Vladimir, I would recommend you start implementing physical plan generation and executor. I would recommend going to https://github.com/ApacheDrill/Brainstorm/wiki/Apache-Drill-Links and then going to the section named "Code generation / Physical plan generation" and start going through the material. Unfortunately Thomas cannot open-source his code but he said it should be a strait-forward implementation as everything covered in the paper.
I could help you with assembly and codegen if you need help and I assume Thomas would be glad to help you with any questions if something is unclear in the paper. >From my quick peek into Impala (https://github.com/cloudera/impala) it seems they also use C/C++ back-end and also codegen with LLVM IR for Hive queries so may be look there for inspiration, may be even beyond inspiration ;) the license is Apache... Cloudera are good guys, contributing a lot into Apache, supporting open-source, Doug is an Apache director so I guess we are safe here. Look into "Cloudera Impala" thread in this mailing list for some more discussion about Impala. Apache Drill has a slightly different goals but code reuse where it make sense is good I guess. And if you would want to help with the executor/storage engine itself you may consider contributing to ZeroVM, we just prepared a new version with easier installation: http://zerovm.org/download/ and stay tuned for private beta of hosted version. ZeroVM is also Apache licensed and Swift which we use as storage engine is also Apache licensed. On Mon, Nov 12, 2012 at 7:23 PM, Vladimir Starostenkov < [email protected]> wrote: > Hello! > > On Jason's slides I saw "C/C++ core with Java support". Are there any more > architecture details? I've already read the "Apache Drill Plan Syntax", but > as I see it's about the stuff you are working on right now in Java. > What about Storage Engine? DRILL-13? Where to help with C++ expertise? >
