Closed eboileau closed 5 years ago
Hi Etienne,
For another project, once I was reasonably confident that I wouldn't need any additional helpers from misc
(similar to rpbp
now), I just copied everything into that project and removed the dependency on misc
(since it ended up not actually being all that much). So one option could be to copy the relevant functions, modules, etc., into ribo_utils
and/or bio_utils
or something like that.
Another option could be to fork misc
and fix all the versions in the fork.
Thanks Brandon for your reply. I aim to spare some time by the end of the summer to address all these issues and others, and also do some cleaning/restructuring of rpbp
, i.e. providing a more light-weight default installation.
Yes, I also think it would be a good idea to copy over what is needed, or fork misc
, so that we are independent of one another; besides, as you alluded to it, there is many functions or even modules that are not used in rpbp
, and since those that are used are relatively robust (being largely wrappers around existing functionalities), this would make sense.
Sounds good. Just let me know if you run into any trouble. I think other than the misc.utils
module, the others are pretty much independent.
I'm closing this now, will take effect in the next release.
I think it would be a good idea to fix versions (i.e. package==version) of all major packages in rpbp and also for riboutils and bio_utils, but how this can be made consistent with misc? Any suggestions @bmmalone, since you are still owner for this repository? We discussed this at some point, but this was done for
pystan
only (not used in misc).This would entail that support is largely/only provided for installation in virtual environments, since otherwise users might end-up with clashes in version requirements for other applications. On the other hand, this would simplify a lot bug tracking, and we could then upgrade versions when needed at major releases, after thorough testing.
I opened this issue because I noticed that
joblib
has changed it's default backend frommultiprocessing
toloky
since version 0.12, and I am not sure yet of the implications (see #95 last comment/bottom, though this could be unrelated)...