pyarpack: python binding based on Boost.Python.Numpy used to expose arpack C++ API
Installation:
-------------
Python3: ~/arpack-ng/build> cmake -DCMAKE_INSTALL_PREFIX=/tmp/local -DPYTHON3=ON -DBOOST_PYTHON_LIBSUFFIX="3" ..
~/arpack-ng/build> make all test
~/arpack-ng/build> make install
Note: Boost must have been compiled for Python3.
Note: installation will install libpyarpack.so in the install directory.
Note: python3 minimal packages must be installed (apt-get install python3-minimal python3-pip).
numpy must be installed (apt-get install python3-numpy) as the A/B matrices used by arpack
solver must be numpy arrays.
Usage:
------
PYTHONPATH must include the directory where libpyarpack.so has been installed.
>> export PYTHONPATH="/tmp/local/lib/pyarpack:${PYTHONPATH}"
>> python
>> import pyarpack
>> help(pyarpack)
You can use sparse or dense matrices, and, play with iterative or direct mode solvers (CG, LU, ...):
1. choose arpack solver with a given mode solver
1.1. if you need to handle sparse matrices
>> from pyarpack import sparseBiCG as pyarpackSlv
1.2. if you need to handle dense matrices
>> from pyarpack import denseBiCG as pyarpackSlv
2. choose arpack data type (float, double, ...)
>> arpackSlv = pyarpackSlv.double()
3. solve the eigen problem
>> arpackSlv.solve(A, B)
4. get eigen values and vectors
>> print(arpackSlv.vec)
>> print(arpackSlv.val)
You can also:
1. restart a solve from the workspace of a previous solve: check out pyarpackRestart.py.in.
2. compute eigen and / or schur vectors.
Note:
1. arpack data type (float, double, ...) must be consistent with A/B numpy dtypes (float32, float64, ...).
at python side, the data MUST be casted in the EXACT expected type (int32, int64, float, double, ...).
otherwise, C++ may not get the data the way it expects them: C++ will not know how to read python data.
if you are not sure how data have been passed from python to C++, set arpackSlv.debug = 1 and check out debug traces.
in other words, pyarpack users MUST :
1.1. create numpy arrays specifying explicitly the type:
>> Aij = np.array([], dtype='complex128')
1.2. filling numpy arrays casting value on append:
>> Aij = np.append(Aij, np.complex128(complex( 200., 200.))) # Casting value on append is MANDATORY or C++ won't get the expected type.
1.3. calling the solver flavor which is consistent with the numpy array data type:
>> arpackSlv = pyarpackSlv.complexDouble() # Caution: complexDouble <=> np.array(..., dtype='complex128')
note: NO data type check can be done at C++ side, the pyarpack user MUST insure data consistency.
2. sparse matrices must be provided in coo format (n, i, j, Mij), that is, as a tuple where:
2.1. n is an integer.
2.2. i, j, Mij are 1 x nnz numpy arrays.
3. dense matrices must be provided in raw format (Mij, rowOrdered), that is, as a tuple where:
3.1. Mij is an n x n numpy array.
3.2. rowOrdered is a boolean (column ordered if False).
4. arpack mode solver are provided by eigen:
4.1. when solver is iterative, A and B can be sparse only.
4.2. when solver is direct, A and B can be sparse or dense.
Examples:
---------
~/arpack-ng> find . -name *.py.in (template files from which python scripts will result)
