On Fri, 30 Nov 2018, Sebastian Wagner wrote: > All, > > looks like the architecture of the mgr supporting independent modules > was quite successful :-). > > In recent times, I'm seeing an increasing amount of coped copied from > other mgr modules. This has the obvious downside that a fix in one > module leaves other modules unfixed. Like for example PR #25337. > > Running `pylint --disable=all --enable=duplicate-code *` within > pybind/mgr reveals, in my opinion, too many duplicates. One also have to > count the places where code was semantically copied (or inspired by). > > One obvious reason is that the mgr modules are prone to replicating > code, as each module is build by many distinct teams for a specific and > independent use case. Also, testing other affected modules can be quite > challenging. > > Therefor, I'd like to raise awareness. Maybe, we can reduce copies in > the future. My main question is what our strategy should be. Where should shared code live? Should we transplant it all into mgr_module.py, or into util.py or similar in the pybind/mgr directory? Current example: I'm using format_bytes() from the status module from the pg_autoscaler module via remote(), which is probably not ideal! s
