"""A file interface for handling local and remote data files.
The goal of datasource is to abstract some of the file system operations
when dealing with data files so the researcher doesn't have to know all the
low-level details. Through datasource, a researcher can obtain and use a
file with one function call, regardless of location of the file.
DataSource is meant to augment standard python libraries, not replace them.
It should work seamlessly with standard file IO operations and the os
module.
DataSource files can originate locally or remotely:
- local files : '/home/guido/src/local/data.txt'
- URLs (http, ftp, ...) : 'http://www.scipy.org/not/real/data.txt'
DataSource files can also be compressed or uncompressed. Currently only
gzip, bz2 and xz are supported.
Example::
>>> # Create a DataSource, use os.curdir (default) for local storage.
>>> from numpy import DataSource
>>> ds = DataSource()
>>>
>>> # Open a remote file.
>>> # DataSource downloads the file, stores it locally in:
>>> # './www.google.com/index.html'
>>> # opens the file and returns a file object.
>>> fp = ds.open('http://www.google.com/') # doctest: +SKIP
>>>
>>> # Use the file as you normally would
>>> fp.read() # doctest: +SKIP
>>> fp.close() # doctest: +SKIP
"""
import os
import shutil
import io
from contextlib import closing
from numpy.core.overrides import set_module
_open = open
def _check_mode(mode, encoding, newline):
"""Check mode and that encoding and newline are compatible.
Parameters
----------
mode : str
File open mode.
encoding : str
File encoding.
newline : str
Newline for text files.
"""
if "t" in mode:
if "b" in mode:
raise ValueError("Invalid mode: %r" % (mode,))
else:
if encoding is not None:
raise ValueError("Argument 'encoding' not supported in binary mode")
if newline is not None:
raise ValueError("Argument 'newline' not supported in binary mode")
# Using a class instead of a module-level dictionary
# to reduce the initial 'import numpy' overhead by
# deferring the import of lzma, bz2 and gzip until needed
# TODO: .zip support, .tar support?
class _FileOpeners:
"""
Container for different methods to open (un-)compressed files.
`_FileOpeners` contains a dictionary that holds one method for each
supported file format. Attribute lookup is implemented in such a way
that an instance of `_FileOpeners` itself can be indexed with the keys
of that dictionary. Currently uncompressed files as well as files
compressed with ``gzip``, ``bz2`` or ``xz`` compression are supported.
Notes
-----
`_file_openers`, an instance of `_FileOpeners`, is made available for
use in the `_datasource` module.
Examples
--------
>>> import gzip
>>> np.lib._datasource._file_openers.keys()
[None, '.bz2', '.gz', '.xz', '.lzma']
>>> np.lib._datasource._file_openers['.gz'] is gzip.open
True
"""
def __init__(self):
self._loaded = False
self._file_openers = {None: io.open}
def _load(self):
if self._loaded:
return
try:
import bz2
self._file_openers[".bz2"] = bz2.open
except ImportError:
pass
try:
import gzip
self._file_openers[".gz"] = gzip.open
except ImportError:
pass
try:
import lzma
self._file_openers[".xz"] = lzma.open
self._file_openers[".lzma"] = lzma.open
except (ImportError, AttributeError):
# There are incompatible backports of lzma that do not have the
# lzma.open attribute, so catch that as well as ImportError.
pass
self._loaded = True
def keys(self):
"""
Return the keys of currently supported file openers.
Parameters
----------
None
Returns
-------
keys : list
The keys are None for uncompressed files and the file extension
strings (i.e. ``'.gz'``, ``'.xz'``) for supported compression
methods.
"""
self._load()
return list(self._file_openers.keys())
def __getitem__(self, key):
self._load()
return self._file_openers[key]
_file_openers = _FileOpeners()
def open(path, mode='r', destpath=os.curdir, encoding=None, newline=None):
"""
Open `path` with `mode` and return the file object.
If ``path`` is an URL, it will be downloaded, stored in the
`DataSource` `destpath` directory and opened from there.
Parameters
----------
path : str
Local file path or URL to open.
mode : str, optional
Mode to open `path`. Mode 'r' for reading, 'w' for writing, 'a' to
append. Available modes depend on the type of object specified by
path. Default is 'r'.
destpath : str, optional
Path to the directory where the source file gets downloaded to for
use. If `destpath` is None, a temporary directory will be created.
The default path is the current directory.
encoding : {None, str}, optional
Open text file with given encoding. The default encoding will be
what `io.open` uses.
newline : {None, str}, optional
Newline to use when reading text file.
Returns
-------
out : file object
The opened file.
Notes
-----
This is a convenience function that instantiates a `DataSource` and
returns the file object from ``DataSource.open(path)``.
"""
ds = DataSource(destpath)
return ds.open(path, mode, encoding=encoding, newline=newline)
@set_module('numpy')
class DataSource:
"""
DataSource(destpath='.')
A generic data source file (file, http, ftp, ...).
DataSources can be local files or remote files/URLs. The files may
also be compressed or uncompressed. DataSource hides some of the
low-level details of downloading the file, allowing you to simply pass
in a valid file path (or URL) and obtain a file object.
Parameters
----------
destpath : str or None, optional
Path to the directory where the source file gets downloaded to for
use. If `destpath` is None, a temporary directory will be created.
The default path is the current directory.
Notes
-----
URLs require a scheme string (``http://``) to be used, without it they
will fail::
>>> repos = np.DataSource()
>>> repos.exists('www.google.com/index.html')
False
>>> repos.exists('http://www.google.com/index.html')
True
Temporary directories are deleted when the DataSource is deleted.
Examples
--------
::
>>> ds = np.DataSource('/home/guido')
>>> urlname = 'http://www.google.com/'
>>> gfile = ds.open('http://www.google.com/')
>>> ds.abspath(urlname)
'/home/guido/www.google.com/index.html'
>>> ds = np.DataSource(None) # use with temporary file
>>> ds.open('/home/guido/foobar.txt')
<open file '/home/guido.foobar.txt', mode 'r' at 0x91d4430>
>>> ds.abspath('/home/guido/foobar.txt')
'/tmp/.../home/guido/foobar.txt'
"""
def __init__(self, destpath=os.curdir):
"""Create a DataSource with a local path at destpath."""
if destpath:
self._destpath = os.path.abspath(destpath)
self._istmpdest = False
else:
import tempfile # deferring import to improve startup time
self._destpath = tempfile.mkdtemp()
self._istmpdest = True
def __del__(self):
# Remove temp directories
if hasattr(self, '_istmpdest') and self._istmpdest:
shutil.rmtree(self._destpath)
def _iszip(self, filename):
"""Test if the filename is a zip file by looking at the file extension.
"""
fname, ext = os.path.splitext(filename)
return ext in _file_openers.keys()
def _iswritemode(self, mode):
"""Test if the given mode will open a file for writing."""
# Currently only used to test the bz2 files.
_writemodes = ("w", "+")
for c in mode:
if c in _writemodes:
return True
return False
def _splitzipext(self, filename):
"""Split zip extension from filename and return filename.
*Returns*:
base, zip_ext : {tuple}
"""
if self._iszip(filename):
return os.path.splitext(filename)
else:
return filename, None
def _possible_names(self, filename):
"""Return a tuple containing compressed filename variations."""
names = [filename]
if not self._iszip(filename):
for zipext in _file_openers.keys():
if zipext:
names.append(filename+zipext)
return names
def _isurl(self, path):
"""Test if path is a net location. Tests the scheme and netloc."""
# We do this here to reduce the 'import numpy' initial import time.
from urllib.parse import urlparse
# BUG : URLs require a scheme string ('http://') to be used.
# www.google.com will fail.
# Should we prepend the scheme for those that don't have it and
# test that also? Similar to the way we append .gz and test for
# for compressed versions of files.
scheme, netloc, upath, uparams, uquery, ufrag = urlparse(path)
return bool(scheme and netloc)
def _cache(self, path):
"""Cache the file specified by path.
Creates a copy of the file in the datasource cache.
"""
# We import these here because importing urllib is slow and
# a significant fraction of numpy's total import time.
from urllib.request import urlopen
from urllib.error import URLError
upath = self.abspath(path)
# ensure directory exists
if not os.path.exists(os.path.dirname(upath)):
os.makedirs(os.path.dirname(upath))
# TODO: Doesn't handle compressed files!
if self._isurl(path):
try:
with closing(urlopen(path)) as openedurl:
with _open(upath, 'wb') as f:
shutil.copyfileobj(openedurl, f)
except URLError:
raise URLError("URL not found: %s" % path)
else:
shutil.copyfile(path, upath)
return upath
Loading ...