On Tue, May 01, 2018 at 11:19:21PM +0200, Johannes Berg wrote: > On Tue, 2018-05-01 at 15:02 -0500, Seth Forshee wrote: > > > > > +import attr > > > > I'm a little hesitant to add use of non-standard libraries if it isn't > > necessary, as some distros have traditionally built the regdb from > > source (not sure how practical that is after the db-as-firmware changes > > though). Do we lose anything critical if we don't use attr? > > I probably suggested the use of attr. It's super useful, for things like > this: > > > > +@attr.s(frozen=True) > > > +class WmmRule(object): > > > + vo_c = attr.ib() > > > + vi_c = attr.ib() > > > + be_c = attr.ib() > > > + bk_c = attr.ib() > > > + vo_ap = attr.ib() > > > + vi_ap = attr.ib() > > > + be_ap = attr.ib() > > > + bk_ap = attr.ib() > > > + > > > + def _as_tuple(self): > > > + return (self.vo_c, self.vi_c, self.be_c, self.bk_c, > > > + self.vo_ap, self.vi_ap, self.be_ap, self.bk_ap) > > Spelling this out as a real object with all the __repr__ and comparisons > etc. gets far more verbose. > > While we can get rid of it in theory, it's a ~100KiB package without any > further dependencies, and the code is better off for it. > > Ultimately it's your decision, but I suspect that python is already such > a big dependency that adding this library is in the noise. I'm more thinking that for e.g. Debian if attr is installed via pip and not a distro package then that might be a problem for their package builds. But let's go ahead and leave it, if there's fallout we can figure that out later. Thanks, Seth
