Open i-pletenev opened 3 weeks ago
Thanks for opening the issues!
I think the bug you are describing is the same I ran into yesterday while fuzzying hictk with the new fuzzer (see https://github.com/paulsengroup/hictk/pull/226 https://github.com/paulsengroup/hictk/pull/227 if you are curious).
If it is the same bug, then the bug should only occur for cis queries where the end coordinate of range2 is greater than the end coordinate of range1.
The bug is due to an error in the code that mirrors interactions from the upper-triangle into the lower triangle.
I will try to write a bugfix by the end of the week. Will ping you here once it is ready :)
Thanks again for opening the issues!
Great, thank you for quick response!
Hi! In order to check consistency between
cooler.Cooler().matrix().fetch()
andhtk.fetch().to_numpy()
I sampled a few random regions fromcooler_test_file.mcool
and looked whether the result is identical. I found that in minority of cases the output differs even when two queried regions don't overlap. Here's an example:I decided to create a second issue, because there could be a different bug here than in #72
Once again, I checked it in both
0.0.5
version and0.0.6.dev50+g6eca918
version from github.