Am 01.10.2013 19:52, schrieb F. B.:
Unlike other things like that, such as PyPy, Cython, ShedSkin, they aim to
be 100% compatible with CPython syntax. Nuitka does not create its own
implementation of the Python language, rather it uses the ast package. of
CPython to parse code, which could be a strong point to keep compatibility.
That's a good thing, though they'd still have to upgrade the system with
each new Python version, so it's not buying the project THAT much.
But yes it makes sure that no deviations creep into the parsing, which
is good.
In Nuitka's TODO list there is an interesting feature: static typing
through decorators, which unlike Cython's *cdef*, do not break CPython
compatibility. Unfortunately it's not yet implemented, but once it will be,
I think the best solution to speed up SymPy would be to clean up Python's
code and put static-variable-decorators. As of now, Python variables are
still translated as PyObject pointers, which shouldn't have great benefits
in term of speed.
The main question, then, is: Is Nuitka viable?
I.e. is it going to stay around as long as Sympy will? How fast do they
fix showstopper bugs? Are type decorators and (hopefully) something
smart for the data representation going to be implemented within our
lifetimes?
Alternatively, we could do our own Python-to-C++ translator. This would
be easier than a full-fledged translator because we'd have to cover only
those constructs that are actually used in Sympy, so we could take
considerable shortcuts (at the cost of needing a list of coding
restrictions, so this option would still come with attached strings).
--
You received this message because you are subscribed to the Google Groups
"sympy" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/sympy.
For more options, visit https://groups.google.com/groups/opt_out.
