer.write(bytes(contents, encoding='utf8'))
def conv_file_hdfs(from_file: str, to_file: str, client: Client):
if not os.path.isfile(from_file):
return
# newpath = os.path.dirname(to_file)
# pathlib.Path(newpath).mkdir(parents=True, exist_ok=True)
MyHdfsFile.safe_make_dir(client, to_file)
f1 = open(from_file, 'r', encoding="gb18030")
# f2 = open(to_file, 'w', encoding="utf-8")
data = ' '
i = 0
while len(data) > 0:
data = f1.readline()
print(data + "\n\n")
print(i)
if i > 0:
# client.write(hdfs_path, data, overwrite=True, append=False) # first line
# client.write(hdfs_path, data, overwrite=False, append=True) # rest line
client.write(to_file, data.encode('utf-8'), overwrite=(i == 1), append=(not (i == 1)))
print("\nWrite")
i = i + 1
client.set_permission(to_file, 777)
f1.close()
print("ok")
class HdfsClient(object):
def __init__(self, url=None):
self.url = url
self.client = Client(url=url)
def ls(self, path):
return self.client.list(path)
def isFile(self, path):
result = self.client.status(path, strict=False)
if result:
return result[TYPE] == FILE
else:
return False
def mkdir(self, path):
self.client.makedirs(path, permission=777)
def isDirectory(self, path):
result = self.client.status(path, strict=False)
if result:
return result[TYPE] == DIRECTORY
else:
return False
def upload(self, localSourcePath, remoteDistPath):
self.client.upload(remoteDistPath, localSourcePath, overwrite=True)
def dowload(self, remoteSourcePath, localDistPath):
self.client.download(remoteSourcePath, localDistPath, overwrite=True)
def put(self, localSourcePath, remoteDistPath):
with open(localSourcePath,
"r") as reader, self.client.write(remoteDistPath) as writer:
data = reader.read(FILE_SIZE)
while data != "":
writer.write(data)
data = reader.read(FILE_SIZE)
def get(self, remoteSourcePath, localDistPath):
with self.client.read(remoteSourcePath,
chunk_size=FILE_SIZE) as reader, open(
localDistPath, "a+") as writer:
for chunk in reader:
writer.write(chunk)
producer = KafkaProducer(bootstrap_servers='116.56.136.57:9092')
def on_send_success(record_metadata):
print('Send successfully!!!!')
print('topic: ', record_metadata.topic)
print('partition: ', record_metadata.partition)
print('offset: ', record_metadata.offset)
def on_send_error(excp):
log.error('I am an errback', exc_info=excp)
# handle exception
nowtime = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
print(nowtime)
while (1):
if (nowtime != datetime.now().strftime("%Y-%m-%d %H:%M:%S")):
nowtime = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
rate = random.uniform(0, 100)
for i in range(0, user_per_second):
data = bytes(nowtime + ";" +
"{:0>8d}".format(random.randint(0, 1000000)) + ";" +
str(random_index([rate, 100 - rate])))
producer.send(
'dzg',
data).add_callback(on_send_success).add_errback(on_send_error)
client.write(hdfs_path, data, overwrite=False, append=True)
class HDFSClient:
def __init__(self,
url,
root=None,
user=None,
proxy=None,
timeout=None,
session=None):
""" 连接hdfs
url: HDFS名称节点的主机名或IP地址及端口号
root: 根路径，此路径将作为传递给客户端的所有HDFS路径的前缀
user: 使用InsecureClient（Base Client）,指定访问hdfs的用户;Client使用默认用户dr.who
proxy: 代理的用户
timeout: 连接超时，转发到请求处理程序
session: request.Session实例，用于发出所有请求
"""
if user:
self.client = InsecureClient(url, user=user)
else:
self.client = Client(url,
root=root,
proxy=proxy,
timeout=timeout,
session=session)
def list_hdfs_file(self, hdfs_path, status=False):
""" 返回目录下的文件
status: 每个文件或目录的属性信息(FileStatus)
return: 列表中包含元组，每个元组是目录名或文件名和属性信息构成
"""
return self.client.list(hdfs_path, status=status)
def walk_hdfs_file(self,
hdfs_path,
depth=0,
status=False,
ignore_missing=False,
allow_dir_changes=False):
""" 深度遍历远程文件系统
hdfs_path: 起始路径。如果该路径不存在，则会引发HdfsError。如果指向文件，则返回的生成器将为空
depth: 探索的最大深度。0为无限制
status: 同时返回每个文件或文件夹的相应FileStatus
ignore_missing: 忽略缺少的嵌套文件夹，而不是引发异常
allow_dir_changes: 允许更改目录列表以影响遍历
return: 生成器，返回值参考python的walk函数
"""
return self.client.walk(hdfs_path,
depth=depth,
status=status,
ignore_missing=ignore_missing,
allow_dir_changes=allow_dir_changes)
def delete_hdfs_file(self, hdfs_path, recursive=False, skip_trash=False):
""" 删除文件
recursive: 递归删除文件或目录，默认情况下，如果尝试删除非空目录，此方法将引发HdfsError
skip_trash: 设置为false时，已删除的路径将被移动到适当的垃圾回收文件夹，而不是被删除
return: 如果删除成功，则此函数返回True；如果hdfs_path之前不存在文件或目录，则返回False
"""
return self.client.delete(hdfs_path,
recursive=recursive,
skip_trash=skip_trash)
def download_hdfs_file(self,
hdfs_path,
local_path,
overwrite=True,
n_threads=1,
temp_dir=None,
**kwargs):
""" 下载文件
hdfs_file: HDFS上要下载的文件或文件夹的路径。如果是文件夹，则将下载该文件夹下的所有文件
local_file: 本地路径。如果它已经存在并且是目录，则文件将在其中下载
overwrite: 覆盖任何现有文件或目录
n_threads: 用于并行化的线程数。值为0（或负数）将使用与文件一样多的线程
temp_dir: 当overwrite = True并且最终目标路径已经存在时，将首先在其下下载文件的目录。下载成功完成后，它将被交换
**kwargs: 关键字参数转发给read()。如果未传递chunk_size参数，则将使用默认值64 kB
return: 方法执行成功，将返回本地下载路径
"""
res = self.client.download(hdfs_path,
local_path,
overwrite=overwrite,
n_threads=n_threads,
temp_dir=temp_dir,
**kwargs)
def upload_hdfs_file(self,
hdfs_path,
local_path,
n_threads=1,
temp_dir=None,
chunk_size=65536,
progress=None,
cleanup=True,
**kwargs):
""" 上传文件
hdfs_path: 目标HDFS路径。如果它已经存在并且是目录，则文件将在其中上传
local_path: 文件或文件夹的本地路径。如果是文件夹，则将上载其中的所有
文件（请注意，这意味着没有文件的文件夹将不会远程创建）
cleanup: 如果上传过程中发生错误，删除所有上传的文件
return: 方法执行成功，将返回状态码，远程上传目录，错误信息
"""
try:
res = self.client.upload(hdfs_path,
local_path,
n_threads=n_threads,
temp_dir=temp_dir,
chunk_size=chunk_size,
progress=progress,
cleanup=cleanup,
overwrite=True)
return 0, res, ''
except HdfsError as e:
return 1, '', str(e)
def makedirs(self, hdfs_path, permission=None):
""" 创建目录，可以递归
permission: 在新创建的目录上设置的八进制权限，这些权限将仅在尚不存在的目录上设置
return: None
"""
self.client.makedirs(hdfs_path, permission=permission)
def parts(self, hdfs_path, parts=None, status=False):
"""
hdfs_path: 远程路径。该目录每个分区最多应包含一个零件文件（否则将任意选择一个文件）
parts: 零件文件编号列表或要选择的零件文件总数。如果是数字，那么将随机选择那么多分区。
默认情况下，将返回所有零件文件。如果部件是列表，但未找到部件之一或需要太多样本，则会引发HdfsError
status: 返回文件的FileStatus
return: 返回对应于路径的零件文件的字典
"""
return self.client.parts(hdfs_path, parts=parts, status=status)
def read_hdfs_file(self, **kwds):
""" 读取文件内容，这个方法必须在一个with块中使用，以便关闭连接
>>> with client.read('foo') as reader:
>>> content = reader.read()
hdfs_path: HDFS路径
offset: 起始字节位置
length: 要处理的字节数。设置为None时会读取整个文件
buffer_size: 用于传输数据的缓冲区大小（以字节为单位）。默认为在HDFS配置中设置的值
encoding: 用于解码请求的编码。默认情况下，返回原始数据
chunk_size: 如果设置为正数，则上下文管理器将返回一个生成器，该生成器生成每个chunk_size字节，
而不是类似文件的对象（除非还设置了定界符）
delimiter: 如果设置，上下文管理器将在每次遇到定界符时返回生成器。此参数要求指定编码
progress: 回调函数，用于跟踪进度，称为每个chunk_size字节（如果未指定块大小，则不可用）。
它将传递两个参数，即要上传的文件的路径和到目前为止已传输的字节数。
完成后，将以-1作为第二个参数调用一次
"""
self.client.read(**kwds)
def write_hdfs_file(self,
hdfs_path,
data=None,
overwrite=False,
permission=None,
blocksize=None,
replication=None,
buffersize=None,
append=False,
encoding=None):
""" 在HDFS上创建文件
data: 要写入的文件内容。 可以是字符串，生成器或文件对象。 最后两个选项将允许流式上传（即无需
将全部内容加载到内存中）。 如果为None，则此方法将返回类似文件的对象，应使用with块调用
它（请参见下面的示例）
permission: 在新创建的文件上设置的八进制权限
append: 附加到文件而不是创建新文件
encoding: 用于序列化写入数据的编码
>>> from json import dump, dumps
>>> records = [
>>> {'name': 'foo', 'weight': 1},
>>> {'name': 'bar', 'weight': 2},
>>> ]
>>> # As a context manager:
>>> with client.write('data/records.jsonl', encoding='utf-8') as writer:
>>> dump(records, writer)
>>> Or, passing in a generator directly:
>>> client.write('data/records.jsonl', data=dumps(records), encoding='utf-8')
"""
self.client.write(hdfs_path,
data=data,
overwrite=overwrite,
permission=permission,
blocksize=blocksize,
replication=replication,
buffersize=buffersize,
append=append,
encoding=encoding)
def rename_or_move(self, hdfs_src_path, hdfs_dst_path):
""" 移动文件或目录
hdfs_src_path: 源路径
hdfs_dst_path: 目标路径，如果路径已经存在并且是目录，则源将移入其中。
如果路径存在并且是文件，或者缺少父目标目录，则此方法将引发HdfsError
"""
self.client.rename(hdfs_src_path, hdfs_dst_path)
def set_owner(self, hdfs_path, owner=None, group=None):
""" 更改文件的所有者，必须至少指定所有者和组之一
owner: 可选，文件的新所有者
group: 可选，文件的新所有组
"""
self.client.set_owner(hdfs_path, owner=owner, group=group)
def set_permission(self, hdfs_path, permission):
""" 更改文件权限
permission: 文件的新八进制权限字符串
"""
self.client.set_permission(hdfs_path, permission)
def set_replication(self, hdfs_path, replication):
""" 设置文件副本
replication: 副本数
"""
self.client.set_replication(hdfs_path, replication)
def set_times(self, hdfs_path, access_time=None, modification_time=None):
""" 更改文件的时间戳
"""
self.client.set_times(hdfs_path,
access_time=access_time,
modification_time=modification_time)
def status_hdfs_file(self, hdfs_path, strict=True):
""" 获取文件的FileStatus
strict: 如果为False，则返回None，而不是如果路径不存在则引发异常
"""
self.client.status(hdfs_path, strict=strict)
def Create_file(self, p_filetype, p_filename, p_formula_str, p_rows, p_encoding='UTF-8'):
try:
m_output = None
m_HDFS_Handler = None
if p_filetype.upper() == "MEM":
m_fs = fs.open_fs('mem://')
m_filename = p_filename
m_output = m_fs.open(path=m_filename, mode='w', encoding=p_encoding)
elif p_filetype.upper() == "FS":
m_filename = p_filename
m_output = open(file=m_filename, mode='w', encoding=p_encoding)
elif p_filetype.upper() == "HDFS":
# HDFS 文件格式： http://node:port/xx/yy/cc.dat
# 注意这里的node和port都是webfs端口，不是rpc端口
m_Protocal = p_filename.split("://")[0]
m_NodePort = p_filename[len(m_Protocal) + 3:].split("/")[0]
m_WebFSURL = m_Protocal + "://" + m_NodePort
m_WebFSDir, m_filename = os.path.split(p_filename[len(m_WebFSURL):])
if self.c_HDFS_ConnectedUser is None:
m_HDFS_Handler = Client(url=m_WebFSURL,
root=m_WebFSDir,
proxy=None,
session=None)
else:
m_HDFS_Handler = InsecureClient(url=m_WebFSURL,
user=self.c_HDFS_ConnectedUser,
root=m_WebFSDir)
else:
raise SQLCliException("Unknown file format.")
m_row_struct = self.parse_formula_str(p_formula_str)
buf = []
if p_filetype.upper() == "HDFS": # 处理HDFS文件写入
# 总是覆盖服务器上的文件, 每10W行提交一次服务器文件, 以避免内存的OOM问题
if p_rows < 100000:
with m_HDFS_Handler.write(hdfs_path=m_filename, overwrite=True) as m_output:
for i in range(0, p_rows):
m_output.write((self.get_final_string(m_row_struct) + "\n").encode())
else:
for i in range(0, p_rows // 100000):
if i == 0:
with m_HDFS_Handler.write(hdfs_path=m_filename, overwrite=True) as m_output:
for j in range(0, 100000):
m_output.write((self.get_final_string(m_row_struct) + "\n").encode())
else:
with m_HDFS_Handler.write(hdfs_path=m_filename, append=True) as m_output:
for j in range(0, 100000):
m_output.write((self.get_final_string(m_row_struct) + "\n").encode())
with m_HDFS_Handler.write(hdfs_path=m_filename, append=True) as m_output:
for i in range(0, p_rows % 100000):
m_output.write((self.get_final_string(m_row_struct) + "\n").encode())
else: # 处理普通文件写入
for i in range(0, p_rows):
buf.append(self.get_final_string(m_row_struct) + '\n')
if len(buf) == 100000: # 为了提高IO效率，每10W条写入文件一次
m_output.writelines(buf)
buf = []
if len(buf) != 0:
m_output.writelines(buf)
m_output.close()
# 重置identity的序列号，保证下次从开头开始
if hasattr(identity, 'x'):
delattr(identity, 'x')
if hasattr(identity_timestamp, 'x'):
delattr(identity_timestamp, 'x')
except SQLCliException as e:
raise SQLCliException(e.message)
except Exception as e:
if "SQLCLI_DEBUG" in os.environ:
print('traceback.print_exc():\n%s' % traceback.print_exc())
print('traceback.format_exc():\n%s' % traceback.format_exc())
raise SQLCliException(repr(e))
def Convert_file(self, p_srcfileType, p_srcfilename, p_dstfileType, p_dstfilename):
try:
if p_srcfileType.upper() == "MEM":
m_srcFSType = "MEM"
if self.g_MemoryFSHandler is None:
self.g_MemoryFSHandler = fs.open_fs('mem://')
m_srcFS = self.g_MemoryFSHandler
else:
m_srcFS = self.g_MemoryFSHandler
m_srcFileName = p_srcfilename
elif p_srcfileType.upper() == "FS":
m_srcFSType = "FS"
m_srcFS = fs.open_fs('./')
m_srcFileName = p_srcfilename
elif p_srcfileType.upper() == "HDFS":
m_srcFSType = "HDFS"
m_srcFullFileName = p_srcfilename
m_Protocal = m_srcFullFileName.split("://")[0]
m_NodePort = m_srcFullFileName[len(m_Protocal) + 3:].split("/")[0]
m_WebFSURL = m_Protocal + "://" + m_NodePort
m_WebFSDir, m_srcFileName = os.path.split(m_srcFullFileName[len(m_WebFSURL):])
m_srcFS = Client(m_WebFSURL, m_WebFSDir, proxy=None, session=None)
else:
m_srcFS = None
m_srcFileName = None
m_srcFSType = "Not Supported"
if p_dstfileType.upper() == "MEM":
m_dstFSType = "MEM"
if self.g_MemoryFSHandler is None:
self.g_MemoryFSHandler = fs.open_fs('mem://')
m_dstFS = self.g_MemoryFSHandler
else:
m_dstFS = self.g_MemoryFSHandler
m_dstFileName = p_dstfilename
elif p_dstfileType.upper() == "FS":
m_dstFSType = "FS"
m_dstFS = fs.open_fs('./')
m_dstFileName = p_dstfilename
elif p_dstfileType.upper() == "HDFS":
m_dstFSType = "HDFS"
m_dstFullFileName = p_dstfilename
m_Protocal = m_dstFullFileName.split("://")[0]
m_NodePort = m_dstFullFileName[len(m_Protocal) + 3:].split("/")[0]
m_WebFSURL = m_Protocal + "://" + m_NodePort
m_WebFSDir, m_dstFileName = os.path.split(m_dstFullFileName[len(m_WebFSURL):])
m_dstFS = Client(m_WebFSURL, m_WebFSDir, proxy=None, session=None)
else:
m_dstFS = None
m_dstFileName = None
m_dstFSType = "Not Supported convert."
if m_srcFSType == "Not Supported" or m_dstFSType == "Not Supported":
raise SQLCliException("Not supported convert. From [" + p_srcfileType + "] to [" + p_dstfileType + "]")
if m_srcFSType in ('MEM', 'FS') and m_dstFSType in ('MEM', 'FS'):
with m_srcFS.openbin(m_srcFileName, "r") as m_reader, m_dstFS.openbin(m_dstFileName, "w") as m_writer:
while True:
m_Contents = m_reader.read(8192)
if len(m_Contents) == 0:
break
m_writer.write(m_Contents)
if m_srcFSType == "HDFS" and m_dstFSType in ('MEM', 'FS'):
with m_srcFS.read(m_srcFileName, "rb") as m_reader, m_dstFS.openbin(m_dstFileName, "w") as m_writer:
while True:
m_Contents = m_reader.read(8192)
if len(m_Contents) == 0:
break
m_writer.write(m_Contents)
# 对于HDFS的写入，每80M提交一次，以避免内存的OOM问题
if m_srcFSType in ('MEM', 'FS') and m_dstFSType == "HDFS":
bHeaderWrite = True
with m_srcFS.openbin(m_srcFileName, "r") as m_reader:
while True:
m_Contents = m_reader.read(8192 * 10240)
if len(m_Contents) == 0:
break
if bHeaderWrite:
with m_dstFS.write(m_dstFileName, overwrite=True) as m_writer:
m_writer.write(m_Contents)
bHeaderWrite = False
else:
with m_dstFS.write(m_dstFileName, append=True) as m_writer:
m_writer.write(m_Contents)
if m_srcFSType == "HDFS" and m_dstFSType == "HDFS":
bHeaderWrite = True
with m_srcFS.read(m_srcFileName) as m_reader:
while True:
m_Contents = m_reader.read(8192 * 10240)
if len(m_Contents) == 0:
break
if bHeaderWrite:
with m_dstFS.write(m_dstFileName, overwrite=True) as m_writer:
m_writer.write(m_Contents)
bHeaderWrite = False
else:
with m_dstFS.write(m_dstFileName, append=True) as m_writer:
m_writer.write(m_Contents)
except HdfsError as he:
# HDFS 会打印整个堆栈信息，所以这里默认只打印第一行的信息
if "SQLCLI_DEBUG" in os.environ:
raise SQLCliException(he.message)
else:
raise SQLCliException(he.message.split('\n')[0])
