def Run(self, cmd_val):
# type: (cmd_value__Argv) -> int
"""
printf: printf [-v var] format [argument ...]
"""
arg_r = args.Reader(cmd_val.argv, spids=cmd_val.arg_spids)
arg_r.Next() # skip argv[0]
arg = PRINTF_SPEC.Parse(arg_r)
fmt, fmt_spid = arg_r.ReadRequired2('requires a format string')
varargs, spids = arg_r.Rest2()
#log('fmt %s', fmt)
#log('vals %s', vals)
arena = self.parse_ctx.arena
if fmt in self.parse_cache:
parts = self.parse_cache[fmt]
else:
line_reader = reader.StringLineReader(fmt, arena)
# TODO: Make public
lexer = self.parse_ctx._MakeLexer(line_reader)
p = _FormatStringParser(lexer)
arena.PushSource(source.ArgvWord(fmt_spid))
try:
parts = p.Parse()
except error.Parse as e:
self.errfmt.PrettyPrintError(e)
return 2 # parse error
finally:
arena.PopSource()
self.parse_cache[fmt] = parts
if 0:
print()
for part in parts:
part.PrettyPrint()
print()
out = []
arg_index = 0
num_args = len(varargs)
backslash_c = False
while True:
for part in parts:
if isinstance(part, printf_part.Literal):
token = part.token
if token.id == Id.Format_EscapedPercent:
s = '%'
else:
s = word_compile.EvalCStringToken(token.id, token.val)
out.append(s)
elif isinstance(part, printf_part.Percent):
flags = None
if len(part.flags) > 0:
flags = ''
for flag_token in part.flags:
flags += flag_token.val
width = None
if part.width:
if part.width.id in (Id.Format_Num, Id.Format_Zero):
width = part.width.val
width_spid = part.width.span_id
elif part.width.id == Id.Format_Star:
if arg_index < num_args:
width = varargs[arg_index]
width_spid = spids[arg_index]
arg_index += 1
else:
width = ''
width_spid = runtime.NO_SPID
else:
raise AssertionError()
try:
width = int(width)
except ValueError:
if width_spid == runtime.NO_SPID:
width_spid = part.width.span_id
self.errfmt.Print(
"printf got invalid number %r for the width",
s,
span_id=width_spid)
return 1
precision = None
if part.precision:
if part.precision.id == Id.Format_Dot:
precision = '0'
precision_spid = part.precision.span_id
elif part.precision.id in (Id.Format_Num,
Id.Format_Zero):
precision = part.precision.val
precision_spid = part.precision.span_id
elif part.precision.id == Id.Format_Star:
if arg_index < num_args:
precision = varargs[arg_index]
precision_spid = spids[arg_index]
arg_index += 1
else:
precision = ''
precision_spid = runtime.NO_SPID
else:
raise AssertionError()
try:
precision = int(precision)
except ValueError:
if precision_spid == runtime.NO_SPID:
precision_spid = part.precision.span_id
self.errfmt.Print(
"printf got invalid number %r for the precision",
s,
span_id=precision_spid)
return 1
if arg_index < num_args:
s = varargs[arg_index]
word_spid = spids[arg_index]
arg_index += 1
else:
s = ''
word_spid = runtime.NO_SPID
typ = part.type.val
if typ == 's':
if precision is not None:
s = s[:precision] # truncate
elif typ == 'q':
s = qsn.maybe_shell_encode(s)
elif typ == 'b':
# Process just like echo -e, except \c handling is simpler.
parts = [] # type: List[str]
lex = match.EchoLexer(s)
while True:
id_, value = lex.Next()
if id_ == Id.Eol_Tok: # Note: This is really a NUL terminator
break
p = word_compile.EvalCStringToken(id_, value)
# Unusual behavior: '\c' aborts processing!
if p is None:
backslash_c = True
break
parts.append(p)
s = ''.join(parts)
elif typ in 'diouxX' or part.type.id == Id.Format_Time:
try:
d = int(s)
except ValueError:
if len(s) >= 1 and s[0] in '\'"':
# TODO: utf-8 decode s[1:] to be more correct. Probably
# depends on issue #366, a utf-8 library.
# Note: len(s) == 1 means there is a NUL (0) after the quote..
d = ord(s[1]) if len(s) >= 2 else 0
elif part.type.id == Id.Format_Time and len(
s) == 0 and word_spid == runtime.NO_SPID:
# Note: No argument means -1 for %(...)T as in Bash Reference
# Manual 4.2 "If no argument is specified, conversion behaves
# as if -1 had been given."
d = -1
else:
# This works around the fact that in the arg recycling case, you have no spid.
if word_spid == runtime.NO_SPID:
self.errfmt.Print(
"printf got invalid number %r for this substitution",
s,
span_id=part.type.span_id)
else:
self.errfmt.Print(
"printf got invalid number %r",
s,
span_id=word_spid)
return 1
if typ in 'di':
s = str(d)
elif typ in 'ouxX':
if d < 0:
e_die(
"Can't format negative number %d with %%%s",
d,
typ,
span_id=part.type.span_id)
if typ == 'u':
s = str(d)
elif typ == 'o':
s = '%o' % d
elif typ == 'x':
s = '%x' % d
elif typ == 'X':
s = '%X' % d
elif part.type.id == Id.Format_Time:
# %(...)T
# Initialize timezone:
# `localtime' uses the current timezone information initialized
# by `tzset'. The function `tzset' refers to the environment
# variable `TZ'. When the exported variable `TZ' is present,
# its value should be reflected in the real environment
# variable `TZ' before call of `tzset'.
#
# Note: unlike LANG, TZ doesn't seem to change behavior if it's
# not exported.
#
# TODO: In Oil, provide an API that doesn't rely on libc's
# global state.
tzcell = self.mem.GetCell('TZ')
if tzcell and tzcell.exported and tzcell.val.tag_(
) == value_e.Str:
tzval = cast(value__Str, tzcell.val)
posix.putenv('TZ', tzval.s)
time.tzset()
# Handle special values:
# User can specify two special values -1 and -2 as in Bash
# Reference Manual 4.2: "Two special argument values may be
# used: -1 represents the current time, and -2 represents the
# time the shell was invoked." from
# https://www.gnu.org/software/bash/manual/html_node/Bash-Builtins.html#index-printf
if d == -1: # the current time
d = time.time()
elif d == -2: # the shell start time
d = shell_start_time
s = time.strftime(typ[1:-2], time.localtime(d))
if precision is not None:
s = s[:precision] # truncate
else:
raise AssertionError()
else:
raise AssertionError()
if width is not None:
if flags:
if '-' in flags:
s = s.ljust(width, ' ')
elif '0' in flags:
s = s.rjust(width, '0')
else:
pass
else:
s = s.rjust(width, ' ')
out.append(s)
else:
raise AssertionError()
if backslash_c: # 'printf %b a\cb xx' - \c terminates processing!
break
if arg_index >= num_args:
break
# Otherwise there are more args. So cycle through the loop once more to
# implement the 'arg recycling' behavior.
result = ''.join(out)
if arg.v:
var_name = arg.v
# Notes:
# - bash allows a[i] here (as in unset and ${!x}), but we haven't
# implemented it.
# - TODO: get the span_id for arg.v!
if not match.IsValidVarName(var_name):
raise error.Usage('got invalid variable name %r' % var_name)
state.SetStringDynamic(self.mem, var_name, result)
else:
sys.stdout.write(result)
return 0
def Run(self, cmd_val):
# type: (cmd_value__Argv) -> int
"""
printf: printf [-v var] format [argument ...]
"""
attrs, arg_r = flag_spec.ParseCmdVal('printf', cmd_val)
arg = arg_types.printf(attrs.attrs)
fmt, fmt_spid = arg_r.ReadRequired2('requires a format string')
varargs, spids = arg_r.Rest2()
#log('fmt %s', fmt)
#log('vals %s', vals)
arena = self.parse_ctx.arena
if fmt in self.parse_cache:
parts = self.parse_cache[fmt]
else:
line_reader = reader.StringLineReader(fmt, arena)
# TODO: Make public
lexer = self.parse_ctx._MakeLexer(line_reader)
parser = _FormatStringParser(lexer)
with alloc.ctx_Location(arena, source.ArgvWord(fmt_spid)):
try:
parts = parser.Parse()
except error.Parse as e:
self.errfmt.PrettyPrintError(e)
return 2 # parse error
self.parse_cache[fmt] = parts
if 0:
print()
for part in parts:
part.PrettyPrint()
print()
out = [] # type: List[str]
arg_index = 0
num_args = len(varargs)
backslash_c = False
while True:
for part in parts:
UP_part = part
if part.tag_() == printf_part_e.Literal:
part = cast(printf_part__Literal, UP_part)
token = part.token
if token.id == Id.Format_EscapedPercent:
s = '%'
else:
s = word_compile.EvalCStringToken(token)
out.append(s)
elif part.tag_() == printf_part_e.Percent:
part = cast(printf_part__Percent, UP_part)
flags = [] # type: List[str]
if len(part.flags) > 0:
for flag_token in part.flags:
flags.append(flag_token.val)
width = -1 # nonexistent
if part.width:
if part.width.id in (Id.Format_Num, Id.Format_Zero):
width_str = part.width.val
width_spid = part.width.span_id
elif part.width.id == Id.Format_Star:
if arg_index < num_args:
width_str = varargs[arg_index]
width_spid = spids[arg_index]
arg_index += 1
else:
width_str = '' # invalid
width_spid = runtime.NO_SPID
else:
raise AssertionError()
try:
width = int(width_str)
except ValueError:
if width_spid == runtime.NO_SPID:
width_spid = part.width.span_id
self.errfmt.Print_("printf got invalid width %r" %
width_str,
span_id=width_spid)
return 1
precision = -1 # nonexistent
if part.precision:
if part.precision.id == Id.Format_Dot:
precision_str = '0'
precision_spid = part.precision.span_id
elif part.precision.id in (Id.Format_Num,
Id.Format_Zero):
precision_str = part.precision.val
precision_spid = part.precision.span_id
elif part.precision.id == Id.Format_Star:
if arg_index < num_args:
precision_str = varargs[arg_index]
precision_spid = spids[arg_index]
arg_index += 1
else:
precision_str = ''
precision_spid = runtime.NO_SPID
else:
raise AssertionError()
try:
precision = int(precision_str)
except ValueError:
if precision_spid == runtime.NO_SPID:
precision_spid = part.precision.span_id
self.errfmt.Print_(
'printf got invalid precision %r' %
precision_str,
span_id=precision_spid)
return 1
#log('index=%d n=%d', arg_index, num_args)
if arg_index < num_args:
s = varargs[arg_index]
word_spid = spids[arg_index]
arg_index += 1
else:
s = ''
word_spid = runtime.NO_SPID
typ = part.type.val
if typ == 's':
if precision >= 0:
s = s[:precision] # truncate
elif typ == 'q':
s = qsn.maybe_shell_encode(s)
elif typ == 'b':
# Process just like echo -e, except \c handling is simpler.
c_parts = [] # type: List[str]
lex = match.EchoLexer(s)
while True:
id_, tok_val = lex.Next()
if id_ == Id.Eol_Tok: # Note: This is really a NUL terminator
break
# TODO: add span_id from argv
tok = Token(id_, runtime.NO_SPID, tok_val)
p = word_compile.EvalCStringToken(tok)
# Unusual behavior: '\c' aborts processing!
if p is None:
backslash_c = True
break
c_parts.append(p)
s = ''.join(c_parts)
elif typ in 'diouxX' or part.type.id == Id.Format_Time:
try:
d = int(s)
except ValueError:
if len(s) >= 1 and s[0] in '\'"':
# TODO: utf-8 decode s[1:] to be more correct. Probably
# depends on issue #366, a utf-8 library.
# Note: len(s) == 1 means there is a NUL (0) after the quote..
d = ord(s[1]) if len(s) >= 2 else 0
elif part.type.id == Id.Format_Time and len(
s) == 0 and word_spid == runtime.NO_SPID:
# Note: No argument means -1 for %(...)T as in Bash Reference
# Manual 4.2 "If no argument is specified, conversion behaves
# as if -1 had been given."
d = -1
else:
if word_spid == runtime.NO_SPID:
# Blame the format string
blame_spid = part.type.span_id
else:
blame_spid = word_spid
self.errfmt.Print_(
'printf expected an integer, got %r' % s,
span_id=blame_spid)
return 1
if typ in 'di':
s = str(d)
elif typ in 'ouxX':
if d < 0:
e_die(
"Can't format negative number %d with %%%s",
d,
typ,
span_id=part.type.span_id)
if typ == 'u':
s = str(d)
elif typ == 'o':
s = mylib.octal(d)
elif typ == 'x':
s = mylib.hex_lower(d)
elif typ == 'X':
s = mylib.hex_upper(d)
elif part.type.id == Id.Format_Time:
# %(...)T
# Initialize timezone:
# `localtime' uses the current timezone information initialized
# by `tzset'. The function `tzset' refers to the environment
# variable `TZ'. When the exported variable `TZ' is present,
# its value should be reflected in the real environment
# variable `TZ' before call of `tzset'.
#
# Note: unlike LANG, TZ doesn't seem to change behavior if it's
# not exported.
#
# TODO: In Oil, provide an API that doesn't rely on libc's
# global state.
tzcell = self.mem.GetCell('TZ')
if tzcell and tzcell.exported and tzcell.val.tag_(
) == value_e.Str:
tzval = cast(value__Str, tzcell.val)
posix.putenv('TZ', tzval.s)
time_.tzset()
# Handle special values:
# User can specify two special values -1 and -2 as in Bash
# Reference Manual 4.2: "Two special argument values may be
# used: -1 represents the current time, and -2 represents the
# time the shell was invoked." from
# https://www.gnu.org/software/bash/manual/html_node/Bash-Builtins.html#index-printf
if d == -1: # the current time
ts = time_.time()
elif d == -2: # the shell start time
ts = self.shell_start_time
else:
ts = d
s = time_.strftime(typ[1:-2], time_.localtime(ts))
if precision >= 0:
s = s[:precision] # truncate
else:
raise AssertionError()
else:
raise AssertionError()
if width >= 0:
if len(flags):
if '-' in flags:
s = s.ljust(width, ' ')
elif '0' in flags:
s = s.rjust(width, '0')
else:
pass
else:
s = s.rjust(width, ' ')
out.append(s)
else:
raise AssertionError()
if backslash_c: # 'printf %b a\cb xx' - \c terminates processing!
break
if arg_index >= num_args:
break
# Otherwise there are more args. So cycle through the loop once more to
# implement the 'arg recycling' behavior.
result = ''.join(out)
if arg.v is not None:
# TODO: get the span_id for arg.v!
v_spid = runtime.NO_SPID
arena = self.parse_ctx.arena
a_parser = self.parse_ctx.MakeArithParser(arg.v)
with alloc.ctx_Location(arena, source.ArgvWord(v_spid)):
try:
anode = a_parser.Parse()
except error.Parse as e:
ui.PrettyPrintError(e, arena) # show parse error
e_usage('Invalid -v expression', span_id=v_spid)
lval = self.arith_ev.EvalArithLhs(anode, v_spid)
if not self.exec_opts.eval_unsafe_arith(
) and lval.tag_() != lvalue_e.Named:
e_usage(
'-v expected a variable name. shopt -s eval_unsafe_arith allows expressions',
span_id=v_spid)
state.SetRef(self.mem, lval, value.Str(result))
else:
mylib.Stdout().write(result)
return 0
def _Format(self, parts, varargs, spids, out):
# type: (List[printf_part_t], List[str], List[int], List[str]) -> int
"""Hairy printf formatting logic."""
arg_index = 0
num_args = len(varargs)
backslash_c = False
while True: # loop over arguments
for part in parts: # loop over parsed format string
UP_part = part
if part.tag_() == printf_part_e.Literal:
part = cast(printf_part__Literal, UP_part)
token = part.token
if token.id == Id.Format_EscapedPercent:
s = '%'
else:
s = word_compile.EvalCStringToken(token)
out.append(s)
elif part.tag_() == printf_part_e.Percent:
# Note: This case is very long, but hard to refactor because of the
# error cases and "recycling" of args! (arg_index, return 1, etc.)
part = cast(printf_part__Percent, UP_part)
# TODO: These calculations are independent of the data, so could be
# cached
flags = [] # type: List[str]
if len(part.flags) > 0:
for flag_token in part.flags:
flags.append(flag_token.val)
width = -1 # nonexistent
if part.width:
if part.width.id in (Id.Format_Num, Id.Format_Zero):
width_str = part.width.val
width_spid = part.width.span_id
elif part.width.id == Id.Format_Star:
if arg_index < num_args:
width_str = varargs[arg_index]
width_spid = spids[arg_index]
arg_index += 1
else:
width_str = '' # invalid
width_spid = runtime.NO_SPID
else:
raise AssertionError()
try:
width = int(width_str)
except ValueError:
if width_spid == runtime.NO_SPID:
width_spid = part.width.span_id
self.errfmt.Print_("printf got invalid width %r" %
width_str,
span_id=width_spid)
return 1
precision = -1 # nonexistent
if part.precision:
if part.precision.id == Id.Format_Dot:
precision_str = '0'
precision_spid = part.precision.span_id
elif part.precision.id in (Id.Format_Num,
Id.Format_Zero):
precision_str = part.precision.val
precision_spid = part.precision.span_id
elif part.precision.id == Id.Format_Star:
if arg_index < num_args:
precision_str = varargs[arg_index]
precision_spid = spids[arg_index]
arg_index += 1
else:
precision_str = ''
precision_spid = runtime.NO_SPID
else:
raise AssertionError()
try:
precision = int(precision_str)
except ValueError:
if precision_spid == runtime.NO_SPID:
precision_spid = part.precision.span_id
self.errfmt.Print_(
'printf got invalid precision %r' %
precision_str,
span_id=precision_spid)
return 1
if arg_index < num_args:
s = varargs[arg_index]
word_spid = spids[arg_index]
arg_index += 1
has_arg = True
else:
s = ''
word_spid = runtime.NO_SPID
has_arg = False
typ = part.type.val
if typ == 's':
if precision >= 0:
s = s[:precision] # truncate
elif typ == 'q':
# TODO: most shells give \' for single quote, while OSH gives $'\''
# this could matter when SSH'ing
s = qsn.maybe_shell_encode(s)
elif typ == 'b':
# Process just like echo -e, except \c handling is simpler.
c_parts = [] # type: List[str]
lex = match.EchoLexer(s)
while True:
id_, tok_val = lex.Next()
if id_ == Id.Eol_Tok: # Note: This is really a NUL terminator
break
# TODO: add span_id from argv
tok = Token(id_, runtime.NO_SPID, tok_val)
p = word_compile.EvalCStringToken(tok)
# Unusual behavior: '\c' aborts processing!
if p is None:
backslash_c = True
break
c_parts.append(p)
s = ''.join(c_parts)
elif part.type.id == Id.Format_Time or typ in 'diouxX':
# %(...)T and %d share this complex integer conversion logic
try:
d = int(
s
) # note: spaces like ' -42 ' accepted and normalized
except ValueError:
# 'a is interpreted as the ASCII value of 'a'
if len(s) >= 1 and s[0] in '\'"':
# TODO: utf-8 decode s[1:] to be more correct. Probably
# depends on issue #366, a utf-8 library.
# Note: len(s) == 1 means there is a NUL (0) after the quote..
d = ord(s[1]) if len(s) >= 2 else 0
# No argument means -1 for %(...)T as in Bash Reference Manual
# 4.2 "If no argument is specified, conversion behaves as if -1
# had been given."
elif not has_arg and part.type.id == Id.Format_Time:
d = -1
else:
blame_spid = word_spid if has_arg else part.type.span_id
self.errfmt.Print_(
'printf expected an integer, got %r' % s,
span_id=blame_spid)
return 1
if part.type.id == Id.Format_Time:
# Initialize timezone:
# `localtime' uses the current timezone information initialized
# by `tzset'. The function `tzset' refers to the environment
# variable `TZ'. When the exported variable `TZ' is present,
# its value should be reflected in the real environment
# variable `TZ' before call of `tzset'.
#
# Note: unlike LANG, TZ doesn't seem to change behavior if it's
# not exported.
#
# TODO: In Oil, provide an API that doesn't rely on libc's
# global state.
tzcell = self.mem.GetCell('TZ')
if tzcell and tzcell.exported and tzcell.val.tag_(
) == value_e.Str:
tzval = cast(value__Str, tzcell.val)
posix.putenv('TZ', tzval.s)
time_.tzset()
# Handle special values:
# User can specify two special values -1 and -2 as in Bash
# Reference Manual 4.2: "Two special argument values may be
# used: -1 represents the current time, and -2 represents the
# time the shell was invoked." from
# https://www.gnu.org/software/bash/manual/html_node/Bash-Builtins.html#index-printf
if d == -1: # the current time
ts = time_.time()
elif d == -2: # the shell start time
ts = self.shell_start_time
else:
ts = d
s = time_.strftime(typ[1:-2], time_.localtime(ts))
if precision >= 0:
s = s[:precision] # truncate
else: # typ in 'diouxX'
# Disallowed because it depends on 32- or 64- bit
if d < 0 and typ in 'ouxX':
e_die(
"Can't format negative number %d with %%%s",
d,
typ,
span_id=part.type.span_id)
if typ == 'o':
s = mylib.octal(d)
elif typ == 'x':
s = mylib.hex_lower(d)
elif typ == 'X':
s = mylib.hex_upper(d)
else: # diu
s = str(d) # without spaces like ' -42 '
# There are TWO different ways to ZERO PAD, and they differ on
# the negative sign! See spec/builtin-printf
zero_pad = 0 # no zero padding
if width >= 0 and '0' in flags:
zero_pad = 1 # style 1
elif precision > 0 and len(s) < precision:
zero_pad = 2 # style 2
if zero_pad:
negative = (s[0] == '-')
if negative:
digits = s[1:]
sign = '-'
if zero_pad == 1:
# [%06d] -42 becomes [-00042] (6 TOTAL)
n = width - 1
else:
# [%6.6d] -42 becomes [-000042] (1 for '-' + 6)
n = precision
else:
digits = s
sign = ''
if zero_pad == 1:
n = width
else:
n = precision
s = sign + digits.rjust(n, '0')
else:
raise AssertionError()
if width >= 0:
if '-' in flags:
s = s.ljust(width, ' ')
else:
s = s.rjust(width, ' ')
out.append(s)
else:
raise AssertionError()
if backslash_c: # 'printf %b a\cb xx' - \c terminates processing!
break
if arg_index >= num_args:
break
# Otherwise there are more args. So cycle through the loop once more to
# implement the 'arg recycling' behavior.
return 0