bpo-38870: fixing unhandled hexescape in docstrings at ast.unparse

Author: 1.v3m

Lib/ast.py

@@ -1088,18 +1088,27 @@ def visit_Name(self, node):
         self.write(node.id)
 
     def _write_docstring(self, node):
+        def esc_char(c):
+            if c in ("\n", "\t"):
+                # In the AST form, we don't know the author's intentation
+                # about how this should be displayed. We'll only escape
+                # \n and \t, because they are more likely to be unescaped
+                # in the source
+                return c
+            else:
+                return c.encode('unicode_escape').decode('ascii')
+
         self.fill()
         if node.kind == "u":
             self.write("u")
 
         value = node.value
         if value:
             # Preserve quotes in the docstring by escaping them
-            value = value.replace("\\", "\\\\")
-            value = value.replace('"""', '""\"')
-            value = value.replace("\r", "\\r")
+            value = "".join(map(esc_char, value))
             if value[-1] == '"':
                 value = value.replace('"', '\\"', -1)
+            value = value.replace('"""', '""\\"')
 
         self.write(f'"""{value}"""')
 

Lib/test/test_unparse.py

@@ -324,7 +324,11 @@ def test_docstrings(self):
             '\\t',
             '\n',
             '\\n',
-            '\r\\r\t\\t\n\\n'
+            '\r\\r\t\\t\n\\n',
+            '""">>> content = \"\"\"blabla\"\"\" <<<"""',
+            r'foo\n\x00'
+            '🐍⛎𩸽üéş^\X\BB\N{LONG RIGHTWARDS SQUIGGLE ARROW}'
+
         )
         for docstring in docstrings:
             # check as Module docstrings for easy testing