1
1
Fork 0
mirror of https://github.com/pypa/pip synced 2023-12-13 21:30:23 +01:00
pip/pip/util.py

481 lines
14 KiB
Python

import sys
import shutil
import os
import posixpath
import stat
import urllib
import urllib2
import re
import pkg_resources
from pip.backwardcompat import WindowsError
from pip.exceptions import InstallationError
from pip.locations import site_packages
__all__ = ['rmtree', 'display_path', 'backup_dir',
'find_command', 'splitext', 'ask', 'Inf',
'url_to_path', 'path_to_url',
'path_to_url2', 'normalize_name',
'format_size', 'is_url', 'is_installable_dir', 'is_archive_file',
'is_svn_page', 'file_contents',
'split_leading_dir', 'has_leading_dir',
'make_path_relative', 'normalize_path',
'get_file_content', 'renames', 'get_terminal_size']
def rmtree(dir):
shutil.rmtree(dir, ignore_errors=True,
onerror=rmtree_errorhandler)
def rmtree_errorhandler(func, path, exc_info):
"""On Windows, the files in .svn are read-only, so when rmtree() tries to
remove them, an exception is thrown. We catch that here, remove the
read-only attribute, and hopefully continue without problems."""
exctype, value = exc_info[:2]
# lookin for a windows error
if exctype is not WindowsError or 'Access is denied' not in str(value):
raise
# file type should currently be read only
if ((os.stat(path).st_mode & stat.S_IREAD) != stat.S_IREAD):
raise
# convert to read/write
os.chmod(path, stat.S_IWRITE)
# use the original function to repeat the operation
func(path)
def display_path(path):
"""Gives the display value for a given path, making it relative to cwd
if possible."""
path = os.path.normcase(os.path.abspath(path))
if path.startswith(os.getcwd() + os.path.sep):
path = '.' + path[len(os.getcwd()):]
return path
def backup_dir(dir, ext='.bak'):
"""Figure out the name of a directory to back up the given dir to
(adding .bak, .bak2, etc)"""
n = 1
extension = ext
while os.path.exists(dir + extension):
n += 1
extension = ext + str(n)
return dir + extension
def splitext(path):
"""Like os.path.splitext, but take off .tar too"""
base, ext = posixpath.splitext(path)
if base.lower().endswith('.tar'):
ext = base[-4:] + ext
base = base[:-4]
return base, ext
def find_command(cmd, paths=None, pathext=None):
"""Searches the PATH for the given command and returns its path"""
if paths is None:
paths = os.environ.get('PATH', []).split(os.pathsep)
if isinstance(paths, basestring):
paths = [paths]
# check if there are funny path extensions for executables, e.g. Windows
if pathext is None:
pathext = os.environ.get('PATHEXT', '.COM;.EXE;.BAT;.CMD')
pathext = [ext for ext in pathext.lower().split(os.pathsep)]
# don't use extensions if the command ends with one of them
if os.path.splitext(cmd)[1].lower() in pathext:
pathext = ['']
# check if we find the command on PATH
for path in paths:
# try without extension first
cmd_path = os.path.join(path, cmd)
for ext in pathext:
# then including the extension
cmd_path_ext = cmd_path + ext
if os.path.exists(cmd_path_ext):
return cmd_path_ext
if os.path.exists(cmd_path):
return cmd_path
return None
def ask(message, options):
"""Ask the message interactively, with the given possible responses"""
while 1:
if os.environ.get('PIP_NO_INPUT'):
raise Exception('No input was expected ($PIP_NO_INPUT set); question: %s' % message)
response = raw_input(message)
response = response.strip().lower()
if response not in options:
print 'Your response (%r) was not one of the expected responses: %s' % (
response, ', '.join(options))
else:
return response
class _Inf(object):
"""I am bigger than everything!"""
def __cmp__(self, a):
if self is a:
return 0
return 1
def __repr__(self):
return 'Inf'
Inf = _Inf()
del _Inf
def url_to_path(url):
"""
Convert a file: URL to a path.
"""
assert url.startswith('file:'), (
"You can only turn file: urls into filenames (not %r)" % url)
path = url[len('file:'):].lstrip('/')
path = urllib.unquote(path)
if _url_drive_re.match(path):
path = path[0] + ':' + path[2:]
else:
path = '/' + path
return path
_drive_re = re.compile('^([a-z]):', re.I)
_url_drive_re = re.compile('^([a-z])[:|]', re.I)
def path_to_url(path):
"""
Convert a path to a file: URL. The path will be made absolute.
"""
path = os.path.normcase(os.path.abspath(path))
if _drive_re.match(path):
path = path[0] + '|' + path[2:]
url = urllib.quote(path)
url = url.replace(os.path.sep, '/')
url = url.lstrip('/')
return 'file:///' + url
def path_to_url2(path):
"""
Convert a path to a file: URL. The path will be made absolute and have
quoted path parts.
"""
path = os.path.normpath(os.path.abspath(path))
drive, path = os.path.splitdrive(path)
filepath = path.split(os.path.sep)
url = '/'.join([urllib.quote(part) for part in filepath])
if not drive:
url = url.lstrip('/')
return 'file:///' + drive + url
_normalize_re = re.compile(r'[^a-z]', re.I)
def normalize_name(name):
return _normalize_re.sub('-', name.lower())
def format_size(bytes):
if bytes > 1000*1000:
return '%.1fMb' % (bytes/1000.0/1000)
elif bytes > 10*1000:
return '%iKb' % (bytes/1000)
elif bytes > 1000:
return '%.1fKb' % (bytes/1000.0)
else:
return '%ibytes' % bytes
def is_url(name):
"""Returns true if the name looks like a URL"""
from pip.vcs import vcs
if ':' not in name:
return False
scheme = name.split(':', 1)[0].lower()
return scheme in ['http', 'https', 'file', 'ftp'] + vcs.all_schemes
def is_installable_dir(path):
"""Return True if `path` is a directory containing a setup.py file."""
if not os.path.isdir(path):
return False
setup_py = os.path.join(path, 'setup.py')
if os.path.isfile(setup_py):
return True
return False
def is_archive_file(name):
"""Return True if `name` is a considered as an archive file."""
archives = ('.zip', '.tar.gz', '.tar.bz2', '.tgz', '.tar', '.pybundle')
ext = splitext(name)[1].lower()
if ext in archives:
return True
return False
def is_svn_page(html):
"""Returns true if the page appears to be the index page of an svn repository"""
return (re.search(r'<title>[^<]*Revision \d+:', html)
and re.search(r'Powered by (?:<a[^>]*?>)?Subversion', html, re.I))
def file_contents(filename):
fp = open(filename, 'rb')
try:
return fp.read()
finally:
fp.close()
def split_leading_dir(path):
path = str(path)
path = path.lstrip('/').lstrip('\\')
if '/' in path and (('\\' in path and path.find('/') < path.find('\\'))
or '\\' not in path):
return path.split('/', 1)
elif '\\' in path:
return path.split('\\', 1)
else:
return path, ''
def has_leading_dir(paths):
"""Returns true if all the paths have the same leading path name
(i.e., everything is in one subdirectory in an archive)"""
common_prefix = None
for path in paths:
prefix, rest = split_leading_dir(path)
if not prefix:
return False
elif common_prefix is None:
common_prefix = prefix
elif prefix != common_prefix:
return False
return True
def make_path_relative(path, rel_to):
"""
Make a filename relative, where the filename path, and it is
relative to rel_to
>>> make_relative_path('/usr/share/something/a-file.pth',
... '/usr/share/another-place/src/Directory')
'../../../something/a-file.pth'
>>> make_relative_path('/usr/share/something/a-file.pth',
... '/home/user/src/Directory')
'../../../usr/share/something/a-file.pth'
>>> make_relative_path('/usr/share/a-file.pth', '/usr/share/')
'a-file.pth'
"""
path_filename = os.path.basename(path)
path = os.path.dirname(path)
path = os.path.normpath(os.path.abspath(path))
rel_to = os.path.normpath(os.path.abspath(rel_to))
path_parts = path.strip(os.path.sep).split(os.path.sep)
rel_to_parts = rel_to.strip(os.path.sep).split(os.path.sep)
while path_parts and rel_to_parts and path_parts[0] == rel_to_parts[0]:
path_parts.pop(0)
rel_to_parts.pop(0)
full_parts = ['..']*len(rel_to_parts) + path_parts + [path_filename]
if full_parts == ['']:
return '.' + os.path.sep
return os.path.sep.join(full_parts)
def normalize_path(path):
"""
Convert a path to its canonical, case-normalized, absolute version.
"""
return os.path.normcase(os.path.realpath(path))
_scheme_re = re.compile(r'^(http|https|file):', re.I)
_url_slash_drive_re = re.compile(r'/*([a-z])\|', re.I)
def geturl(urllib2_resp):
"""
Use instead of urllib.addinfourl.geturl(), which appears to have
some issues with dropping the double slash for certain schemes
(e.g. file://). This implementation is probably over-eager, as it
always restores '://' if it is missing, and it appears some url
schemata aren't always followed by '//' after the colon, but as
far as I know pip doesn't need any of those.
"""
url = urllib2_resp.geturl()
scheme, rest = url.split(':', 1)
if rest.startswith('//'):
return url
else:
return '%s//%s' % (scheme, rest)
def get_file_content(url, comes_from=None):
"""Gets the content of a file; it may be a filename, file: URL, or
http: URL. Returns (location, content)"""
match = _scheme_re.search(url)
if match:
scheme = match.group(1).lower()
if (scheme == 'file' and comes_from
and comes_from.startswith('http')):
raise InstallationError(
'Requirements file %s references URL %s, which is local'
% (comes_from, url))
if scheme == 'file':
path = url.split(':', 1)[1]
path = path.replace('\\', '/')
match = _url_slash_drive_re.match(path)
if match:
path = match.group(1) + ':' + path.split('|', 1)[1]
path = urllib.unquote(path)
if path.startswith('/'):
path = '/' + path.lstrip('/')
url = path
else:
## FIXME: catch some errors
resp = urllib2.urlopen(url)
return geturl(resp), resp.read()
f = open(url)
content = f.read()
f.close()
return url, content
def renames(old, new):
"""Like os.renames(), but handles renaming across devices."""
# Implementation borrowed from os.renames().
head, tail = os.path.split(new)
if head and tail and not os.path.exists(head):
os.makedirs(head)
shutil.move(old, new)
head, tail = os.path.split(old)
if head and tail:
try:
os.removedirs(head)
except OSError:
pass
def in_venv():
"""
Return True if we're running inside a virtualenv, False otherwise.
"""
return hasattr(sys, 'real_prefix')
def is_local(path):
"""
Return True if path is within sys.prefix, if we're running in a virtualenv.
If we're not in a virtualenv, all paths are considered "local."
"""
if not in_venv():
return True
return normalize_path(path).startswith(normalize_path(sys.prefix))
def dist_is_local(dist):
"""
Return True if given Distribution object is installed locally
(i.e. within current virtualenv).
Always True if we're not in a virtualenv.
"""
return is_local(dist_location(dist))
def get_installed_distributions(local_only=True, skip=('setuptools', 'pip', 'python')):
"""
Return a list of installed Distribution objects.
If ``local_only`` is True (default), only return installations
local to the current virtualenv, if in a virtualenv.
``skip`` argument is an iterable of lower-case project names to
ignore; defaults to ('setuptools', 'pip', 'python'). [FIXME also
skip virtualenv?]
"""
if local_only:
local_test = dist_is_local
else:
local_test = lambda d: True
return [d for d in pkg_resources.working_set if local_test(d) and d.key not in skip]
def egg_link_path(dist):
"""
Return the path where we'd expect to find a .egg-link file for
this distribution. (There doesn't seem to be any metadata in the
Distribution object for a develop egg that points back to its
.egg-link and easy-install.pth files).
This won't find a globally-installed develop egg if we're in a
virtualenv.
"""
return os.path.join(site_packages, dist.project_name) + '.egg-link'
def dist_location(dist):
"""
Get the site-packages location of this distribution. Generally
this is dist.location, except in the case of develop-installed
packages, where dist.location is the source code location, and we
want to know where the egg-link file is.
"""
egg_link = egg_link_path(dist)
if os.path.exists(egg_link):
return egg_link
return dist.location
def get_terminal_size():
"""Returns a tuple (x, y) representing the width(x) and the height(x)
in characters of the terminal window."""
def ioctl_GWINSZ(fd):
try:
import fcntl
import termios
import struct
cr = struct.unpack('hh', fcntl.ioctl(fd, termios.TIOCGWINSZ,
'1234'))
except:
return None
return cr
cr = ioctl_GWINSZ(0) or ioctl_GWINSZ(1) or ioctl_GWINSZ(2)
if not cr:
try:
fd = os.open(os.ctermid(), os.O_RDONLY)
cr = ioctl_GWINSZ(fd)
os.close(fd)
except:
pass
if not cr:
cr = (os.environ.get('LINES', 25), os.environ.get('COLUMNS', 80))
return int(cr[1]), int(cr[0])
# Insurance against "creative" interpretation of the RFC:
# http://bugs.python.org/issue8732
def urlopen(url):
if isinstance(url, basestring):
url = urllib2.Request(url, headers={'Accept-encoding': 'identity'})
return urllib2.urlopen(url)