def testLooksLikeGlob(self):
# The way to test bash behavior is:
# $ shopt -s nullglob; argv [ # not a glob
# $ shopt -s nullglob; argv [] # is a glob
# $ shopt -s nullglob; argv [][ # is a glob
CASES = [
(r'[]', True),
(r'[a]', True),
(r'[][', True),
(r'][', False), # no balanced pair
(r'\[]', False), # no balanced pair
(r'[', False), # no balanced pair
(r']', False), # no balanced pair
(r'echo', False),
(r'status=0', False),
(r'*', True),
(r'\*', False),
(r'\*.sh', False),
('\\', False),
('*\\', True),
('?', True),
]
for pat, expected in CASES:
self.assertEqual(expected, glob_.LooksLikeGlob(pat),
'%s: expected %r' % (pat, expected))
def DoUnarySuffixOp(s, op, arg):
"""Helper for ${x#prefix} and family."""
# Fast path for constant strings.
if not glob_.LooksLikeGlob(arg):
if op.op_id in (Id.VOp1_Pound, Id.VOp1_DPound): # const prefix
if s.startswith(arg):
return s[len(arg):]
else:
return s
elif op.op_id in (Id.VOp1_Percent, Id.VOp1_DPercent): # const suffix
if s.endswith(arg):
return s[:-len(arg)]
else:
return s
elif op.op_id == Id.VOp1_Comma: # Only lowercase the first letter
if arg != '':
raise NotImplementedError("%s can't have an argument" % op.op_id)
return s[0].lower() + s[1:]
elif op.op_id == Id.VOp1_DComma:
if arg != '':
raise NotImplementedError("%s can't have an argument" % op.op_id)
return s.lower()
elif op.op_id == Id.VOp1_Caret: # Only uppercase the first letter
if arg != '':
raise NotImplementedError("%s can't have an argument" % op.op_id)
return s[0].upper() + s[1:]
elif op.op_id == Id.VOp1_DCaret:
if arg != '':
raise NotImplementedError("%s can't have an argument" % op.op_id)
return s.upper()
else: # e.g. ^ ^^ , ,,
raise AssertionError(op.op_id)
# For patterns, do fnmatch() in a loop.
#
# TODO: The loop needs to iterate over code points, not bytes!
# - The forward case can probably be handled in a similar manner.
# - The backward case might be handled by pre-calculating an array of start
# positions with _NextUtf8Char.
#
# TODO: Another potential fast path:
#
# v=aabbccdd
# echo ${v#*b} # strip shortest prefix
#
# If the whole thing doesn't match '*b*', then no test can succeed. So we
# can fail early. Conversely echo ${v%%c*} and '*c*'.
#
# (Although honestly this whole construct is nuts and should be deprecated.)
n = len(s)
if op.op_id == Id.VOp1_Pound: # shortest prefix
# 'abcd': match 'a', 'ab', 'abc', ...
for i in xrange(1, n+1):
#log('Matching pattern %r with %r', arg, s[:i])
if libc.fnmatch(arg, s[:i]):
return s[i:]
else:
return s
elif op.op_id == Id.VOp1_DPound: # longest prefix
# 'abcd': match 'abc', 'ab', 'a'
for i in xrange(n, 0, -1):
#log('Matching pattern %r with %r', arg, s[:i])
if libc.fnmatch(arg, s[:i]):
return s[i:]
else:
return s
elif op.op_id == Id.VOp1_Percent: # shortest suffix
# 'abcd': match 'abc', 'ab', 'a'
for i in xrange(n-1, -1, -1):
#log('Matching pattern %r with %r', arg, s[:i])
if libc.fnmatch(arg, s[i:]):
return s[:i]
else:
return s
elif op.op_id == Id.VOp1_DPercent: # longest suffix
# 'abcd': match 'abc', 'bc', 'c', ...
for i in xrange(0, n):
#log('Matching pattern %r with %r', arg, s[:i])
if libc.fnmatch(arg, s[i:]):
return s[:i]
else:
return s
else:
raise NotImplementedError("Can't use %s with pattern" % op.op_id)
def DoUnarySuffixOp(s, op, arg, extglob):
# type: (str, suffix_op__Unary, str, bool) -> str
"""Helper for ${x#prefix} and family."""
# Fast path for constant strings.
if not glob_.LooksLikeGlob(arg):
# It doesn't look like a glob, but we glob-escaped it (e.g. [ -> \[). So
# reverse it. NOTE: We also do this check in Globber.Expand(). It would
# be nice to somehow store the original string rather tahn
# escaping/unescaping.
arg = glob_.GlobUnescape(arg)
if op.op_id in (Id.VOp1_Pound, Id.VOp1_DPound): # const prefix
# explicit check for non-empty arg (len for mycpp)
if len(arg) and s.startswith(arg):
return s[len(arg):]
else:
return s
elif op.op_id in (Id.VOp1_Percent, Id.VOp1_DPercent): # const suffix
# need explicit check for non-empty arg (len for mycpp)
if len(arg) and s.endswith(arg):
return s[:-len(arg)]
else:
return s
# These operators take glob arguments, we don't implement that obscure case.
elif op.op_id == Id.VOp1_Comma: # Only lowercase the first letter
if arg != '':
# TODO: location info for op
e_die("%s can't have an argument", ui.PrettyId(op.op_id))
if len(s):
return s[0].lower() + s[1:]
else:
return s
elif op.op_id == Id.VOp1_DComma:
if arg != '':
e_die("%s can't have an argument", ui.PrettyId(op.op_id))
return s.lower()
elif op.op_id == Id.VOp1_Caret: # Only uppercase the first letter
if arg != '':
e_die("%s can't have an argument", ui.PrettyId(op.op_id))
if len(s):
return s[0].upper() + s[1:]
else:
return s
elif op.op_id == Id.VOp1_DCaret:
if arg != '':
e_die("%s can't have an argument", ui.PrettyId(op.op_id))
return s.upper()
else: # e.g. ^ ^^ , ,,
raise AssertionError(op.op_id)
# For patterns, do fnmatch() in a loop.
#
# TODO:
# - Another potential fast path:
# v=aabbccdd
# echo ${v#*b} # strip shortest prefix
#
# If the whole thing doesn't match '*b*', then no test can succeed. So we
# can fail early. Conversely echo ${v%%c*} and '*c*'.
#
# (Although honestly this whole construct is nuts and should be deprecated.)
n = len(s)
if op.op_id == Id.VOp1_Pound: # shortest prefix
# 'abcd': match '', 'a', 'ab', 'abc', ...
i = 0
while True:
assert i <= n
#log('Matching pattern %r with %r', arg, s[:i])
if libc.fnmatch(arg, s[:i], extglob):
return s[i:]
if i >= n:
break
i = _NextUtf8Char(s, i)
return s
elif op.op_id == Id.VOp1_DPound: # longest prefix
# 'abcd': match 'abc', 'ab', 'a'
i = n
while True:
assert i >= 0
#log('Matching pattern %r with %r', arg, s[:i])
if libc.fnmatch(arg, s[:i], extglob):
return s[i:]
if i == 0:
break
i = PreviousUtf8Char(s, i)
return s
elif op.op_id == Id.VOp1_Percent: # shortest suffix
# 'abcd': match 'abcd', 'abc', 'ab', 'a'
i = n
while True:
assert i >= 0
#log('Matching pattern %r with %r', arg, s[:i])
if libc.fnmatch(arg, s[i:], extglob):
return s[:i]
if i == 0:
break
i = PreviousUtf8Char(s, i)
return s
elif op.op_id == Id.VOp1_DPercent: # longest suffix
# 'abcd': match 'abc', 'bc', 'c', ...
i = 0
while True:
assert i <= n
#log('Matching pattern %r with %r', arg, s[:i])
if libc.fnmatch(arg, s[i:], extglob):
return s[:i]
if i >= n:
break
i = _NextUtf8Char(s, i)
return s
else:
raise NotImplementedError(ui.PrettyId(op.op_id))
def DoUnarySuffixOp(s, op, arg):
"""Helper for ${x#prefix} and family."""
# Fast path for constant strings.
if not glob_.LooksLikeGlob(arg):
# It doesn't look like a glob, but we glob-escaped it (e.g. [ -> \[). So
# reverse it. NOTE: We also do this check in Globber.Expand(). It would
# be nice to somehow store the original string rather tahn
# escaping/unescaping.
arg = glob_.GlobUnescape(arg)
if op.op_id in (Id.VOp1_Pound, Id.VOp1_DPound): # const prefix
# explicit check for non-empty arg (len for mycpp)
if len(arg) and s.startswith(arg):
return s[len(arg):]
else:
return s
elif op.op_id in (Id.VOp1_Percent, Id.VOp1_DPercent): # const suffix
# need explicit check for non-empty arg (len for mycpp)
if len(arg) and s.endswith(arg):
return s[:-len(arg)]
else:
return s
# These operators take glob arguments, we don't implement that obscure case.
elif op.op_id == Id.VOp1_Comma: # Only lowercase the first letter
if arg != '':
raise NotImplementedError("%s can't have an argument" %
op.op_id)
return s[0].lower() + s[1:]
elif op.op_id == Id.VOp1_DComma:
if arg != '':
raise NotImplementedError("%s can't have an argument" %
op.op_id)
return s.lower()
elif op.op_id == Id.VOp1_Caret: # Only uppercase the first letter
if arg != '':
raise NotImplementedError("%s can't have an argument" %
op.op_id)
return s[0].upper() + s[1:]
elif op.op_id == Id.VOp1_DCaret:
if arg != '':
raise NotImplementedError("%s can't have an argument" %
op.op_id)
return s.upper()
else: # e.g. ^ ^^ , ,,
raise AssertionError(op.op_id)
# For patterns, do fnmatch() in a loop.
#
# TODO: The loop needs to iterate over code points, not bytes!
# - The forward case can probably be handled in a similar manner.
# - The backward case might be handled by pre-calculating an array of start
# positions with _NextUtf8Char.
#
# TODO: Another potential fast path:
#
# v=aabbccdd
# echo ${v#*b} # strip shortest prefix
#
# If the whole thing doesn't match '*b*', then no test can succeed. So we
# can fail early. Conversely echo ${v%%c*} and '*c*'.
#
# (Although honestly this whole construct is nuts and should be deprecated.)
n = len(s)
if op.op_id == Id.VOp1_Pound: # shortest prefix
# 'abcd': match 'a', 'ab', 'abc', ...
for i in xrange(1, n + 1):
#log('Matching pattern %r with %r', arg, s[:i])
if libc.fnmatch(arg, s[:i]):
return s[i:]
else:
return s
elif op.op_id == Id.VOp1_DPound: # longest prefix
# 'abcd': match 'abc', 'ab', 'a'
for i in xrange(n, 0, -1):
#log('Matching pattern %r with %r', arg, s[:i])
if libc.fnmatch(arg, s[:i]):
return s[i:]
else:
return s
elif op.op_id == Id.VOp1_Percent: # shortest suffix
# 'abcd': match 'abc', 'ab', 'a'
for i in xrange(n - 1, -1, -1):
#log('Matching pattern %r with %r', arg, s[:i])
if libc.fnmatch(arg, s[i:]):
return s[:i]
else:
return s
elif op.op_id == Id.VOp1_DPercent: # longest suffix
# 'abcd': match 'abc', 'bc', 'c', ...
for i in xrange(0, n):
#log('Matching pattern %r with %r', arg, s[:i])
if libc.fnmatch(arg, s[i:]):
return s[:i]
else:
return s
else:
raise NotImplementedError("Can't use %s with pattern" % op.op_id)
