Spurious deprecation warning of vector array interface

[Jan Blechta]

Pull request #403 introduced some deprecation warnings which are spurious. Note also of double implementation of GenericVector.__iter__ which is probably related to this problem.

from dolfin import *
import numpy as np

mesh = UnitSquareMesh(8, 8)
V = FunctionSpace(mesh, 'P', 1)
u = Function(V)

u.vector()[:] *= u.vector()  # Ok
u.vector()[:] *= np.ones((u.vector().size(),))  # Deprecated?
u.vector()[:] * np.ones((u.vector().size(),))  # Deprecated?

Thanks Sigvald Marholm for reporting https://www.allanswered.com/post/rejvg/#vnmbo.

[Tormod Landet]

I am testing the fix now (3c92512). Don't know how I missed that routine, pretty sure I grepped through all the code to find instances of array() ... Good that this was caught before release

[Tormod Landet]

Tested now. With the fix this does not print any warning. Will merge to next and master

[Tormod Landet]

Fix in master

[Jan Blechta]

Usage of deprecated code still in plotting:

dolfin$ grep -r array\(\) site-packages/dolfin/ python/dolfin
site-packages/dolfin/compilemodules/compilemodule.py:import_array();
site-packages/dolfin/mesh/boundarysubdomainfinder.py:            subdomains[boundary_cell_map.array()[key]] =  counter
site-packages/dolfin/common/plotting.py:        C = fvec.array()  # NB! Assuming here dof ordering matching cell numbering
site-packages/dolfin/common/plotting.py:        C = obj.array()
python/dolfin/common/plotting.py:        C = obj.array()
dolfin$ python3 demo/documented/mixed-poisson/python/demo_mixed-poisson.py 
Solving linear variational problem.
/home/jan/dev/fenics-master/src/dolfin/local.master/lib/python3.6/site-packages/dolfin/cpp/la.py:1331: DeprecationWarning: GenericVector.array() is being deprecated, use GenericVector.get_local()
  "use GenericVector.get_local()", DeprecationWarning)

[Tormod Landet]

C = obj.array() should be kept since obj is a meshfunction

I will fix C = fvec.array()

[Tormod Landet]

Fix issue #961

Hopefully last time, grep shows nothing suspicious in ./dolfin/swig, ./python or ./site-packages after this

→ <<cset bbfa212d7247>>

[Jan Blechta]

grep -r array\(\) test/ shows few suspicious items and CI shows warnings:

python/jit/test_jit.py::test_compile_extension_module
  /usr/lib/python3/dist-packages/dolfin/cpp/la.py:1331: DeprecationWarning: GenericVector.array() is being deprecated, use GenericVector.get_local()
    "use GenericVector.get_local()", DeprecationWarning)
  /usr/lib/python3/dist-packages/dolfin/cpp/la.py:1331: DeprecationWarning: GenericVector.array() is being deprecated, use GenericVector.get_local()
    "use GenericVector.get_local()", DeprecationWarning)
  /usr/lib/python3/dist-packages/dolfin/cpp/la.py:1331: DeprecationWarning: GenericVector.array() is being deprecated, use GenericVector.get_local()
    "use GenericVector.get_local()", DeprecationWarning)
  /usr/lib/python3/dist-packages/dolfin/cpp/la.py:1331: DeprecationWarning: GenericVector.array() is being deprecated, use GenericVector.get_local()
    "use GenericVector.get_local()", DeprecationWarning)

python/multimesh/test_volume.py::test_volume_2d
  /usr/lib/python3/dist-packages/dolfin/cpp/la.py:1331: DeprecationWarning: GenericVector.array() is being deprecated, use GenericVector.get_local()
    "use GenericVector.get_local()", DeprecationWarning)

python/multimesh/test_volume.py::test_volume_2d_4_meshes
  /usr/lib/python3/dist-packages/dolfin/cpp/la.py:1331: DeprecationWarning: GenericVector.array() is being deprecated, use GenericVector.get_local()
    "use GenericVector.get_local()", DeprecationWarning)

python/multimesh/test_volume.py::test_volume_2d_six_meshes
  /usr/lib/python3/dist-packages/dolfin/cpp/la.py:1331: DeprecationWarning: GenericVector.array() is being deprecated, use GenericVector.get_local()
    "use GenericVector.get_local()", DeprecationWarning)

[Tormod Landet]

I only updated the tests that are relevant on pybind11. Are anyone going to run the unit tests with SWIG after the 2017.2 release? I thought all that code would be deleted ASAP?

[Jan Blechta]

Sure, if all these cases are to be ditched with SWIG deletion, then it is fine.