hi Sebastien, Seems reasonable, and the integration between the libraries could be useful (although npyio could be made to depend on the Arrow Go libraries)
If there is the desire from the Arrow Go developers to merge in that codebase, it seems perhaps old enough that we might have to do an IP clearance -- we've done enough of these that it's become fairly routine at this point. I've conducted so many IP clearances (7, according to http://incubator.apache.org/ip-clearance/) that I would really appreciate if another PMC member could assume ownership for doing one. - Wes On Fri, Dec 7, 2018 at 2:50 AM Sebastien Binet <bi...@cern.ch> wrote: > > hi there, > > Back in the days, for my particle physics work, I had to be able to read > and write numpy data files[1] (for interop with existing analysis > pipelines.) > > I was wondering whether I couldn't integrate this Go package with Apache > Arrow: > - https://github.com/sbinet/npyio > > and put it under github.com/apache/arrow/go as either > - github.com/apache/arrow/go/npyio, or > - github.com/apache/arrow/go/arrow/npyio (as a sibling of the already > existing csv package.) > > that new npyio package would have builtin support for array.Interface and > tensor.Interface. > > thoughts? > > cheers, > -s > > [1]: http://docs.scipy.org/doc/numpy/neps/npy-format.html
