Fix for the definition of miget_dimension_sizes.

[carlohamalainen]

The full definition from MINC-2.2.00 is:

extern int miget_dimension_sizes(const midimhandle_t dimensions[], int array_length,
             unsigned int sizes[]);

The definition in libpyminc2.py has int_sizes (C's unsigned long long) for the last parameter, but it should be uint_sizes.

Test script:

from pyminc.volumes.factory import *
import numpy as np
import os

tmp = '/tmp/test.mnc'

try:
    os.remove(tmp)
except OSError:
    pass

shape = (10, 20)

vol = volumeFromDescription(tmp, ['yspace', 'xspace'], shape, [-5, -5], [0.2, 0.2])
vol.data = np.random.random(shape)
print vol.data.shape # should say (10, 20)
vol.writeFile()
vol.closeVolume()

Result:

$ python test_creation.py
Traceback (most recent call last):
  File "test_creation.py", line 15, in <module>
    vol.data = np.random.random(shape)
  File "/usr/local/lib/python2.7/dist-packages/pyminc/volumes/volumes.py", line 63, in setdata
    raise IncorrectDimsException
pyminc.volumes.volumes.IncorrectDimsException

After the fix:

$ python test_creation.py
(10, 20)

I tested scripts/pyminc_test2.py and it seemed to run successfully:

$ ./pyminc_test2.py /tmp/test.mnc numpy /tmp/test_out.mnc
4
Opening volume took 0.01000 seconds
Voxel: 0.000000
sizes: 110 217 181
0
Getting hyperslab took 0.15000 seconds
55
0
(4320470,)
(110, 217, 181)
0
computing ten neighbourhood averages using numpy method
Iterations 0 took 0.260 seconds
Iterations 1 took 0.230 seconds
Iterations 2 took 0.220 seconds
Iterations 3 took 0.230 seconds
Iterations 4 took 0.230 seconds
Iterations 5 took 0.220 seconds
Iterations 6 took 0.230 seconds
Iterations 7 took 0.230 seconds
Iterations 8 took 0.220 seconds
Iterations 9 took 0.230 seconds
done averaging. Took a total of 2.300 seconds, averaging 0.230 seconds
before volume
after volume
max and min 87 0
(110, 217, 181)
Creating new volume took 0.36000 seconds
before setting hyperslab
0
0.0
0 255 0 63.2519589304
int32 uint8
[[[0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]
  ...,
  [0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]]

 [[0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]
  ...,
  [0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]]

 [[0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]
  ...,
  [0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]]

 ...,
 [[0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]
  ...,
  [0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]]

 [[0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]
  ...,
  [0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]]

 [[0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]
  ...,
  [0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]
  [0 0 0 ..., 0 0 0]]]
Setting hyperslab took 0.09000 seconds

$ mincstats /tmp/test.mnc
File:              /tmp/test.mnc
Mask file:         (null)
Total voxels:      4320470
# voxels:          4320470
% of total:        100
Volume (mm3):      4320470
Min:               0
Max:               102.5024482
Sum:               102940629.4
Sum^2:             5570867979
Mean:              23.82625718
Variance:          721.7221255
Stddev:            26.86488648
CoM_voxel(z,y,x):  42.81716948 105.8562036 90.67176645
CoM_real(x,y,z):   0.6717664491 -20.14379639 30.81716948

Histogram:         (null)
Total voxels:      4320470
# voxels:          4320470
% of total:        100
Median:            0.03855289613
Majority:          0.02562561259
BiModalT:          28.26505089
PctT [  0%]:       0.02562561259
Entropy :          5.90236189

$ mincstats /tmp/test_out.mnc
File:              /tmp/test_out.mnc
Mask file:         (null)
Total voxels:      4320470
# voxels:          4320470
% of total:        100
Volume (mm3):      4320470
Min:               0
Max:               87
Sum:               93236088.69
Sum^2:             4408692362
Mean:              21.58008011
Variance:          554.7198926
Stddev:            23.55249228
CoM_voxel(z,y,x):  42.56045625 105.8436785 90.68415457
CoM_real(x,y,z):   0.6841545723 -20.15632149 30.56045625

Histogram:         (null)
Total voxels:      4320470
# voxels:          4320470
% of total:        100
Median:            8.808750357
Majority:          0.02174999937
BiModalT:          24.90374947
PctT [  0%]:       0.02174999937
Entropy :          3.946002744

Note: all testing was done on Debian Wheezy 64bit; I got segfaults on Debian Wheezy 32bit with the latest commit and also my patches. Is pyminc compatible with 32bit systems?

[mcvaneede]

Hi Carlo,

Currently pyminc is geared towards Vladimir's development branch from libminc, because that is the one that is installed when you use the minc-tookit repository (https://github.com/BIC-MNI/minc-toolkit). There the function definition for miget_dimension_sizes has changed:

https://github.com/BIC-MNI/libminc/blob/fa3ea0a6c2758db6b0fbc2ed0737e92b9382ebfb/libsrc2/minc2_api.h#L552-553

and uses unsigned long long. There are a number of other functions that will have the same discrepancy with the master branch. Potentially this repository could have a master and develop branch that mimic the changes in limbic.

Cheers, Matthijs

[carlohamalainen]

Hi Matthijs,

Thanks for clarifying. For the moment I'll switch to the BIC-MNI/minc-toolkit repository so you can close this pull request.

Cheers,

-- Carlo

[mcvaneede]

I've updated the README of the repository to indicate that pyminc is currently tailored to libminc's development branch. Closing this issue.