search

gwastro / pycbc

Core package to analyze gravitational-wave data, find signals, and study their parameters. This package was used in the first direct detection of gravitational waves (GW150914), and is used in the ongoing analysis of LIGO/Virgo data.
http://pycbc.org
GNU General Public License v3.0
315 stars 350 forks source link

Infinite loop in ordering of waveform transforms #3762

Closed a-r-williamson closed 3 years ago

a-r-williamson commented 3 years ago

When applying the rigid rotational transform being worked on in #3756 I hit an infinite loop in the order_transforms function. This happened in a case when I defined priors on the cartesian spin components, which are both input to and output by the transform, and this was the only transform being applied. As a result this line resolves as False and we get stuck.

I would be interested in hearing suggestions about how to overcome this. Is this a bug that could recur in other cases that we should fix? (There look to be other transforms that have parameters common between their inputs and outputs, e.g.). Or would it be preferable to change our new transform to avoid this?

cdcapano commented 3 years ago

Interesting... I forgot we had transforms that can have shared input and output parameters (we really haven't used the other spin transforms; they were written as a part of a master's student's project, then never really put to use). The ordering function should be able to handle that. I think the easiest solution is to modify order_transforms to pull out new outputs rather than all outputs. In other words, change line 2268 to:

outputs = set().union(*[set(t.outputs)-set(t.inputs) for t in transforms])

I think that should do it.