Skip to content

Master

maxigas requested to merge maxigas/uzta:master into master

The first commit fixed English language use and typos.

The second commit introduces the job library into the dependencies and uses it to parse group and user ontologies.

The job library can be installed on a current debian system like this:

apt install python3-joblib

Parallelization with the job library brings down the time of execution for the test case (which is -- i believe -- the Calafou group and the efkin user) from 30 minutes to 20 minutes. In order to achieve this, the reader functions have to accept an additional parameter holding the session object, and a new session have to be passed to each higher level function. Therefore, we have to log in and get a new session cookie each time we read a new ontology.

In conjecture that further speed up can be achieved in the future if parallelisation is switched from the ontologies to the lower level functions, because the former run on maximum 10 threads, since there are maximum 10 ontologies (e.g. in groups -- in users there are less).

Nota bene, however, that the real situation may be more complicated since on a dual core CPU jobs=2 yields similar results to jobs=10 or even jobs=64 -- but this does probably not invalidate the above conjecture.

The third commit fixes a typo in a comment.

Merge request reports