[Pyrex] Subclassing - design question

[Helmut Jarausch]

Please note anywhere in the docs, that
the __new__ method of a class which
should be subclassable should have
only a single required parameter and
a placeholder for additional ones.

E.g. the corrected version the failing example 

cdef class B:
  cdef int Seen
  def __new__(self,*dummy):
    self.Seen=17

cdef class D(B):
  def __new__(self,int DS):
    self.Seen= DS

just works fine.
Helmut.



On  8 Mar, Helmut Jarausch wrote:
> Hi,
> 
> given
> 
> cdef class B:
>   cdef int Seen
>   def __new__(self):
>     self.Seen=17
> 
> cdef class D(B):
>   def __new__(self,int DS):
>     .....
> 
> In Python the __init__ method of a base class must be called
> explicity (similarly to the base class initialization call
> in a C++ constructor)
> This is necessary since the base' __init__ can be called
> with different arguments.
> 
> But in Pyrex the __new__ method of the base class is automatically
> called by the python runtime library, so Pyrex can only fill 
> ONE possible call into the PyTypeObject table.
> 
> Currently, function generate_new_function (if base_type:) of Nodes.py
> just copies it's own argument list. So, IHMO, Pyrex would need a
> new syntactic possiblity to let the user specify the arguments going
> to the base class.
> 
> This shoe is too big for me (literal translation of a German saying)
> 
> Helmut.
> 

-- 
Helmut Jarausch

Lehrstuhl fuer Numerische Mathematik
RWTH - Aachen University
D 52056 Aachen, Germany