Thought came to mind chatting with @frheault regarding #33.
Particularly with fast implementations that can omit certain bounds checking, malformed data can result in very unusual bugs.
What would help here, similarly to MRtrix3/mrtrix3#2418, would be to have a dedicated command that performs a more exhaustive check of the raw data. So e.g. make sure that streamlines don't overlap, don't exceed the bound reported in the header, that sort of thing. Streamlines with zero or one vertex should be counted and reported; such things should actually be legal and are highly useful in some circumstances, but in other circumstances they may be indicative of a problem.
arokem
commented
8 months ago
Thought came to mind chatting with @frheault regarding #33.
Particularly with fast implementations that can omit certain bounds checking, malformed data can result in very unusual bugs.
What would help here, similarly to MRtrix3/mrtrix3#2418, would be to have a dedicated command that performs a more exhaustive check of the raw data. So e.g. make sure that streamlines don't overlap, don't exceed the bound reported in the header, that sort of thing. Streamlines with zero or one vertex should be counted and reported; such things should actually be legal and are highly useful in some circumstances, but in other circumstances they may be indicative of a problem.