add_cutdata seems to overwrite data for left hemisphere with data from right hemisphere.

[N-HEDGER]

Hi pycortex. I seem to have uncovered an issue that doesn't seem to have been raised so far. It overlaps somewhat with #398

Problem

To illustrate this problem clearly, here I add some data that should show blue for the left hemisphere and red for the right hemisphere.

hemdat=np.repeat([0,1], 10242) # 10242 vertices in fsaverage5

hemdatv=cortex.Vertex(hemdat,subject='fsaverage5',vmin=0,vmax=1,cmap='BuBkRd')
cortex.webshow(hemdatv)

Here is the proof, in cortex.webshow()

However. When I call the cut_surface command, specifying the left hemisphere:

my_retinotopy=cortex.Dataset()
my_retinotopy.views['hem']=hemdatv
my_retinotopy.description='my_retinotopy'  
cortex.segment.cut_surface(cx_subject='fsaverage5',hemi='lh',name='retinotopy_flat',fs_subject='fsaverage5',data=my_retinotopy)

we see that the left hemisphere is coloured red, not blue as in the webviewer.

Culprit lines

My guess is that the problem lies here:

pycortex/cortex/blender/__init__.py

Lines 97 to 100 in 499965a

	lcolor = blendlib._repack(left)
	rcolor = blendlib._repack(right)
	print(len(lcolor), len(rcolor))
	blendlib.add_vcolor((lcolor, rcolor), mesh, name)

Here, despite the fact that we are only cutting the left hemisphere, it seems as though data for both hemispheres are added to blender. Presumably this is what causes the right hemispheres data to appear on the left (perhaps the left data is over-written by the right data).

Proposed solution.

1. in segment.py, hemi could be added as an argument to add_cutdata, as follows

pycortex/cortex/segment.py

Line 185 in 499965a

blender.add_cutdata(fname, data, name=data.description)

blender.add_cutdata(fname, data,hemi, name=data.description)

2. The hem input can then be added accordingly in add_cutdata itself.

a.

pycortex/cortex/blender/__init__.py

Line 43 in 499965a

def add_cutdata(fname, braindata, name="retinotopy", projection="nearest", mesh="hemi"):

def add_cutdata(fname, braindata,hemi, name="retinotopy", projection="nearest", mesh="hemi"):

b.

pycortex/cortex/blender/__init__.py

Line 65 in 499965a

add_cutdata(fname, data, name=view_name, projection=projection, mesh=mesh)

add_cutdata(fname, data,hemi, name=view_name, projection=projection, mesh=mesh)

3. Then, we replace the end of the add_cutdata function

pycortex/cortex/blender/__init__.py

Lines 86 to 104 in 499965a

	p.pack_array(lcolor.ravel(), p.pack_double)
	p.pack_array(rcolor.ravel(), p.pack_double)
	with tempfile.NamedTemporaryFile() as tf:
	tf.write(p.get_buffer())
	tf.flush()
	code = """with open('{tfname}', 'rb') as fp:
	u = xdrlib.Unpacker(fp.read())
	mesh = u.unpack_string().decode('utf-8')
	name = u.unpack_string().decode('utf-8')
	left = u.unpack_array(u.unpack_double)
	right = u.unpack_array(u.unpack_double)
	lcolor = blendlib._repack(left)
	rcolor = blendlib._repack(right)
	print(len(lcolor), len(rcolor))
	blendlib.add_vcolor((lcolor, rcolor), mesh, name)
	""".format(tfname=tf.name)
	_call_blender(fname, code)
	return

with something like this, to only add one set of data depending on the hemi argument.

    if hemi=='lh':
        dat2add=lcolor
    else:
        dat2add=rcolor

    p.pack_array(dat2add.ravel(), p.pack_double)


    with tempfile.NamedTemporaryFile() as tf:
        tf.write(p.get_buffer())
        tf.flush()
        code = """with open('{tfname}', 'rb') as fp:
            u = xdrlib.Unpacker(fp.read())
            mesh = u.unpack_string().decode('utf-8')
            name = u.unpack_string().decode('utf-8')
            hemdat= u.unpack_array(u.unpack_double)
            ccolor = blendlib._repack(hemdat)
            blendlib.add_vcolor(ccolor, mesh, name)
        """.format(tfname=tf.name)

        print(code)
        _call_blender(fname, code)

    return 

This seems to produce the desired result.

Since the code that makes python talk to blender is a little confusing to me - these are tentative suggestions, but they seem to do the job.

Happy to address this myself - but any comments welcome.

[marklescroart]

Pardon the late reply here - the root issue is that no one in the original pycortex development team ever tried to make cuts based on anything besides volumetric functional localizer (mostly retinotopy) data, so this is all good-enough code. These fixes seem reasonable; I'd suggest you submit a pull request and we can test them out. Thanks!