Fix scaling: take chromsizes from header

[Phlya]

Scaling without a view was broken! Docs said it would use the chromsizes from the header, but there was no code that was doing it...

[Phlya]

pairtools/tests/test_scaling.py

Line 27 in 8c41a26

) # double unmapped pairs are currently ignored by lib.scaling

Are single unmapped pairs somehow supposed to contribute to scaling? I think this might be why the test fails...

[golobor]

nah, they are never counted.

…

On Fri, 3 May 2024 at 15:34, Ilya Flyamer ***@***.***> wrote: https://github.com/open2c/pairtools/blob/8c41a26b3f3a3905e307f3baa4b547203394dca1/tests/test_scaling.py#L27 Are single unmapped pairs somehow supposed to contribute to scaling? I think this might be why the test fails... — Reply to this email directly, view it on GitHub <#239 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAG64CVEOSBP5SA34U4CKWLZAOG7NAVCNFSM6AAAAABHFQMVTKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDAOJTGAZTKMJSG4> . You are receiving this because your review was requested.Message ID: ***@***.***>

[Phlya]

Then why does the test assert 9 total pairs? It doesn't make sense to me, but somehow I guess this test was passing before?

[Phlya]

(As an aside, in the file there is a pair with both sides beyond the end of the chromosome... should that actually error or warn at least?)

[Phlya]

OK, I think I know where the problem is: when not chromsizes are provided, internally they are created from the data, and then there is a fake chrom "!" which I guess just behaves like any other chromosome...

pairtools/pairtools/lib/scaling.py

Line 143 in 8c41a26

if chromsizes is None:

So I would say the test is wrong in this case?

[Phlya]

this is the main change

[Phlya]

@golobor can we merge this?

[Phlya]

Shall we merge this? I realize the PR is full of blackification, sorry it's annoying...