source: git/Singular/dyn_modules/python/README @ 4bde6b

For binary distributions remember to set the environment vraiable
PYTHONPATH to the directories containing the Singular python modules
(interpreter.py, utils.py, usually the Singular LIB directory)
and to your python directory (usually .)

Example for bash etc.:
export PYTHONPATH=".:/usr/local/Singular/3-0-2/LIB"

Requirements for building:
--------------------------
install scons (the build tool)
install boost::python
install boost
dynamic modules working
have a working Singular build directory
NOSTREAMIO disabled in factory (rebuild also claptmpl.o in Singular)
install python-numeric (for matrix report)

Suggested additional modules:
probstat: fast (C-Impl.) combinatorics (permutations, cartesian products)
lxml for openmath


supported interpreter values:

INPUT:
int
poly
vector
number
matrix
ideal
list
string
ring
intvec

Output:
int
poly
vector
number
matrix
ideal
list
string
ring
intvec

typemappings:
interpreter, python internal, python official name
int -> int
poly -> Poly ->Polynomial
vector -> Vector -> Vector
number -> Number ->Number
matrix -> Numeric.array
ideal -> Ideal ->Ideal
list -> list
string -> str
ring -> ring(in a Ring, but this does nothing) ->ring
BUILD:
--------
just type:
scons

adjust PYTHONPATH:
-----------------
python must be able to find interpreter.py
so either
export PYTHONPATH="<singroot>/modules/python"
but
for me simply
export PYTHONPATH=
 also works

Test:
-----
../../Singular/Singular <testfile>
where testfile is in
powertester
polytester interpretertester
if an exceptions occurs, review the code it may be wanted for examples the last line in interpretertester

TASKS:
------
- push and pop rings, automatically change rings, if Singular functions are executed
-Map more types
-throw Python exception if something fails
-testing
-write some module presenting the power of python
  compared to the Singular interpreter language
- make modgen work properly
- make the functionality available build in the kernel for the windows platform
- documentation
- provide feedback for missing functionality
-smart pointer for ring in Poly

RULES:
-----

If you want to help, which would make me happy, follow some simple rules:

- divide the work in layers, so what is a problem of a better C++-Interface should be done in the Singular kernel (compare Number.h, Poly.h PowerSeries.h), the Pythonwrapper itself should be ultrathin
-raw operations (+,-,* creation of res...) should map directly to the Singular kernel (prefereable nice C++ wrapper classes in the kernel)
- Wrapper classes in C++ should behave like canonical C++-Classes, no additional res.-management (with the exception of new, delete) should be required
- never use currRing implicitly, always write out the currRing argument, if this isn't done consequently in my code, it would be nice of you to fix it
- regularly check in, check out from CVS


NEWS:
*global_functions has be renamed to singular_globals_proxy
the old name exists for short time for compat.

*it is now possible to use the singular_globals_proxy for
    -reading global variables
    -writing global variables, if the var exists with the correct type in the interpreter
    -calling global interpreter and builtin functions (not builtins, for example groebner but not std

successfully tested systems:
linux: x86, x86_64
Mac OS X:G4,
windows (ask Oliver)