I replied on the other thread (missed this one). I believe that since the C++ codebase uses native endianness, we would need to add byte-swapping logic in the IPC code path to support big endian systems.
On Mon, Aug 1, 2016 at 9:31 AM, Sanjay Rao <getsanjay...@live.com> wrote: > Hi Micah, > Thanks, I am trying to write Java examples with Java libraries of Apache > Arrow on Big Endian system, I already got JARs built, this is part of a > research I am doing. > I ran "mvn test" on Big Endian system and it failed in TestEndianess, kindly > let me know if at least Java version of Apache Arrow will run on Big Endian > system. > Thanks,Sanjay > >> From: emkornfi...@gmail.com >> Date: Mon, 1 Aug 2016 09:23:51 -0700 >> Subject: Re: Is LittleEndian system mandatory for using Apache Arrow ? >> To: dev@arrow.apache.org >> >> Hi Sanjay, >> There was some discussion on the mailing list a little while ago about this >> [1]. We discussed making endianness part of the IPC metadata. I don't >> think we've gotten far enough in the implementation to take action on the >> discussion though. >> >> Right now I think the C++ code at least is endian agnostic, but it hasn't >> been tested against a big-endian system as far as I know. As referenced in >> the prior thread (and I don't think its been changed), there is an explicit >> check for endianness when constructing vectors in Java >> >> Cheers, >> Micah >> >> >> https://mail-archives.apache.org/mod_mbox/arrow-dev/201604.mbox/%3CCAK7Z5T_0T9hOa=mbJ=V30jc_Em169e=qjpyqnsblrakb8dm...@mail.gmail.com%3E >> >> On Mon, Aug 1, 2016 at 8:57 AM, Sanjay Rao <getsanjay...@live.com> wrote: >> >> > Hi, >> > I read that Integer representation is assumed to LittleEndian in Arrow, >> > does this mean we cannot use Arrow in Big Endian Systems like SPARC ? >> > >> > Thanks,Sanjay >
