I'd like to see a proof of concept of moving data from R to Pandas and back using Apache Arrow. Something similar was asked not too long ago: Apache Arrow: shared in-memory data object. But I thought I'd jump in and see if there are any updates/worked examples/clearer explanations.
I passed along this link to some of our Solutions Engineers to take a look at, seeing if we can get some examples!
Neal on the Arrow team pointed out an issue they are working on.
That would be a good item to watch, comment, and/or contribute to.
Thanks! An example of zero copy with reticulate would be awesome too. I'll follow that issue with interest!
But for now I'd settle for independently in entirely separate R and Python processes manually passing the memory address from one to the other. But, in particular a whole table/tibble (Single arrays and atomic items are most of the examples I've seen so far). But it looks like a fair bit of boilerplate needs to be written too.