MantisBT - VTK |
| View Issue Details |
|
| ID | Project | Category | View Status | Date Submitted | Last Update |
| 0012730 | VTK | (No Category) | public | 2011-11-16 06:18 | 2011-11-17 08:21 |
|
| Reporter | Bryn Lloyd | |
| Assigned To | David Gobbi | |
| Priority | normal | Severity | minor | Reproducibility | have not tried |
| Status | closed | Resolution | fixed | |
| Platform | | OS | | OS Version | |
| Product Version | 5.8.0 | |
| Target Version | | Fixed in Version | | |
| Project | TBD |
| Type | incorrect functionality |
|
| Summary | 0012730: Importing large arrays from numpy inyo VTK does not work. |
| Description | Importing large arrays from numpy into VTK does not work.
This was already mentioned on the users list, and reproduced by David Gobbi.
Example 1:
==========
from numpy import zeros
import vtk
a = zeros(2**32 - 1).astype(uint8) #to create a massive
array im = vtk.vtkImageImport() #believe this
is the standard method to convert to vtk
im.CopyImportVoidPointer(a.tostring(),a.size) #do conversion
throws the following error:
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
SystemError: error return without exception set
Example 2:
==========
from vtk.util import numpy_support as vn
import vtk
import numpy
a = numpy.zeros(2**32-1, numpy.uint8)
aV = vn.numpy_to_vtk(pp, deep=False)
The last line throws the error:
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "P:\vtk-5.6.1\x64\Wrapping\Python\vtk\util\numpy_support.py", line 165, in numpy_to_vtk
result_array.SetVoidArray(z_flat, len(z_flat), 1)
OverflowError: size does not fit in an int
|
| Steps To Reproduce | |
| Additional Information | |
| Tags | No tags attached. |
| Relationships | |
| Attached Files | |
|
| Issue History |
| Date Modified | Username | Field | Change |
| 2011-11-16 06:18 | Bryn Lloyd | New Issue | |
| 2011-11-16 06:31 | Bryn Lloyd | Note Added: 0027697 | |
| 2011-11-16 10:27 | David Gobbi | Note Added: 0027698 | |
| 2011-11-16 10:27 | David Gobbi | Note Edited: 0027698 | bug_revision_view_page.php?bugnote_id=27698#r396 |
| 2011-11-16 10:35 | David Gobbi | Note Added: 0027699 | |
| 2011-11-16 11:43 | David Gobbi | Assigned To | => David Gobbi |
| 2011-11-16 11:43 | David Gobbi | Status | backlog => tabled |
| 2011-11-16 11:47 | David Gobbi | Note Added: 0027704 | |
| 2011-11-17 08:10 | Bryn Lloyd | Note Added: 0027705 | |
| 2011-11-17 08:21 | David Gobbi | Note Added: 0027706 | |
| 2011-11-17 08:21 | David Gobbi | Status | tabled => closed |
| 2011-11-17 08:21 | David Gobbi | Resolution | open => fixed |
|
Notes |
|
|
(0027697)
|
|
Bryn Lloyd
|
|
2011-11-16 06:31
|
|
I just noticed that the second example was run using VTK 5.6.1. I checked in the docs and see that now vtkIdType is used, instead of int in the function SetVoidArray.
Sorry for the confusion. Using VTK 5.8 I get the same error also for the second example:
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "P:\vtk-5.8.0-epd-712\x64\Wrapping\Python\vtk\util\numpy_support.py", lin
e 160, in numpy_to_vtk
result_array.SetVoidArray(z_flat, len(z_flat), 1)
SystemError: error return without exception set |
|
|
|
(0027698)
|
|
David Gobbi
|
|
2011-11-16 10:27
|
|
The problem might not be due to large arrays, it might have something to do specifically with the value "2**32-1". For example, the following code with "2**32+1" or "2**32" works fine on my mac. Bryn, can you confirm?
====
import numpy
import vtk
a = numpy.zeros(2**32 + 1, numpy.uint8)
im = vtk.vtkImageImport()
im.CopyImportVoidPointer(a, len(a))
====
|
|
|
|
(0027699)
|
|
David Gobbi
|
|
2011-11-16 10:35
|
|
|
On further inspection, it seems to be a sign issue. On my machine it fails for any array size between 2**31 and 2**32-1, which correspond to values that can be represented by unsigned int but not by signed int. |
|
|
|
(0027704)
|
|
David Gobbi
|
|
2011-11-16 11:47
|
|
I have merged commit dee7431302a85b94bc112bde7855e06e26f35b54 to fix this bug.
http://vtk.org/gitweb?p=VTK.git;a=commitdiff;h=dee74313 [^]
diff --git a/Wrapping/Python/vtkPythonArgs.cxx b/Wrapping/Python/vtkPythonArgs.cxx
index 8302337..29f59d9 100644
--- a/Wrapping/Python/vtkPythonArgs.cxx
+++ b/Wrapping/Python/vtkPythonArgs.cxx
@@ -228,9 +228,11 @@ bool vtkPythonGetValue(PyObject *o, const void *&a)
if (b->bf_getsegcount(o, NULL) == 1)
{
void *p;
- int s = (int)b->bf_getreadbuffer(o, 0, &p);
- if (s >= 0)
+ Py_ssize_t sz = b->bf_getreadbuffer(o, 0, &p);
+ if (sz >= 0 && sz <= VTK_INT_MAX)
{
+ // check for pointer mangled as string
+ int s = (int)sz;
a = vtkPythonUtil::UnmanglePointer((char *)p, &s, "void_p");
if (s >= 0)
{
@@ -247,6 +249,12 @@ bool vtkPythonGetValue(PyObject *o, const void *&a)
PyErr_SetString(PyExc_TypeError, "cannot get a void pointer");
}
}
+ else if (sz >= 0)
+ {
+ // directly use the pointer to the buffer contents
+ a = p;
+ return true;
+ }
return false;
}
PyErr_SetString(PyExc_TypeError, "buffer must be single-segment"); |
|
|
|
(0027705)
|
|
Bryn Lloyd
|
|
2011-11-17 08:10
|
|
|
I tested the fix on a Linux machine and it seems to work. Thanks! |
|
|
|
(0027706)
|
|
David Gobbi
|
|
2011-11-17 08:21
|
|
|
I'll recommend this patch for VTK 5.8.1. |
|