1.1.0: reports having data_scale and has Nones as scale/offset about nibabel

[matthew-brett]

Ah - has_data_slope refers to the class. ie nifti1 headers have a data slope. Probably a bit confusing...

[matthew-brett]

@yarikoptic - any comments here? Maybe renaming the class attribute? To something like 'can_use_slope' or something?

[yarikoptic]

Ah -- "has_data_slope refers to the class" -- sorry I have missed it I guess... Now that I know that it refers to description of a format not to a particular instance, meaning became clearer, so if you like feel free to leave it alone. Alternatively indeed it could be renamed -- not sure how many people use it though (and how many have the same incorrect understanding behind as I had) so may be break of API would be actually helpful.

May be it is worth unifying specification of format 'features'? I see that nibabel/imageclasses.py defines a class_map (with e.g. has_affine) and there is also has_data_intercept ... what if classes define something like

.properties = dict(ext=['.mgz', '.mgh'], has_affine=True, has_data_intercept=False, ....)

and then class_map just contains 'class' and the rest get assembled from classes' .properties?

Just an idea...

[matthew-brett]

On Fri, Mar 2, 2012 at 10:11 AM, Yaroslav Halchenko
[email protected]
wrote:

Ah -- "has_data_slope refers to the class" -- sorry I have missed it I guess... Now that I know that it refers to description of a format not to a particular instance, meaning became clearer, so if you like feel free to leave it alone. Alternatively indeed it could be renamed -- not sure how many people use it though (and how many have the same incorrect understanding behind as I had) so may be break of API would be actually helpful.

May be it is worth unifying specification of format 'features'?  I see that nibabel/imageclasses.py  defines a class_map (with e.g. has_affine) and there is also has_data_intercept ... what if classes define something like

.properties = dict(ext=['.mgz', '.mgh'], has_affine=True, has_data_intercept=False, ....)

and then class_map just contains 'class' and the rest get assembled from classes' .properties?

Yes... Could be nice. The disadvantage of that is that you lose the
inheritance. For example. nifti inherits from SPMAnalyze and
therefore gets 'has_data_slope'. But with a dictionary you have to
fill each class fresh.

[yarikoptic]

yes -- inheritance is lost... but could be easily compensated to mimic
inheritance:

class A(object):
    properties = {1: 'a',
                  'persist': 1}

class B(A):
    properties = A.properties.copy()
    properties[1] = 'b'     # changed
    properties[2] = 'new'

print A.properties
print B.properties

results in desired

{1: 'a', 'persist': 1}
{1: 'b', 2: 'new', 'persist': 1}

Yes... Could be nice. The disadvantage of that is that you lose the
inheritance. For example. nifti inherits from SPMAnalyze and
therefore gets 'has_data_slope'. But with a dictionary you have to
fill each class fresh.

=------------------------------------------------------------------=
Keep in touch www.onerussian.com
Yaroslav Halchenko www.ohloh.net/accounts/yarikoptic