Why not implementing __repr__ and __str__ for Nifti1Image?

[dangom]

Hi guys,

I was wondering whether there is a reason why the Nifti1Image class doesn't implement repr nor str. I think it'd be a great addition, and very informative for end users to be able to get simple info on the REPL instead of the default <nibabel.nifti1.Nifti1Image at 0x17d895ab0>. But maybe there is a motivation behind that?

Perhaps topic for another issue, but it'd be lovely if that simple info printed on the str would be implemented as properties. So say we print dtype, voxel size and TR, then calling nifti_image.TR would also print the TR (and similarly nifti_image.TR = 3. would change it by adapting the header).

what do you think?!

Example:

import nibabel

def example_str(self):
    dimension = "4D" if len(self.dataobj.shape) == 4 else "3D"

    info = f"""
    Nifti1Image (type: {self.dataobj.dtype})

    Dimension: {dimension}
    Shape: {self.dataobj.shape}
    Voxel Dimensions: {self.header["pixdim"][1:4]}
    Repetition Time: {self.header["pixdim"][4]:.3f}
    """
    return info


nibabel.nifti1.Nifti1Image.__str__ = example_str

Already improves things a lot:

[effigies]

__str__ is implemented in SpatialImage:

>>> print(nb.Nifti1Image(np.zeros((5,5,5)), np.eye(4)))
<class 'nibabel.nifti1.Nifti1Image'>
data shape (5, 5, 5)
affine: 
[[1. 0. 0. 0.]
 [0. 1. 0. 0.]
 [0. 0. 1. 0.]
 [0. 0. 0. 1.]]
metadata:
<class 'nibabel.nifti1.Nifti1Header'> object, endian='<'
sizeof_hdr      : 348
data_type       : b''
db_name         : b''
extents         : 0
session_error   : 0
regular         : b''
dim_info        : 0
dim             : [3 5 5 5 1 1 1 1]
intent_p1       : 0.0
intent_p2       : 0.0
intent_p3       : 0.0
intent_code     : none
datatype        : float64
bitpix          : 64
slice_start     : 0
pixdim          : [1. 1. 1. 1. 1. 1. 1. 1.]
vox_offset      : 0.0
scl_slope       : nan
scl_inter       : nan
slice_end       : 0
slice_code      : unknown
xyzt_units      : 0
cal_max         : 0.0
cal_min         : 0.0
slice_duration  : 0.0
toffset         : 0.0
glmax           : 0
glmin           : 0
descrip         : b''
aux_file        : b''
qform_code      : unknown
sform_code      : aligned
quatern_b       : 0.0
quatern_c       : 0.0
quatern_d       : 0.0
qoffset_x       : 0.0
qoffset_y       : 0.0
qoffset_z       : 0.0
srow_x          : [1. 0. 0. 0.]
srow_y          : [0. 1. 0. 0.]
srow_z          : [0. 0. 1. 0.]
intent_name     : b''
magic           : b'n+1'

I would be open to improving the __repr__, though I would expect it to (usually) be <80 characters to keep the REPL usable.

[dangom]

amazing!! I must've overlooked the str implementation. That's pretty good, thank you. Do you think it'd be possible to add some provenance info to it, such as whether the nifti was loaded directly from a file, and if so print its filename?

If <80 characters then it'd be hard to have it output something that could be used to instantiate the object upon eval. But we could take inspiration from ANTsPy:

[effigies]

Yeah, I wouldn't expect a repr to be eval-able. But we could report filenames. nibabel.filebasedimages.FileBasedImage has a get_filename() method.