gh-126072: do not add None to co_consts if there is no docstring (GH-126101)

Author: xuantengh

Doc/library/inspect.rst

@@ -1700,6 +1700,14 @@ which is a bitmap of the following flags:
 
    .. versionadded:: 3.6
 
+.. data:: CO_HAS_DOCSTRING
+
+   The flag is set when there is a docstring for the code object in
+   the source code. If set, it will be the first item in
+   :attr:`~codeobject.co_consts`.
+
+   .. versionadded:: 3.14
+
 .. note::
    The flags are specific to CPython, and may not be defined in other
    Python implementations.  Furthermore, the flags are an implementation

Doc/reference/datamodel.rst

@@ -1536,9 +1536,9 @@ Other bits in :attr:`~codeobject.co_flags` are reserved for internal use.
 
 .. index:: single: documentation string
 
-If a code object represents a function, the first item in
-:attr:`~codeobject.co_consts` is
-the documentation string of the function, or ``None`` if undefined.
+If a code object represents a function and has a docstring,
+the first item in :attr:`~codeobject.co_consts` is
+the docstring of the function.
 
 Methods on code objects
 ~~~~~~~~~~~~~~~~~~~~~~~

Include/cpython/code.h

@@ -174,6 +174,11 @@ struct PyCodeObject _PyCode_DEF(1);
 
 #define CO_NO_MONITORING_EVENTS 0x2000000
 
+/* Whether the code object has a docstring,
+   If so, it will be the first item in co_consts
+*/
+#define CO_HAS_DOCSTRING 0x4000000
+
 /* This should be defined if a future statement modifies the syntax.
    For example, when a keyword is added.
 */

Include/internal/pycore_symtable.h

@@ -123,6 +123,7 @@ typedef struct _symtable_entry {
     unsigned ste_comp_iter_target : 1; /* true if visiting comprehension target */
     unsigned ste_can_see_class_scope : 1; /* true if this block can see names bound in an
                                              enclosing class scope */
+    unsigned ste_has_docstring : 1; /* true if docstring present */
     int ste_comp_iter_expr; /* non-zero if visiting a comprehension range expression */
     _Py_SourceLocation ste_loc; /* source location of block */
     struct _symtable_entry *ste_annotation_block; /* symbol table entry for this entry's annotations */

Lib/dis.py

@@ -151,16 +151,17 @@ def distb(tb=None, *, file=None, show_caches=False, adaptive=False, show_offsets
 # list of CO_* constants. It is also used by pretty_flags to
 # turn the co_flags field into a human readable list.
 COMPILER_FLAG_NAMES = {
-     1: "OPTIMIZED",
-     2: "NEWLOCALS",
-     4: "VARARGS",
-     8: "VARKEYWORDS",
-    16: "NESTED",
-    32: "GENERATOR",
-    64: "NOFREE",
-   128: "COROUTINE",
-   256: "ITERABLE_COROUTINE",
-   512: "ASYNC_GENERATOR",
+            1: "OPTIMIZED",
+            2: "NEWLOCALS",
+            4: "VARARGS",
+            8: "VARKEYWORDS",
+           16: "NESTED",
+           32: "GENERATOR",
+           64: "NOFREE",
+          128: "COROUTINE",
+          256: "ITERABLE_COROUTINE",
+          512: "ASYNC_GENERATOR",
+    0x4000000: "HAS_DOCSTRING",
 }
 
 def pretty_flags(flags):

Lib/inspect.py

@@ -56,6 +56,7 @@
     "CO_OPTIMIZED",
     "CO_VARARGS",
     "CO_VARKEYWORDS",
+    "CO_HAS_DOCSTRING",
     "ClassFoundException",
     "ClosureVars",
     "EndOfBlock",
@@ -409,6 +410,7 @@ def iscode(object):
         co_flags            bitmap: 1=optimized | 2=newlocals | 4=*arg | 8=**arg
                             | 16=nested | 32=generator | 64=nofree | 128=coroutine
                             | 256=iterable_coroutine | 512=async_generator
+                            | 0x4000000=has_docstring
         co_freevars         tuple of names of free variables
         co_posonlyargcount  number of positional only arguments
         co_kwonlyargcount   number of keyword only arguments (not including ** arg)

Lib/test/test_code.py

@@ -17,7 +17,7 @@
 freevars: ()
 nlocals: 2
 flags: 3
-consts: ('None', '<code object g>')
+consts: ('<code object g>',)
 
 >>> dump(f(4).__code__)
 name: g
@@ -86,7 +86,7 @@
 cellvars: ()
 freevars: ()
 nlocals: 0
-flags: 3
+flags: 67108867
 consts: ("'doc string'", 'None')
 
 >>> def keywordonly_args(a,b,*,k1):
@@ -123,6 +123,61 @@
 flags: 3
 consts: ('None',)
 
+>>> def has_docstring(x: str):
+...     'This is a one-line doc string'
+...     x += x
+...     x += "hello world"
+...     # co_flags should be 0x4000003 = 67108867
+...     return x
+
+>>> dump(has_docstring.__code__)
+name: has_docstring
+argcount: 1
+posonlyargcount: 0
+kwonlyargcount: 0
+names: ()
+varnames: ('x',)
+cellvars: ()
+freevars: ()
+nlocals: 1
+flags: 67108867
+consts: ("'This is a one-line doc string'", "'hello world'")
+
+>>> async def async_func_docstring(x: str, y: str):
+...     "This is a docstring from async function"
+...     import asyncio
+...     await asyncio.sleep(1)
+...     # co_flags should be 0x4000083 = 67108995
+...     return x + y
+
+>>> dump(async_func_docstring.__code__)
+name: async_func_docstring
+argcount: 2
+posonlyargcount: 0
+kwonlyargcount: 0
+names: ('asyncio', 'sleep')
+varnames: ('x', 'y', 'asyncio')
+cellvars: ()
+freevars: ()
+nlocals: 3
+flags: 67108995
+consts: ("'This is a docstring from async function'", 'None')
+
+>>> def no_docstring(x, y, z):
+...     return x + "hello" + y + z + "world"
+
+>>> dump(no_docstring.__code__)
+name: no_docstring
+argcount: 3
+posonlyargcount: 0
+kwonlyargcount: 0
+names: ()
+varnames: ('x', 'y', 'z')
+cellvars: ()
+freevars: ()
+nlocals: 3
+flags: 3
+consts: ("'hello'", "'world'")
 """
 
 import copy

Lib/test/test_compile.py

@@ -834,7 +834,7 @@ def f():
                 return "unused"
 
         self.assertEqual(f.__code__.co_consts,
-                         (None, "used"))
+                         (True, "used"))
 
     @support.cpython_only
     def test_remove_unused_consts_extended_args(self):
@@ -852,9 +852,9 @@ def test_remove_unused_consts_extended_args(self):
         eval(compile(code, "file.py", "exec"), g)
         exec(code, g)
         f = g['f']
-        expected = tuple([None, ''] + [f't{i}' for i in range(N)])
+        expected = tuple([''] + [f't{i}' for i in range(N)])
         self.assertEqual(f.__code__.co_consts, expected)
-        expected = "".join(expected[2:])
+        expected = "".join(expected[1:])
         self.assertEqual(expected, f())
 
     # Stripping unused constants is not a strict requirement for the
@@ -1244,7 +1244,7 @@ def return_genexp():
                     y)
         genexp_lines = [0, 4, 2, 0, 4]
 
-        genexp_code = return_genexp.__code__.co_consts[1]
+        genexp_code = return_genexp.__code__.co_consts[0]
         code_lines = self.get_code_lines(genexp_code)
         self.assertEqual(genexp_lines, code_lines)
 

Lib/test/test_compiler_assemble.py

@@ -84,7 +84,7 @@ def inner():
                 return x
             return inner() % 2
 
-        inner_code = mod_two.__code__.co_consts[1]
+        inner_code = mod_two.__code__.co_consts[0]
         assert isinstance(inner_code, types.CodeType)
 
         metadata = {