Patrick Hoefler

Results 345 comments of Patrick Hoefler

Change default shuffle method without a cluster to tasks

> So you're saying that the performance of tasks comparable to disk on larger-than-memory datasets. Is that correct? What I am saying is that this is not something we should...

0.0.6: pep517 build and compile time warnings and .c and .h files are added into .whl

Do you have time to take a look?

Zetas- Ram crashes dask (GPU/CPU)

@dask/gpu

Index Query Hangs

Thanks! Could you created your dataframe in a way that we can reproduce this?

Avoid the costly rechunking for 1D-arrays in cumreduction

You don't have to check for performance, you can look at the graph and check that we don't have any rechunks in there for example

Avoid the costly rechunking for 1D-arrays in cumreduction

thanks

Document that `Series.unique` order differs from Pandas

This is expected. We are triggering a shuffle under the hood to avoid overloading a single worker. And a shuffle won't preserve the input order. I'll label this as a...

test_tokenize failures in 2024.8.1

Investigations are welcome

different `run_spec` between consecutive calls to `update_graph` | zarr-formatted xarray

Any chance you could try with the newest dask release? The MCVE doesn't warn for me anymore

Error using db.from_delayed with sparse arrays

Thanks for your report. Any advice on how we can make this work without adding scipy as a dependency for bags?