python3: getting defined functions from a code object?

Question

In python3, I have the following code:

path = '/path/to/file/containing/python/code'
source = open(path, 'r').read()
codeobject = compile(source, path, 'exec')

I have examined codeobject, but I don't see any way to get a list of all the functions that are defined within that object.

I know I can search the source string for lines that begin with def, but I want to get this info from the code object, if at all possible.

What am I missing?

Answer 1

A code object is a nested structure; functions are created when the code object is executed, with their bodies embedded as separate code objects that are part of the constants:

>>> example = '''\
... def foobar():
...     print('Hello world!')
... '''
>>> codeobject = compile(example, '', 'exec')
>>> codeobject
<code object <module> at 0x11049ff60, file "", line 1>
>>> codeobject.co_consts
(<code object foobar at 0x11049fe40, file "", line 1>, 'foobar', None)
>>> codeobject.co_consts[0]
<code object foobar at 0x11049fe40, file "", line 1>
>>> codeobject.co_consts[0].co_name
'foobar'

When you disassemble the top-level code object you can see that the function objects are created from such code objects:

>>> import dis
>>> dis.dis(codeobject)
  1           0 LOAD_CONST               0 (<code object foobar at 0x11049fe40, file "", line 1>)
              2 LOAD_CONST               1 ('foobar')
              4 MAKE_FUNCTION            0
              6 STORE_NAME               0 (foobar)
              8 LOAD_CONST               2 (None)
             10 RETURN_VALUE

The MAKE_FUNCTION opcode takes the code object from the stack, as well as the function name and any default argument values from the stack; you can see the LOAD_CONST opcodes preceding it that put the code object and name there.

Not all code objects are functions however:

>>> compile('[i for i in range(10)]', '', 'exec').co_consts
(<code object <listcomp> at 0x1105cb030, file "", line 1>, '<listcomp>', 10, None)
>>> compile('class Foo: pass', '', 'exec').co_consts
(<code object Foo at 0x1105cb0c0, file "", line 1>, 'Foo', None)

If you wanted to list what functions are loaded in the bytecode, your best bet is to use the disassembly, not look for code objects:

import dis
from itertools import islice

# old itertools example to create a sliding window over a generator
def window(seq, n=2):
    """Returns a sliding window (of width n) over data from the iterable
       s -> (s0,s1,...s[n-1]), (s1,s2,...,sn), ...
    """
    it = iter(seq)
    result = tuple(islice(it, n))
    if len(result) == n:
        yield result    
    for elem in it:
        result = result[1:] + (elem,)
        yield result

def extract_functions(codeobject):
    codetype = type(codeobject)
    signature = ('LOAD_CONST', 'LOAD_CONST', 'MAKE_FUNCTION', 'STORE_NAME')
    for op1, op2, op3, op4 in window(dis.get_instructions(codeobject), 4):
        if (op1.opname, op2.opname, op3.opname, op4.opname) == signature:
            # Function loaded
            fname = op2.argval
            assert isinstance(op1.argval, codetype)
            yield fname, op1.argval

This generates (name, codeobject) tuples for all functions that are loaded in a given code object.