[Edu-sig] Learn to Program in Ten Years

Kirby Urner urnerk at qwest.net
Sun Dec 26 01:02:07 CET 2004


> Other finds from a day of lazy exploration (us Jews have time on
> Christmas):
> 

Tell me about it.  I've been napping, browsing... even added a "me" face to
my blog (taken aboard a roller coaster of some kind).

>

SKIPPING DOWN -- lot's of interest, especially Lush...

> It is interpreted when given untyped variable information, and compiled
> when given typed variables.  And works nicely with inline C code.
> 
> Wouldn't that be nice if Python...
> 

I think this idea of a Core Python that doesn't go with typing, makes a lot
of sense.  Then there could be this annotation which, applied to Python,
would take it into static.  Maybe even some special commenting would do the
trick, e.g.:

  i = 3 # int
  j = 4 # int
  
Probably not though.  I just don't want Core Python to sprout all these
declarations, because the whole point of late binding is you can't really
know -- the code makes up new types at runtime, so they can't have been
declared (pre runtime, there wouldn't be any words for 'em).  Something like
that anyway.  Declaring types really pins you down and suddenly, it's really
not Python.

> Not totally off-topic in any case since it does have bindings to the
> Python C API. The demos are written for Python2.2.  I actually got into
> getting them to run on Python2.3 (change LONG_LONG to PY_LONG_LONG
> (thank you Google) - but when it got to trying to fix error messages it
> was generating related to _PyObject_GC_UnTrack I threw in the towel.
> Especially since I am not sure where the Lush/Python link is supposed
> to get you.

As I was explaining to Tim at our most recent official ILP meeting, in
theory, any Turing Complete language can implement any other Turing Complete
language.  You can interpret Python with a C program, but you could also
interpret C with a Python program.  That sounds wrong, only because the chip
itself (the Intel or AMD or Motorola thingy) is also a Turing Machine, and a
really critical one.  Languages that go straight to the chip somehow, have
to be there.  PyPy is an attempt to deepen the Python layer.  You could also
design a chip that executes Python byte codes natively somehow -- that'd be
interesting to think about.

> 
> __
> A little effort at sharing my finds of the day.
> 
> I hope no one objects.
> 
> Art

Not at all, good finds.  Back to napping.

Kirby




More information about the Edu-sig mailing list