Import of `as_seqdict`

[simonvh]

@siebrenf Is it necessary to have as_seqdict import in __init__.py?

It gives an error when genomepy is imported in another module due to a circular import of pybedtools (I think). Removing it from __init__.py fixes the issue.

[siebrenf]

It is not necessary.

I simply thought it was nice to expose this function to the surface level API as
genomepy.as_seqdict() vs genomepy.genome.seqdict.as_seqdict()

[simonvh]

Yes, that's what I thought. I would indeed be nicer to have a shorter import, but this somehow results in weird import issues.

[siebrenf]

Bad news. Without importing as_seqdict in anywhere in genomepy it will not be accessible. When importing it, it will always cause a circular import error.

bottom line is we can't import pybedtools inside genomepy, unless its an older version.

[siebrenf]

We could make the bybedtools option available if pybedools has already been imported with
if "pybedtools" in sys.modules:

[simonvh]

I think #174 may fix it. It's not too pretty, but I think importing in any order should work now. I also tried putting the import back in genomepy/__init__.py but that doesn't work.

[siebrenf]

I think it doesnt work :(

>>>import genomepy
>>>genomepy.genome.seqdict
Traceback (most recent call last):
  File "<input>", line 1, in <module>
AttributeError: module 'genomepy.genome' has no attribute 'seqdict'

As far as I understand it, we have to import the as_seqdict function, or its module, somewhere in genomepy. Otherwise it isn't reachable.
That also means the code is read, thus pybedtools is imported, and the recursion happens again...

So I think the "solution" it to remove the try import.

[simonvh]

This does work:

>>> from genomepy.genome.seqdict import as_seqdict
>>>

[siebrenf]

Guess it's an "advanced feature" then ;)

Should we still move the seqdict module to a higher level?
I mean change it from from genomepy.genome.seqdict import as_seqdict to from genomepy.seq import as_seqdict

[simonvh]

Yes, I think that's a good idea.