summaryrefslogtreecommitdiff
path: root/Lib/distutils/util.py
diff options
context:
space:
mode:
authorNeal Norwitz <nnorwitz@gmail.com>2007-04-17 08:48:32 +0000
committerNeal Norwitz <nnorwitz@gmail.com>2007-04-17 08:48:32 +0000
commit9d72bb452bced3a100f07f8a9e30c4495a9ec41a (patch)
treec39762a764fcc16f2cfc42e2504e58ff31e159e6 /Lib/distutils/util.py
parentff11334927ee616d765b54a3851016b76a20bcec (diff)
downloadcpython-git-9d72bb452bced3a100f07f8a9e30c4495a9ec41a.tar.gz
Remove functions in string module that are also string methods. Also remove:
* all calls to functions in the string module (except maketrans) * everything from stropmodule except for maketrans() which is still used
Diffstat (limited to 'Lib/distutils/util.py')
-rw-r--r--Lib/distutils/util.py26
1 files changed, 12 insertions, 14 deletions
diff --git a/Lib/distutils/util.py b/Lib/distutils/util.py
index 16d8bcba4d..6f15ce8b5e 100644
--- a/Lib/distutils/util.py
+++ b/Lib/distutils/util.py
@@ -42,10 +42,9 @@ def get_platform ():
# Convert the OS name to lowercase, remove '/' characters
# (to accommodate BSD/OS), and translate spaces (for "Power Macintosh")
- osname = string.lower(osname)
- osname = string.replace(osname, '/', '')
- machine = string.replace(machine, ' ', '_')
- machine = string.replace(machine, '/', '-')
+ osname = osname.lower().replace('/', '')
+ machine = machine.replace(' ', '_')
+ machine = machine.replace('/', '-')
if osname[:5] == "linux":
# At least on Linux/Intel, 'machine' is the processor --
@@ -139,7 +138,7 @@ def convert_path (pathname):
if pathname[-1] == '/':
raise ValueError, "path '%s' cannot end with '/'" % pathname
- paths = string.split(pathname, '/')
+ paths = pathname.split('/')
while '.' in paths:
paths.remove('.')
if not paths:
@@ -178,7 +177,7 @@ def change_root (new_root, pathname):
return os.path.join(new_root, pathname)
else:
# Chop off volume name from start of path
- elements = string.split(pathname, ":", 1)
+ elements = pathname.split(":", 1)
pathname = ":" + elements[1]
return os.path.join(new_root, pathname)
@@ -281,7 +280,7 @@ def split_quoted (s):
# bit of a brain-bender to get it working right, though...
if _wordchars_re is None: _init_regex()
- s = string.strip(s)
+ s = s.strip()
words = []
pos = 0
@@ -294,7 +293,7 @@ def split_quoted (s):
if s[end] in string.whitespace: # unescaped, unquoted whitespace: now
words.append(s[:end]) # we definitely have a word delimiter
- s = string.lstrip(s[end:])
+ s = s[end:].lstrip()
pos = 0
elif s[end] == '\\': # preserve whatever is being escaped;
@@ -354,7 +353,7 @@ def strtobool (val):
are 'n', 'no', 'f', 'false', 'off', and '0'. Raises ValueError if
'val' is anything else.
"""
- val = string.lower(val)
+ val = val.lower()
if val in ('y', 'yes', 't', 'true', 'on', '1'):
return 1
elif val in ('n', 'no', 'f', 'false', 'off', '0'):
@@ -445,7 +444,7 @@ files = [
#if prefix:
# prefix = os.path.abspath(prefix)
- script.write(string.join(map(repr, py_files), ",\n") + "]\n")
+ script.write(",\n".join(map(repr, py_files)) + "]\n")
script.write("""
byte_compile(files, optimize=%r, force=%r,
prefix=%r, base_dir=%r,
@@ -507,7 +506,6 @@ def rfc822_escape (header):
"""Return a version of the string escaped for inclusion in an
RFC-822 header, by ensuring there are 8 spaces space after each newline.
"""
- lines = string.split(header, '\n')
- lines = map(string.strip, lines)
- header = string.join(lines, '\n' + 8*' ')
- return header
+ lines = [x.strip() for x in header.split('\n')]
+ sep = '\n' + 8*' '
+ return sep.join(lines)