Skip to content

jingw/pyhdfs

Repository files navigation

Python HDFS client

Because the world needs yet another way to talk to HDFS from Python.

Usage

This library provides a Python client for WebHDFS. NameNode HA is supported by passing in both NameNodes. Responses are returned as nice Python classes, and any failed operation will raise some subclass of HdfsException matching the Java exception.

Example usage:

>>> fs = pyhdfs.HdfsClient(hosts='nn1.example.com:50070,nn2.example.com:50070', user_name='someone')
>>> fs.list_status('/')
[FileStatus(pathSuffix='benchmarks', permission='777', type='DIRECTORY', ...), FileStatus(...), ...]
>>> fs.listdir('/')
['benchmarks', 'hbase', 'solr', 'tmp', 'user', 'var']
>>> fs.mkdirs('/fruit/x/y')
True
>>> fs.create('/fruit/apple', 'delicious')
>>> fs.append('/fruit/apple', ' food')
>>> with contextlib.closing(fs.open('/fruit/apple')) as f:
...     f.read()
...
b'delicious food'
>>> fs.get_file_status('/fruit/apple')
FileStatus(length=14, owner='someone', type='FILE', ...)
>>> fs.get_file_status('/fruit/apple').owner
'someone'
>>> fs.get_content_summary('/fruit')
ContentSummary(directoryCount=3, fileCount=1, length=14, quota=-1, spaceConsumed=14, spaceQuota=-1)
>>> list(fs.walk('/fruit'))
[('/fruit', ['x'], ['apple']), ('/fruit/x', ['y'], []), ('/fruit/x/y', [], [])]
>>> fs.exists('/fruit/apple')
True
>>> fs.delete('/fruit')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File ".../pyhdfs.py", line 525, in delete
  ...
pyhdfs.HdfsPathIsNotEmptyDirectoryException: `/fruit is non empty': Directory is not empty
>>> fs.delete('/fruit', recursive=True)
True
>>> fs.exists('/fruit/apple')
False
>>> issubclass(pyhdfs.HdfsFileNotFoundException, pyhdfs.HdfsIOException)
True

The methods and return values generally map directly to WebHDFS endpoints. The client also provides convenience methods that mimic Python os methods and HDFS CLI commands (e.g. walk and copy_to_local).

pyhdfs logs all HDFS actions at the INFO level, so turning on INFO level logging will give you a debug record for your application.

For more information, see the full API docs.

Installing

pip install pyhdfs

Python 3 is required.

Development testing

http://codecov.io/github/jingw/pyhdfs/coverage.svg?branch=master Documentation Status

First run install-hdfs.sh x.y.z, which will download, extract, and run the HDFS NN/DN processes in the current directory. (Replace x.y.z with a real version.) Then run the following commands. Note they will create and delete hdfs://localhost/tmp/pyhdfs_test.

Commands:

python3 -m venv env
source env/bin/activate
pip install -e .
pip install -r dev_requirements.txt
pytest