ast.unparse doesn't observe the new PEP701 string delimiter rules

[tonybaloney]

Bug report

Checklist

• I am confident this is a bug in CPython, not a bug in a third-party project
• I have searched the CPython issue tracker,
  and am confident this bug has not been reported before

CPython versions tested on:

3.12, CPython main branch

Operating systems tested on:

macOS

Output from running 'python -VV' on the command line:

Python 3.12.0b4 (v3.12.0b4:97a6a41816, Jul 11 2023, 11:19:02) [Clang 13.0.0 (clang-1300.0.29.30)] on darwin

A clear and concise description of the bug:

The ast.unparse Unparser doesn't seem to respect PEP701, for example, if you use double quotes in an f-string then unparse the AST it will use a backslash--

>>> import ast
>>> code = 'f" something { my_dict["key"] } something else "'
>>> tree = ast.parse(code)
>>> ast.unparse(tree)
'f" something {my_dict[\'key\']} something else "'

Furthermore, if you use the nested f-string example in the PEP, it crashes completely when unparsing the AST

>>> f"{f"{f"{f"{f"{f"{1+1}"}"}"}"}"}"
'2'
>>> code = 'f"{f"{f"{f"{f"{f"{1+1}"}"}"}"}"}"'
>>> import ast
>>> tree = ast.parse(code)
>>> tree
<ast.Module object at 0x10992fed0>
>>> ast.unparse(tree)
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
 File "/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/ast.py", line 1777, in unparse
 return unparser.visit(ast_obj)
 ^^^^^^^^^^^^^^^^^^^^^^^
 File "/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/ast.py", line 859, in visit
 self.traverse(node)
 File "/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/ast.py", line 850, in traverse
 super().visit(node)
 File "/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/ast.py", line 407, in visit
 return visitor(node)
 ^^^^^^^^^^^^^
 File "/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/ast.py", line 874, in visit_Module
 self._write_docstring_and_traverse_body(node)
 File "/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/ast.py", line 867, in _write_docstring_and_traverse_body
 self.traverse(node.body)
 File "/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/ast.py", line 848, in traverse
 self.traverse(item)
 File "/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/ast.py", line 850, in traverse
 super().visit(node)
 File "/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/ast.py", line 407, in visit
 return visitor(node)
 ^^^^^^^^^^^^^
 File "/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/ast.py", line 889, in visit_Expr
 self.traverse(node.value)
 File "/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/ast.py", line 850, in traverse
 super().visit(node)
 File "/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/ast.py", line 407, in visit
 return visitor(node)
 ^^^^^^^^^^^^^
 File "/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/ast.py", line 1240, in visit_JoinedStr
 self._write_fstring_inner(value)
 File "/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/ast.py", line 1268, in _write_fstring_inner
 self.visit_FormattedValue(node)
 File "/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/ast.py", line 1281, in visit_FormattedValue
 raise ValueError(
ValueError: Unable to avoid backslash in f-string expression part

Linked PRs

• gh-108469: Update ast.unparse for unescaped quote support from PEP701 [3.12] #108553
• [3.12] gh-108469: Update ast.unparse for unescaped quote support from PEP701 [3.12] (GH-108553) #108960

[tonybaloney]

CC @pablogsal @isidentical

[pablogsal]

@isidentical can you take a look?

[tonybaloney]

Related TODO https://github.com/python/cpython/blame/main/Lib/test/test_tokenize.py#L1862-L1865

[isidentical]

@isidentical can you take a look?

Yep!

[isidentical]

@tonybaloney just saw the #108553, are you interested on providing a fix in the PR too or is it a test case PR? I'd be happy to help / answer questions regarding the old f-string unparsing logic if you have any.

cpython/Lib/ast.py

Lines 1226 to 1260 in 4116592

	def visit_JoinedStr(self, node):
	self.write("f")
	if self._avoid_backslashes:
	with self.buffered() as buffer:
	self._write_fstring_inner(node)
	return self._write_str_avoiding_backslashes("".join(buffer))
	# If we don't need to avoid backslashes globally (i.e., we only need
	# to avoid them inside FormattedValues), it's cosmetically preferred
	# to use escaped whitespace. That is, it's preferred to use backslashes
	# for cases like: f"{x}\n". To accomplish this, we keep track of what
	# in our buffer corresponds to FormattedValues and what corresponds to
	# Constant parts of the f-string, and allow escapes accordingly.
	fstring_parts = []
	for value in node.values:
	with self.buffered() as buffer:
	self._write_fstring_inner(value)
	fstring_parts.append(
	("".join(buffer), isinstance(value, Constant))
	)
	new_fstring_parts = []
	quote_types = list(_ALL_QUOTES)
	for value, is_constant in fstring_parts:
	value, quote_types = self._str_literal_helper(
	value,
	quote_types=quote_types,
	escape_special_whitespace=is_constant,
	)
	new_fstring_parts.append(value)
	value = "".join(new_fstring_parts)
	quote_type = quote_types[0]
	self.write(f"{quote_type}{value}{quote_type}")

[tonybaloney]

@tonybaloney just saw the #108553, are you interested on providing a fix in the PR too or is it a test case PR? I'd be happy to help / answer questions regarding the old f-string unparsing logic if you have any.

cpython/Lib/ast.py

Lines 1226 to 1260 in 4116592

	def visit_JoinedStr(self, node):
	self.write("f")
	if self._avoid_backslashes:
	with self.buffered() as buffer:
	self._write_fstring_inner(node)
	return self._write_str_avoiding_backslashes("".join(buffer))
	# If we don't need to avoid backslashes globally (i.e., we only need
	# to avoid them inside FormattedValues), it's cosmetically preferred
	# to use escaped whitespace. That is, it's preferred to use backslashes
	# for cases like: f"{x}\n". To accomplish this, we keep track of what
	# in our buffer corresponds to FormattedValues and what corresponds to
	# Constant parts of the f-string, and allow escapes accordingly.
	fstring_parts = []
	for value in node.values:
	with self.buffered() as buffer:
	self._write_fstring_inner(value)
	fstring_parts.append(
	("".join(buffer), isinstance(value, Constant))
	)
	new_fstring_parts = []
	quote_types = list(_ALL_QUOTES)
	for value, is_constant in fstring_parts:
	value, quote_types = self._str_literal_helper(
	value,
	quote_types=quote_types,
	escape_special_whitespace=is_constant,
	)
	new_fstring_parts.append(value)
	value = "".join(new_fstring_parts)
	quote_type = quote_types[0]
	self.write(f"{quote_type}{value}{quote_type}")

@isidentical I'd be happy to submit a PR. I looked at the code and couldn't really work out what it was doing or why though. There are some private methods for escaping strings or using a different string delimiter which I guess would be redundant from 3.12 but couldn't see if that was the only change needed

[isidentical]

There are some private methods for escaping strings or using a different string delimiter which I guess would be redundant from 3.12 but couldn't see if that was the only change needed

I think so. Most of that code was trying to get rid of backslashes and figure out what quote we can use for nested f-strings which doesn't create a problem anymore.

[sunmy2019]

We'd better move fast. We won't make it to the Sept 4th release. Hurry for Oct 2rd release.