Moved weave

1 files changed, 116 insertions, 0 deletions

diff --git a/scipy/weave/examples/vtk_example.py b/scipy/weave/examples/vtk_example.py
new file mode 100644
index 000000000..b9daf631c
--- /dev/null
+++ b/scipy/weave/examples/vtk_example.py
@@ -0,0 +1,116 @@
+""" A simple example to show how to use weave with VTK.  This lets one
+create VTK objects using the standard VTK-Python API (via 'import
+vtk') and then accelerate any of the computations by inlining C++ code
+inside Python.
+
+Please note the use of the `inc_dirs` and the `lib_dirs` variables in
+the call to weave.inline.  Point these to where your VTK headers are
+and where the shared libraries are.
+
+For every VTK object encountered the corresponding VTK header is
+automatically added to the C++ code generated.  If you need to add
+other headers specified like so::
+
+ headers=['"vtkHeader1.h"', '"vtkHeader2.h"']
+
+in the keyword arguments to weave.inline.  Similarly, by default,
+vtkCommon is linked into the generated module.  If you need to link to
+any of the other vtk libraries add something like so::
+
+ libraries=['vtkHybrid', 'vtkFiltering']
+
+in the keyword arguments to weave.inline.  For example::
+
+ weave.inline(code, ['arr', 'v_arr'],
+              include_dirs = ['/usr/local/include/vtk'],
+              library_dirs = ['/usr/local/lib/vtk'],
+              headers=['"vtkHeader1.h"', '"vtkHeader2.h"'],
+              libraries=['vtkHybrid', 'vtkFiltering'])
+
+
+This module has been tested to work with VTK-4.2 and VTK-4.4 under
+Linux.  YMMV on other platforms.
+
+
+Author: Prabhu Ramachandran
+Copyright (c) 2004, Prabhu Ramachandran
+License: BSD Style.
+
+"""
+
+import weave
+import vtk
+import scipy_base.numerix
+
+import sys
+import time
+
+
+# Please change these to suit your needs.  If not, this example will
+# not compile.
+inc_dirs = ['/usr/local/include/vtk', '/usr/include/vtk']
+lib_dirs = ['/usr/local/lib/vtk', '/usr/lib/vtk']
+
+
+def simple_test():
+    """A simple example of how you can access the methods of a VTK
+    object created from Python in C++ using weave.inline.
+
+    """
+    
+    a = vtk.vtkStructuredPoints()
+    a.SetOrigin(1.0, 1.0, 1.0)
+    print "sys.getrefcount(a) = ", sys.getrefcount(a)
+
+    code=r"""
+    printf("a->ClassName() == %s\n", a->GetClassName());
+    printf("a->GetReferenceCount() == %d\n", a->GetReferenceCount());
+    double *origin = a->GetOrigin();
+    printf("Origin = %f, %f, %f\n", origin[0], origin[1], origin[2]);
+    """
+    weave.inline(code, ['a'], include_dirs=inc_dirs, library_dirs=lib_dirs)
+
+    print "sys.getrefcount(a) = ", sys.getrefcount(a)
+    
+
+def array_test():
+    """Tests if a large scipy_base.numerix array can be copied into a
+    vtkFloatArray rapidly by using weave.inline.
+
+    """
+
+    # Create a large scipy_base.numerix array.
+    arr = scipy_base.numerix.arange(0, 10, 0.0001, 'f')
+    print "Number of elements in array = ", arr.shape[0]
+
+    # Copy it into a vtkFloatArray and time the process.
+    v_arr = vtk.vtkFloatArray()
+    ts = time.clock()
+    for i in range(arr.shape[0]):
+        v_arr.InsertNextValue(arr[i])
+    print "Time taken to do it in pure Python =", time.clock() - ts    
+
+    # Now do the same thing using weave.inline
+    v_arr = vtk.vtkFloatArray()
+    code = """
+    int size = Narr[0];
+    for (int i=0; i<size; ++i)
+        v_arr->InsertNextValue(arr[i]);
+    """
+    ts = time.clock()
+    # Note the use of the include_dirs and library_dirs.
+    weave.inline(code, ['arr', 'v_arr'], include_dirs=inc_dirs,
+                 library_dirs=lib_dirs)    
+    print "Time taken to do it using Weave =", time.clock() - ts
+
+    # Test the data to make certain that we have done it right.
+    print "Checking data."
+    for i in range(v_arr.GetNumberOfTuples()):
+        val = (v_arr.GetValue(i) -arr[i] )
+        assert (val < 1e-6), "i = %d, val= %f"%(i, val)
+    print "OK."
+
+
+if __name__ == "__main__":    
+    simple_test()
+    array_test()