--
You received this message because you are subscribed to the Google Groups "xarray" group.
To unsubscribe from this group and stop receiving emails from it, send an email to xarray+unsubscribe@googlegroups.com.
To post to this group, send email to xar...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/xarray/267463a1-95c3-44ec-b0d4-1e2da9743c5d%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
The short answer is that nothing has been added into xarray proper yet. I think this GitHub issue (https://github.com/pydata/xarray/issues/585) is still up to date.
For cases where the inner function takes and returns numpy arrays, there is a good integration point with apply_ufunc (not yet public API). Your email prompted me to finish up this work into a pull request, which will probably make it into a future release: https://github.com/pydata/xarray/pull/1517Cases where the inner function takes and returns xarray objects are harder. The challenge is that for these cases all the xarray metadata (variable names, dimension names and sizes, index labels, attributes, etc.,) needs to be known in order to construct the result out of dask.array objects. This could either be done by providing some sort of explicit schema (which is rather painful), or could perhaps be inferred from the result of evaluating the inner function on a single block.I still think this could be a viable option, I just never got around to figuring it out. Pull requests would be welcome.Best,Stephan
On Tue, Aug 22, 2017 at 11:46 AM Noah Brenowitz <nbr...@gmail.com> wrote:
I would like to be able to map some function in a parallel fashion over some subsets of the data. This could either be achieved by a groupby/apply or a wrapper around dask's map_blocks function. Ideally, the inner function would take DataArray objects as an input. However, there are bunch of issues/pull requests that seem relevant to this, and I was having trouble parsing the discussion. Is there any basic support for embarrassingly parallel computations in xarray that are more complicated than simple arithmetic?--Thanks,Noah
You received this message because you are subscribed to the Google Groups "xarray" group.
To unsubscribe from this group and stop receiving emails from it, send an email to xarray+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/xarray/b0cb3816-8580-40da-8b52-1a1260ac5e0e%40googlegroups.com.