Sometimes, map is a sequence, sometimes not?

Question

In python3, when I try to execute the following lines:

$ python3
Python 3.3.5 (default, Mar 18 2014, 02:00:02) 
[GCC 4.2.1 20070831 patched [FreeBSD]] on freebsd9
Type "help", "copyright", "credits" or "license" for more information.
>>> def fail(arg):
...   raise Exception
... 
>>> def fail_type(arg):
...   raise TypeError
... 
>>> identity = lambda x: x
>>> m = [0,1]
>>> print("{:8.2f}{:8.2f}".format(*map(identity, m)))
    0.00    1.00
>>> print("{:8.2f}{:8.2f}".format(*map(fail, m)))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "<stdin>", line 2, in fail
Exception
>>> print("{:8.2f}{:8.2f}".format(*map(fail_type, m)))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: format() argument after * must be a sequence, not map
>>> list(map(identity, m))
[0, 1]
>>> list(map(fail, m))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "<stdin>", line 2, in fail
Exception
>>> list(map(fail_type, m))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "<stdin>", line 2, in fail_type
TypeError
>>> 

Why does python throw that in my opinion highly misleading TypeError: format() argument after * must be a sequence, not map for this builtin function? It makes it even more confusing to see that an explicit raise for other Exceptions is caught an passed through. Is there an assumption in * error handling that a TypeError must be due to the argumentlist not being a sequence?

Answer 1

The TypeError exception match occurs in ext_do_call (Python/ceval.c, line 4507), which is called to handle the opcodes CALL_FUNCTION_VAR, CALL_FUNCTION_KW, and CALL_FUNCTION_VAR_KW. It's assuming PySequence_Tuple raised the TypeError because the argument wasn't iterable, as is raised by PyObject_GetIter.

There is an opened Python bug in "patch review" stage for 3 years: Function calls taking a generator as star argument can mask TypeErrors in the generator