Closed glennhickey closed 5 years ago
You can add -r 1 to the seqwish call. Is the cyclicity causing problems?
On Mon, Jul 8, 2019, 15:26 Glenn Hickey notifications@github.com wrote:
How feasible would it be to force one of the input sequences (ex hg38) to be acyclic in the output graph?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/ekg/seqwish/issues/19?email_source=notifications&email_token=AABDQELSJM2IK2EJFLXIG6TP6M6ATA5CNFSM4H63V7H2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4G53IJNQ, or mute the thread https://github.com/notifications/unsubscribe-auth/AABDQEMKYA4ZDI4FLPJTR3LP6M6ATANCNFSM4H63V7HQ .
Yeah, they're causing trouble for vg deconstruct
, it seems. I was assuming that cycles get spanned by snarls (which is true), but I also have logic where if a snarl is too big to handle, it recurses on child snarls. These snarls can be within cyclces, and I'm not handling that case properly. I'm trying to figure out the best way to fix deconstruct now, but was curious if I could just flatten things out in the meantime
That makes sense. You can try this. It will prevent cycling by stopping the transitive closure if it encounters the same sequence twice.
On Mon, Jul 8, 2019, 15:46 Glenn Hickey notifications@github.com wrote:
Yeah, they're causing trouble for vg deconstruct, it seems. I was assuming that cycles get spanned by snarls (which is true), but I also have logic where if a snarl is too big to handle, it recurses on child snarls. These snarls can be within cyclces, and I'm not handling that case properly. I'm trying to figure out the best way to fix deconstruct now, but was curious if I could just flatten things out in the meantime
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/ekg/seqwish/issues/19?email_source=notifications&email_token=AABDQEIY6Z5JN3KNZHTVWJDP6NAM5A5CNFSM4H63V7H2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODZNEFVA#issuecomment-509231828, or mute the thread https://github.com/notifications/unsubscribe-auth/AABDQEJIUYJGEBNOQYKX6ATP6NAM5ANCNFSM4H63V7HQ .
Did -r 1 solve this problem for you?
Hmm, I honestly don't remember. I think I fixed deconstruct to be more robust (judging from the name of this commit https://github.com/vgteam/vg/commit/ad7d26c8007d20ea14a74bec1a75fc29ff4a5e82#diff-d905c05327c6f07e310f14d41c42d84f0 ).
I do remember being blocked on #18. I guess I should try that data with the latest seqish (it's been a while) to see if still comes up
How feasible would it be to force one of the input sequences (ex hg38) to be acyclic in the output graph?