I see. So, what are the backward compatibility guarantees Arrow has moving forward?
On Mon, Nov 2, 2020 at 9:52 AM Wes McKinney <wesmck...@gmail.com> wrote: > No, you'd have to follow the project's pull requests or JIRA issues, > that's the only place where these things are discussed (except > occasionally on the mailing list). > > On Mon, Nov 2, 2020 at 8:50 AM Niranda Perera <niranda.per...@gmail.com> > wrote: > > > > Thanks Wes. Is there a document/ link for architectural decisions like > > this? (apart from a release change log, that is) > > > > On Mon, Nov 2, 2020 at 9:26 AM Wes McKinney <wesmck...@gmail.com> wrote: > > > > > Indeed, we made a change to cause buffers[0] to always be null when > > > the null count is 0, which has always been permitted by the columnar > > > format specification (and in 0.16.0 and prior it was inconsistently > > > null depending on how the array was created). > > > > > > On Mon, Nov 2, 2020 at 8:22 AM Niranda Perera < > niranda.per...@gmail.com> > > > wrote: > > > > > > > > Hi, > > > > > > > > We have been using arrow v0.16 and recently upgraded to v2.0. One > main > > > > difference we saw was that in v2.0, primitive arrays' buffer vector > > > > contains an empty buffer at idx 0 (whereas in 0.16 this was a null > bitmap > > > > IINM). > > > > Was this an architectural decision, may be to reduce the memory > footprint > > > > of the arrays? > > > > > > > > -- > > > > Niranda Perera > > > > @n1r44 <https://twitter.com/N1R44> > > > > +1 812 558 8884 / +94 71 554 8430 > > > > https://www.linkedin.com/in/niranda > > > > > > > > > -- > > Niranda Perera > > @n1r44 <https://twitter.com/N1R44> > > +1 812 558 8884 / +94 71 554 8430 > > https://www.linkedin.com/in/niranda > -- Niranda Perera @n1r44 <https://twitter.com/N1R44> +1 812 558 8884 / +94 71 554 8430 https://www.linkedin.com/in/niranda
