def decode(b, errors='strict'):
u, length = utf_8.decode(b, errors)
tokens = tokenize_rt.src_to_tokens(u)
new_tokens = []
for token in tokens:
if token.name == 'NUMBER':
new_tokens.extend(
tokenize_rt.src_to_tokens("blurse({})".format(token.src)))
else:
new_tokens.append(token)
return tokenize_rt.tokens_to_src(new_tokens), length
def _upgrade(source: str) -> str:
ast_obj = _ast_parse(source)
visitor = _FindAssignment()
visitor.visit(ast_obj)
tokens = src_to_tokens(source)
_mutate_found(tokens, visitor)
return tokens_to_src(tokens)
def _fix_six(contents_text):
try:
ast_obj = ast_parse(contents_text)
except SyntaxError:
return contents_text
visitor = FindSixUsage()
visitor.visit(ast_obj)
tokens = src_to_tokens(contents_text)
for i, token in reversed_enumerate(tokens):
if token.offset in visitor.simple_names:
node = visitor.simple_names[token.offset]
tokens[i] = Token('CODE', SIX_SIMPLE_ATTRS[node.id])
elif token.offset in visitor.simple_attrs:
node = visitor.simple_attrs[token.offset]
if tokens[i + 1].src == '.' and tokens[i + 2].src == node.attr:
tokens[i:i + 3] = [Token('CODE', SIX_SIMPLE_ATTRS[node.attr])]
elif token.offset in visitor.remove_decorators:
if tokens[i - 1].src == '@':
end = i + 1
while tokens[end].name != 'NEWLINE':
end += 1
del tokens[i - 1:end + 1]
return tokens_to_src(tokens)
def _has_trailing_semicolon(src: str) -> Tuple[str, bool]:
"""
Check if cell has trailing semicolon.
Parameters
----------
src
Notebook cell source.
Returns
-------
bool
Whether notebook has trailing semicolon.
"""
tokens = tokenize_rt.src_to_tokens(src)
trailing_semicolon = False
for idx, token in tokenize_rt.reversed_enumerate(tokens):
if not token.src.strip(" \n") or token.name == "COMMENT":
continue
if token.name == "OP" and token.src == ";":
tokens[idx] = token._replace(src="")
trailing_semicolon = True
break
if not trailing_semicolon:
return src, False
return tokenize_rt.tokens_to_src(tokens), True
def test_src_to_tokens_octal_literal_normalization():
ret = src_to_tokens('0755\n')
assert ret == [
Token('NUMBER', '0755', line=1, utf8_byte_offset=0),
Token('NEWLINE', '\n', line=1, utf8_byte_offset=4),
Token('ENDMARKER', '', line=2, utf8_byte_offset=0),
]
def decode(b, errors='strict'):
import tokenize_rt # pip install future-fstrings[rewrite]
u, length = utf_8.decode(b, errors)
tokens = tokenize_rt.src_to_tokens(u)
to_replace = []
start = end = seen_f = None
for i, token in enumerate(tokens):
if start is None:
if token.name == 'STRING':
start, end = i, i + 1
seen_f = _is_f(token)
elif token.name == 'STRING':
end = i + 1
seen_f |= _is_f(token)
elif token.name not in tokenize_rt.NON_CODING_TOKENS:
if seen_f:
to_replace.append((start, end))
start = end = seen_f = None
for start, end in reversed(to_replace):
try:
tokens[start:end] = _make_fstring(tokens[start:end])
except TokenSyntaxError as e:
msg = str(e.e)
line = u.splitlines()[e.token.line - 1]
bts = line.encode('UTF-8')[:e.token.utf8_byte_offset]
indent = len(bts.decode('UTF-8'))
raise SyntaxError(msg + '\n\n' + line + '\n' + ' ' * indent + '^')
return tokenize_rt.tokens_to_src(tokens), length
def _fix_src(contents_text: str, min_version: Tuple[int, ...]) -> str:
try:
ast_obj = ast_parse(contents_text)
except SyntaxError:
return contents_text
callbacks = visit(FUNCS, ast_obj, min_version)
tokens = src_to_tokens(contents_text)
for i, token in _changing_list(tokens):
# DEDENT is a zero length token
if not token.src:
continue
# though this is a defaultdict, by using `.get()` this function's
# self time is almost 50% faster
for callback in callbacks.get(token.offset, ()):
callback(i, tokens)
if token.src in START_BRACES:
fix_brace(
tokens,
find_simple(i, tokens),
add_comma=False,
remove_comma=False,
)
return tokens_to_src(tokens)
def _fix_format_literals(contents_text):
tokens = src_to_tokens(contents_text)
to_replace = []
string_start = None
string_end = None
seen_dot = False
for i, token in enumerate(tokens):
if string_start is None and token.name == 'STRING':
string_start = i
string_end = i + 1
elif string_start is not None and token.name == 'STRING':
string_end = i + 1
elif string_start is not None and token.src == '.':
seen_dot = True
elif seen_dot and token.src == 'format':
to_replace.append((string_start, string_end))
string_start, string_end, seen_dot = None, None, False
elif token.name not in NON_CODING_TOKENS:
string_start, string_end, seen_dot = None, None, False
for start, end in reversed(to_replace):
src = tokens_to_src(tokens[start:end])
new_src = _rewrite_string_literal(src)
tokens[start:end] = [Token('STRING', new_src)]
return tokens_to_src(tokens)
def _fix_fstrings(contents_text):
try:
ast_obj = ast_parse(contents_text)
except SyntaxError:
return contents_text
visitor = FindSimpleFormats()
visitor.visit(ast_obj)
tokens = src_to_tokens(contents_text)
for i, token in reversed(tuple(enumerate(tokens))):
node = visitor.found.get(Offset(token.line, token.utf8_byte_offset))
if node is None:
continue
if _is_bytestring(token.src): # pragma: no cover (py2-only)
continue
paren = i + 3
if tokens_to_src(tokens[i + 1:paren + 1]) != '.format(':
continue
# we don't actually care about arg position, so we pass `node`
victims = _victims(tokens, paren, node, gen=False)
end = victims.ends[-1]
# if it spans more than one line, bail
if tokens[end].line != token.line:
continue
tokens[i] = token._replace(src=_to_fstring(token.src, node))
del tokens[i + 1:end + 1]
return tokens_to_src(tokens)
def _fix_percent_format(contents_text):
try:
ast_obj = ast_parse(contents_text)
except SyntaxError:
return contents_text
visitor = FindPercentFormats()
visitor.visit(ast_obj)
if not visitor.found:
return contents_text
tokens = src_to_tokens(contents_text)
for i, token in reversed_enumerate(tokens):
node = visitor.found.get(token.offset)
if node is None:
continue
# no .format() equivalent for bytestrings in py3
# note that this code is only necessary when running in python2
if _is_bytestring(tokens[i].src): # pragma: no cover (py2-only)
continue
if isinstance(node.right, ast.Tuple):
_fix_percent_format_tuple(tokens, i, node)
elif isinstance(node.right, ast.Dict):
_fix_percent_format_dict(tokens, i, node)
return tokens_to_src(tokens)
def _fix_py2_compatible(contents_text):
try:
ast_obj = ast_parse(contents_text)
except SyntaxError:
return contents_text
visitor = Py2CompatibleVisitor()
visitor.visit(ast_obj)
if not any((
visitor.dicts,
visitor.sets,
visitor.set_empty_literals,
visitor.is_literal,
)):
return contents_text
tokens = src_to_tokens(contents_text)
for i, token in reversed_enumerate(tokens):
if token.offset in visitor.dicts:
_process_dict_comp(tokens, i, visitor.dicts[token.offset])
elif token.offset in visitor.set_empty_literals:
_process_set_empty_literal(tokens, i)
elif token.offset in visitor.sets:
_process_set_literal(tokens, i, visitor.sets[token.offset])
elif token.offset in visitor.is_literal:
_process_is_literal(tokens, i, visitor.is_literal[token.offset])
return tokens_to_src(tokens)
def _fix_plugins(contents_text: str, settings: Settings) -> str:
try:
ast_obj = ast_parse(contents_text)
except SyntaxError:
return contents_text
callbacks = visit(FUNCS, ast_obj, settings)
if not callbacks:
return contents_text
try:
tokens = src_to_tokens(contents_text)
except tokenize.TokenError: # pragma: no cover (bpo-2180)
return contents_text
_fixup_dedent_tokens(tokens)
for i, token in reversed_enumerate(tokens):
if not token.src:
continue
# though this is a defaultdict, by using `.get()` this function's
# self time is almost 50% faster
for callback in callbacks.get(token.offset, ()):
callback(i, tokens)
return tokens_to_src(tokens)
def remove_trailing_semicolon(src: str) -> Tuple[str, bool]:
"""Remove trailing semicolon from Jupyter notebook cell.
For example,
fig, ax = plt.subplots()
ax.plot(x_data, y_data); # plot data
would become
fig, ax = plt.subplots()
ax.plot(x_data, y_data) # plot data
Mirrors the logic in `quiet` from `IPython.core.displayhook`, but uses
``tokenize_rt`` so that round-tripping works fine.
"""
from tokenize_rt import (
src_to_tokens,
tokens_to_src,
reversed_enumerate,
)
tokens = src_to_tokens(src)
trailing_semicolon = False
for idx, token in reversed_enumerate(tokens):
if token.name in TOKENS_TO_IGNORE:
continue
if token.name == "OP" and token.src == ";":
del tokens[idx]
trailing_semicolon = True
break
if not trailing_semicolon:
return src, False
return tokens_to_src(tokens), True
def fix_file(filename: str, show_diff: bool = False, dry_run: bool = False) -> int:
with open(filename, 'rb') as f:
contents_bytes = f.read()
try:
contents_text = contents_bytes.decode()
except UnicodeDecodeError:
print(f'{filename} is non-utf8 (not supported)')
return 1
tokens = tokenize_rt.src_to_tokens(contents_text)
tokens_no_comments = _remove_comments(tokens)
src_no_comments = tokenize_rt.tokens_to_src(tokens_no_comments)
if src_no_comments == contents_text:
return 0
with tempfile.NamedTemporaryFile(
dir=os.path.dirname(filename),
prefix=os.path.basename(filename),
suffix='.py',
) as tmpfile:
tmpfile.write(src_no_comments.encode())
tmpfile.flush()
flake8_results = _run_flake8(tmpfile.name)
if any('E999' in v for v in flake8_results.values()):
print(f'{filename}: syntax error (skipping)')
return 0
for i, token in tokenize_rt.reversed_enumerate(tokens):
if token.name != 'COMMENT':
continue
if NOQA_RE.search(token.src):
_rewrite_noqa_comment(tokens, i, flake8_results)
elif NOQA_FILE_RE.match(token.src) and not flake8_results:
if i == 0 or tokens[i - 1].name == 'NEWLINE':
del tokens[i: i + 2]
else:
_remove_comment(tokens, i)
newsrc = tokenize_rt.tokens_to_src(tokens)
if newsrc != contents_text:
if (show_diff or dry_run):
diff = difflib.unified_diff(
contents_text.splitlines(keepends=True),
newsrc.splitlines(keepends=True),
fromfile=filename,
tofile=filename,
)
print(''.join(diff), end='')
if (not dry_run):
print(f'Rewriting {filename}')
with open(filename, 'wb') as f:
f.write(newsrc.encode())
return 1
else:
return 0
def _fix_tokens(contents_text: str, min_version: Version) -> str:
remove_u = (min_version >= (3, )
or _imports_future(contents_text, 'unicode_literals'))
try:
tokens = src_to_tokens(contents_text)
except tokenize.TokenError:
return contents_text
for i, token in reversed_enumerate(tokens):
if token.name == 'NUMBER':
tokens[i] = token._replace(src=_fix_long(_fix_octal(token.src)))
elif token.name == 'STRING':
tokens[i] = _fix_ur_literals(tokens[i])
if remove_u:
tokens[i] = _remove_u_prefix(tokens[i])
tokens[i] = _fix_escape_sequences(tokens[i])
elif token.src == '(':
_fix_extraneous_parens(tokens, i)
elif token.src == 'format' and i > 0 and tokens[i - 1].src == '.':
_fix_format_literal(tokens, i - 2)
elif token.src == 'encode' and i > 0 and tokens[i - 1].src == '.':
_fix_encode_to_binary(tokens, i)
elif (min_version >= (3, ) and token.utf8_byte_offset == 0
and token.line < 3 and token.name == 'COMMENT'
and tokenize.cookie_re.match(token.src)):
del tokens[i]
assert tokens[i].name == 'NL', tokens[i].name
del tokens[i]
elif token.src == 'from' and token.utf8_byte_offset == 0:
_fix_import_removals(tokens, i, min_version)
return tokens_to_src(tokens).lstrip()
def test_src_to_tokens_string_prefix_normalization(prefix):
src = f"{prefix}'foo'\n"
ret = src_to_tokens(src)
assert ret == [
Token('STRING', f"{prefix}'foo'", line=1, utf8_byte_offset=0),
Token('NEWLINE', '\n', line=1, utf8_byte_offset=5 + len(prefix)),
Token('ENDMARKER', '', line=2, utf8_byte_offset=0),
]
def test_src_to_tokens_long_literal_normalization(postfix):
src = f'123{postfix}\n'
ret = src_to_tokens(src)
assert ret == [
Token('NUMBER', f'123{postfix}', line=1, utf8_byte_offset=0),
Token('NEWLINE', '\n', line=1, utf8_byte_offset=4),
Token('ENDMARKER', '', line=2, utf8_byte_offset=0),
]
def _fix_calls(contents_text: str) -> str:
try:
ast_obj = ast_parse(contents_text)
except SyntaxError:
return contents_text
visitor = Visitor()
visitor.visit(ast_obj)
if not visitor.calls:
return contents_text
try:
tokens = src_to_tokens(contents_text)
except tokenize.TokenError: # pragma: no cover (bpo-2180)
return contents_text
for i, token in reversed_enumerate(tokens):
if token.offset in visitor.calls:
visitor.calls.discard(token.offset)
# search forward for the opening brace
while tokens[i].src != '(':
i += 1
call_start = i
i += 1
brace_depth = 1
start = -1
end = -1
while brace_depth:
if tokens[i].src in {'(', '{', '['}:
if brace_depth == 1:
start = i
brace_depth += 1
elif tokens[i].src in {')', '}', ']'}:
brace_depth -= 1
if brace_depth == 1:
end = i
i += 1
assert start != -1
assert end != -1
call_end = i - 1
# dedent everything inside the brackets
for i in range(call_start, call_end):
if (tokens[i - 1].name == 'NL'
and tokens[i].name == UNIMPORTANT_WS):
tokens[i] = tokens[i]._replace(src=tokens[i].src[4:])
del tokens[end + 1:call_end]
del tokens[call_start + 1:start]
return tokens_to_src(tokens)
def decode(b, errors="strict"):
non_coding_tokens = frozenset(
("COMMENT", tokenize_rt.ESCAPED_NL, "NL", tokenize_rt.UNIMPORTANT_WS))
u, length = utf_8.decode(b, errors)
tokens = tokenize_rt.src_to_tokens(u)
to_replace = []
started = -1
end = -1
for i in range(0, 1 + len(tokens)):
if i < len(tokens):
token = tokens[i]
else:
token = None
if token:
if fstr(token):
if started < 0:
started = i
continue
end = i
if started >= 0:
if peek_is_fstr(tokens, i + 1):
continue
if peek_is_str(tokens, i + 1):
continue
if token is None:
pass
elif token.name in non_coding_tokens or peek_is_str(tokens, i):
#multiline f-string+str
continue
to_replace.append((started, end))
started = -1
for start, end in reversed(to_replace):
if end - start > 1:
#move ending line away from format of multiline fstrings
if tokens[end - 1].name in non_coding_tokens:
end -= 1
try:
tokens[start:end] = _make_fstring(tokens[start:end])
except TokenSyntaxError as e:
msg = str(e.e)
line = u.splitlines()[e.token.line - 1]
bts = line.encode("UTF-8")[:e.token.utf8_byte_offset]
indent = len(bts.decode("UTF-8"))
raise SyntaxError(msg + "\n\n" + line + "\n" + " " * indent + "^")
return tokenize_rt.tokens_to_src(tokens), length
def fix_file(filename: str) -> int:
with open(filename, 'rb') as f:
contents_bytes = f.read()
try:
contents_text = contents_bytes.decode()
except UnicodeDecodeError:
print(f'{filename} is non-utf8 (not supported)')
return 1
tokens = tokenize_rt.src_to_tokens(contents_text)
tokens_no_comments = _remove_comments(tokens)
src_no_comments = tokenize_rt.tokens_to_src(tokens_no_comments)
if src_no_comments == contents_text:
return 0
fd, path = tempfile.mkstemp(
dir=os.path.dirname(filename),
prefix=os.path.basename(filename),
suffix='.py',
)
try:
with open(fd, 'wb') as f:
f.write(src_no_comments.encode())
flake8_results = _run_flake8(path)
finally:
os.remove(path)
if any('E999' in v for v in flake8_results.values()):
print(f'{filename}: syntax error (skipping)')
return 0
for i, token in tokenize_rt.reversed_enumerate(tokens):
if token.name != 'COMMENT':
continue
if NOQA_RE.search(token.src):
_rewrite_noqa_comment(tokens, i, flake8_results)
elif NOQA_FILE_RE.match(token.src) and not flake8_results:
if i == 0 or tokens[i - 1].name == 'NEWLINE':
del tokens[i:i + 2]
else:
_remove_comment(tokens, i)
newsrc = tokenize_rt.tokens_to_src(tokens)
if newsrc != contents_text:
print(f'Rewriting {filename}')
with open(filename, 'wb') as f:
f.write(newsrc.encode())
return 1
else:
return 0
def test_src_to_tokens_simple():
src = 'x = 5\n'
ret = src_to_tokens(src)
assert ret == [
Token('NAME', 'x', line=1, utf8_byte_offset=0),
Token(UNIMPORTANT_WS, ' ', line=None, utf8_byte_offset=None),
Token('OP', '=', line=1, utf8_byte_offset=2),
Token(UNIMPORTANT_WS, ' ', line=None, utf8_byte_offset=None),
Token('NUMBER', '5', line=1, utf8_byte_offset=4),
Token('NEWLINE', '\n', line=1, utf8_byte_offset=5),
Token('ENDMARKER', '', line=2, utf8_byte_offset=0),
]
def _fix_octal_literals(contents_text):
def _fix_octal(s):
if not s.startswith('0') or not s.isdigit() or s == len(s) * '0':
return s
else: # pragma: no cover (py2 only)
return '0o' + s[1:]
tokens = src_to_tokens(contents_text)
for i, token in enumerate(tokens):
if token.name == 'NUMBER':
tokens[i] = token._replace(src=_fix_octal(token.src))
return tokens_to_src(tokens)
def test_src_to_tokens_escaped_nl_no_left_ws():
src = ('x =\\\n' ' 5\n')
ret = src_to_tokens(src)
assert ret == [
Token('NAME', 'x', line=1, utf8_byte_offset=0),
Token(UNIMPORTANT_WS, ' ', line=None, utf8_byte_offset=None),
Token('OP', '=', line=1, utf8_byte_offset=2),
Token(ESCAPED_NL, '\\\n', line=None, utf8_byte_offset=None),
Token(UNIMPORTANT_WS, ' ', line=None, utf8_byte_offset=None),
Token('NUMBER', '5', line=2, utf8_byte_offset=4),
Token('NEWLINE', '\n', line=2, utf8_byte_offset=5),
Token('ENDMARKER', '', line=3, utf8_byte_offset=0),
]
def _fix_unicode_literals(contents_text, py3_plus):
if not py3_plus and not _imports_unicode_literals(contents_text):
return contents_text
tokens = src_to_tokens(contents_text)
for i, token in enumerate(tokens):
if token.name != 'STRING':
continue
match = STRING_PREFIXES_RE.match(token.src)
prefix = match.group(1)
rest = match.group(2)
new_prefix = prefix.replace('u', '').replace('U', '')
tokens[i] = Token('STRING', new_prefix + rest)
return tokens_to_src(tokens)
def test_reversed_enumerate():
tokens = src_to_tokens('x = 5\n')
ret = reversed_enumerate(tokens)
assert next(ret) == (6, Token('ENDMARKER', '', line=2, utf8_byte_offset=0))
rest = list(ret)
assert rest == [
(5, Token(name='NEWLINE', src='\n', line=1, utf8_byte_offset=5)),
(4, Token('NUMBER', '5', line=1, utf8_byte_offset=4)),
(3, Token(UNIMPORTANT_WS, ' ')),
(2, Token('OP', '=', line=1, utf8_byte_offset=2)),
(1, Token(UNIMPORTANT_WS, ' ')),
(0, Token('NAME', 'x', line=1, utf8_byte_offset=0)),
]
def _fix_dictcomps(contents_text):
try:
ast_obj = ast_parse(contents_text)
except SyntaxError:
return contents_text
visitor = FindDictsVisitor()
visitor.visit(ast_obj)
if not visitor.dicts:
return contents_text
tokens = src_to_tokens(contents_text)
for i, token in reversed_enumerate(tokens):
if token.offset in visitor.dicts:
_process_dict_comp(tokens, i, visitor.dicts[token.offset])
return tokens_to_src(tokens)
def _fix_dictcomps(contents_text):
try:
ast_obj = ast_parse(contents_text)
except SyntaxError:
return contents_text
visitor = FindDictsVisitor()
visitor.visit(ast_obj)
if not visitor.dicts:
return contents_text
tokens = src_to_tokens(contents_text)
for i, token in reversed(tuple(enumerate(tokens))):
key = (token.line, token.utf8_byte_offset)
if key in visitor.dicts:
_process_dict_comp(tokens, i, visitor.dicts[key])
return tokens_to_src(tokens)
def _fix_sets(contents_text):
try:
ast_obj = ast_parse(contents_text)
except SyntaxError:
return contents_text
visitor = FindSetsVisitor()
visitor.visit(ast_obj)
if not visitor.sets and not visitor.set_empty_literals:
return contents_text
tokens = src_to_tokens(contents_text)
for i, token in reversed_enumerate(tokens):
if token.offset in visitor.set_empty_literals:
_process_set_empty_literal(tokens, i)
elif token.offset in visitor.sets:
_process_set_literal(tokens, i, visitor.sets[token.offset])
return tokens_to_src(tokens)
def _fix_src(contents_text, py35_plus, py36_plus):
try:
ast_obj = ast_parse(contents_text)
except SyntaxError:
return contents_text
visitor = FindNodes()
visitor.visit(ast_obj)
tokens = src_to_tokens(contents_text)
for i, token in _changing_list(tokens):
# DEDENT is a zero length token
if not token.src:
continue
key = Offset(token.line, token.utf8_byte_offset)
fixes = []
if key in visitor.calls:
for call in visitor.calls[key]:
# Only fix stararg calls if asked to
add_comma = not call.star_args or py35_plus
fixes.append((add_comma, _find_call(call, i, tokens)))
elif key in visitor.funcs:
func = visitor.funcs[key]
add_comma = not func.star_args or py36_plus
# functions can be treated as calls
fixes.append((add_comma, _find_call(func, i, tokens)))
elif key in visitor.literals:
fixes.append((True, _find_simple(i, tokens)))
# Handle parenthesized things, unhug of tuples, and comprehensions
elif token.src in START_BRACES:
fixes.append((False, _find_simple(i, tokens)))
for add_comma, fix_data in fixes:
if fix_data is not None:
_fix_brace(fix_data, add_comma, tokens)
# need to handle tuples afterwards as tuples report their starting
# starting index as the first element, which may be one of the above
# things.
if key in visitor.tuples:
fix_data = _find_tuple(i, tokens)
if fix_data is not None:
_fix_brace(fix_data, True, tokens)
return tokens_to_src(tokens)
def _fix_escape_sequences(contents_text):
last_name = None
tokens = src_to_tokens(contents_text)
for i, token in enumerate(tokens):
if token.name == 'NAME':
last_name = token
continue
elif token.name != 'STRING':
last_name = None
continue
match = STRING_PREFIXES_RE.match(token.src)
prefix = match.group(1)
rest = match.group(2)
if last_name is not None: # pragma: no cover (py2 bug)
actual_prefix = (last_name.src + prefix).lower()
else: # pragma: no cover (py3 only)
actual_prefix = prefix.lower()
if 'r' in actual_prefix or '\\' not in rest:
continue
if 'b' in actual_prefix:
valid_escapes = ESCAPE_STARTS_BYTES
else:
valid_escapes = ESCAPE_STARTS
escape_sequences = {m[1] for m in ESCAPE_RE.findall(rest)}
has_valid_escapes = escape_sequences & valid_escapes
has_invalid_escapes = escape_sequences - valid_escapes
def cb(match):
matched = match.group()
if matched[1] in valid_escapes:
return matched
else:
return r'\{}'.format(matched)
if has_invalid_escapes and (has_valid_escapes or 'u' in actual_prefix):
tokens[i] = token._replace(src=prefix + ESCAPE_RE.sub(cb, rest))
elif has_invalid_escapes and not has_valid_escapes:
tokens[i] = token._replace(src=prefix + 'r' + rest)
return tokens_to_src(tokens)