Jan, I'm not sure you and Richard concur, or maybe you and I don't concur on the meaning of the word "concur." ;^)
"A closed branch should be a branch in which all leaves are closed" is not the definition Richard advocated. If I understand him correctly, he said to use the "/brlist definition" which considers only the newest leaf, because "[i]f you have multiple branches with the same name, then you probably deserve whatever it is that happens," and "[t]he /brlist definition is easier to compute." I'll do either one, I just need buy-in because backward compatibility is being broken, albeit in an obscure corner case. Though I definitely agree with Richard about the ease of computation. Look at my code, see what it takes to get the classic branch ls behavior, look at the massive slowdown incurred. As for multiple branches, they're not unheard-of, e.g. "mistake" and "pending-review". At the risk of derailing this thread, I admit I am quite curious about that slowdown. Am I doing something wrong? Would an index help? Is this an inherently slow thing to do? Or... is SQLite calculating the closed column prior to grouping rather than waiting until the last minute, and furthermore failing to notice that it will always have the same value for all rows within a given group since it's a function of only joined tables and the argument to GROUP BY? One final thing. Richard, I don't have commit access to SQLite since I never requested it, but I did sign its contributor agreement at the same time as I signed Fossil's agreement. Would you welcome my attempt to incorporate your words about the special behavior of max() and min() into the SQLite documentation? On Nov 19, 2016 11:55, "Jan Nijtmans" <jan.nijtm...@gmail.com> wrote: > Op 19 nov. 2016 6:25 p.m. schreef "Richard Hipp": > > On 11/19/16, Andy Goth wrote: > > > Going forward, which definition do we want? > > > > I think the definition used by /brlist. > > > > Reasoning: > > (1) I don't think the difference is all that important. If you have > > multiple branches with the same name, then you probably deserve > > whatever it is that happens. > > (2) The /brlist definition is easier to compute > > I concur. A closed branch should be a branch in which all leaves are > closed. That's what I would expect. > > Regards, > Jan Nijtmans > > _______________________________________________ > fossil-dev mailing list > fossil-dev@mailinglists.sqlite.org > http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/fossil-dev > >
_______________________________________________ fossil-dev mailing list fossil-dev@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/fossil-dev
