Though I understand the implication, but as there are currently no readers that can deal with larger than RAM npy images, I suggest keeping the whole buffer approach for the first commit.

Thanks! Dov

Øyvind Kolås wrote:

On Tue, May 28, 2013 at 5:27 PM, Dov Grobgeld wrote:

I finished a saver operation for floating point npy images. It may e.g. be used to debug floating point operations (e.g. gaussian-blur) with numerical python. The resulting image may be read into numpy as follows:

   import numpy    img = numpy.load('image.npy')

It may also be viewed in my image viewer giv (though it currently does not support color npy images).

Is it ok to commit it?

It would be good to avoid fetching the whole buffer with gegl_buffer_get; GeglBuffer can store larger than RAM images and a malloc this large might fail. It would be better to use a loop either over scanlines of the image or groups of scanlines (32 or 64 at a time).

/Øyvind K.

-- «The future is already here. It's just not very evenly distributed»                                                  -- William Gibson http://pippin.gimp.org/                            http://ffii.org/

Ok. I commit it after slicing up the processing. Figured it might be a good exercise. :-)

Regards,
Dov

On Wed, May 29, 2013 at 9:51 PM, Øyvind Kolås wrote:

On Wed, May 29, 2013 at 6:28 PM, Dov Grobgeld wrote:

Though I understand the implication, but as there are currently no readers that can deal with larger than RAM npy images, I suggest keeping the whole buffer approach for the first commit.

It is fine as a first commit, but the concern of not needing buffer-sized allocations is a general concern in GEGL, this isn't really a critical place; but dealing with it would be nice. :).

/Ø