ResetImagePixels: Assertion `image != (Image *) NULL' failed.

[DanKrt82]

I all

I have a problem using wandImage.from_array using a big nparray.

That's my code (typeToUse is np.uint8, h_swath is 1024 and x is 31496):

np_array = np.ones(shape=(h_swath, x), dtype=typeToUse) * 255
with wandImage(wandImage.from_array(np_array.astype(np.uint8), b'I')) as img:
print(img.size)
print('width =', img.width)
print('height =', img.height)
print(img.format)
print(img.depth)
print(img.page)
print(img.strip)
print(img.strip)
img.type = 'grayscale'
img.depth = 2
img.resolution = (int(dpiX), int(dpiY))
img.compression = 'undefined'
img.save(filename=out_path + '/Slice_Pass_' + str(swath) + '.tif')

On wandImage.from_array call the script dump and the error is the seguent:

python3: ../../magick/image.c:2300: ResetImagePixels: Assertion `image != (Image *) NULL' failed.
Aborted (core dumped)

I use that code on a windows installation and I don't have that problem, now I install all the environment on a Ubuntu machine (5.8.0-50-generic #56~20.04.1-Ubuntu SMP) using that package:

$ dpkg -l | grep wand
ii libmagickwand-6-headers 8:6.9.10.23+dfsg-2.1ubuntu11.2 all image manipulation library - headers files
ii libmagickwand-6.q16-6:amd64 8:6.9.10.23+dfsg-2.1ubuntu11.2 amd64 image manipulation library -- quantum depth Q16
ii libmagickwand-6.q16-dev:amd64 8:6.9.10.23+dfsg-2.1ubuntu11.2 amd64 image manipulation library - development files (Q16)
ii libmagickwand-dev 8:6.9.10.23+dfsg-2.1ubuntu11.2 all image manipulation library -- dummy package

Any suggest to avoid that problem? I also use gdb, but I don't found any helpfull indication.

Thank you

[emcconville]

The error means that the pixel data was removed from memory before the rest cycle completed. Possibly due to limited resources, but I would suspect that Python's fastbin saw wandImage(wandImage.from_array(...)) and optimized it away. Try this...

with wandImage.from_array(np_array, b'R') as img:
    ....

Note: I'm assuming you want b'R' as this is a grayscale image, not b'I' which would be index.

[DanKrt82]

Hi @emcconville,

thank you for your reply. I forgot to write that if I use small images all it's working fine, do you know if there is some configuration to do, for example ulimit and some other setting on the linux systems?

[emcconville]

I can suggest looking into configuring resources with policy.xml, but a better solution might be rewriting the code to use temp files. Something as simple as...

from tempfile import TemporaryFile

with TemporaryFile() as tfd:
    np_array.tofile(tfd)
    tfd.seek(0)
    with Image(file=tfd, depth=8, format='R', width=x, height=h_swath) as img:
        img.save(filename=out_path + '/Slice_Pass_' + str(swath) + '.tif')

[DanKrt82]

I already change the policy configuration, but nothing change. Tomorrow I'll test the temporary file like you suggested and I'll give you a feedback.

Thank you very much

[DanKrt82]

Using temporary files and changing the configuration using policy.xml my code is working also with big image (~10Gb).

Thank you