[FIX] Bugs found after #1572 and #1591 #1623
Conversation
Improved the way interfaces were initialized to prevent that lists of lists of int are not set. Fixes nipy#1625
oesteban
changed the title
|
|
||
| """ | ||
|
|
||
| reg = Registration(from_file='./smri_ants_registration_settings.json') | ||
| reg = Registration(from_file=example_data('smri_ants_registration_settings.json')) |
There was a problem hiding this comment.
moved this settings file from examples to the example data, since now I use it in tests
Current coverage is 70.88% (diff: 94.66%)@@ master #1623 diff @@
==========================================
Files 1020 1020
Lines 51240 51277 +37
Methods 0 0
Messages 0 0
Branches 7259 7265 +6
==========================================
+ Hits 36314 36347 +33
- Misses 13840 13849 +9
+ Partials 1086 1081 -5
|
|
@satra please have a look, this is ready to merge |
| dict_withhash.append((name, | ||
| self._get_sorteddict(val, True, hash_method=hash_method, | ||
| hash_files=hash_files))) | ||
| for name, val in sorted(self.get_traitsfree().items()): |
There was a problem hiding this comment.
the reason i used get in the previous version was to speed up the hashval computation. since get_traitsfree requires a dictionary cleanup. if using get doesn't change assumptions here, it may be cleaner.
There was a problem hiding this comment.
Ok, back to the original 👍
|
@satra, @chrisfilo this is ready, and fixes a couple of well-identified problems with the python 2/3 compatibility. I have added 3 tests based on ants.Registration to make sure we cover the (possibly) most complex case of nipype interfaces. |
Found corner cases that failed when calculating the hash of dicts - fixes #1620.
Also found that list of lists of ints do not work in #1591.