Closed eleftherioszisis closed 4 months ago
Why aren't the r
values the same?
Hmm, was reading the emits
, that's weird. I will have a look.
Hmm, was reading the emits, that's weird. I will have a look.
The difference in r
was actually a clue; switching to the double
build of MorphIO and the test passes.
It looks like we lose too much precision reading in those values as float32
s rather than float64
.
I don't really want to switch MorphIO wheels to the float64 build, because of the extra memory usage. Another option is to make the readers initially use doubles, and narrow the values when keeping the internal tree, but that would probably cause problems when round tripping. I think the best is to make it a relative tolerance - I will have to give it more thought.
This should be fixed with #494. I'm going to close this, please reopen if there are more cases like this.
It seems that the check for the three-point soma cylinder is too strict with its epsilon:
emits:
Using the values from the swc:
However, the point written in SWC is
1317.6848819729996
.This was encountered by @ssssarah with the SEU morphologies. Given that they raise exceptions on warnings to ensure there are no artifacts, this currently raises although the morphology is correct.