Hi I prefre b.
I think it's more right to be compatible with users of the tree model and to be consistent with relational databases for the table model. Aligning the table and tree aligned time series semantically is something that can be sacrificed. Best ------------- Xinyu Tan On 2024/09/02 01:51:39 Yuan Tian wrote: > Hi all, > > As we discussed in [1], we need to decide whether to automatically filter > out rows that are entirely null in the result set during a scan by the > storage engine. > > There are three reasonable options: > > [a] keep table model same as tree model ( automatically filter out rows > that are entirely null ) > [b] table model don't automatically filter out rows that are entirely null, > but keep the tree model same as before. > [c] table model and aligned time series in tree model don't automatically > filter out rows that are entirely null, but keep nonAligned time series in > tree model same as before > > > Personally, I prefer [c], because in the storage engine, table model shares > the same things as aligned time series in tree model, [b] will break it up > while scanning. > > Feel free to vote for the option you personally want to choose. > > > > > > [1] https://lists.apache.org/thread/40gk1k1q3styjsf4s95ydlm5f2l2lcr4 > > > Best regards, > -------------------------- > Yuan Tian >
