Deepak Cherian

@gjoseph92 (and everyone else involved here), thank you! How do I test it out ;)? ``` with dask.config.set({"distributed.scheduler.worker-saturation": 1.5}): result.compute() ``` Is this right? What range of values should I...

They are still hyperlinks, but all of them link to the first page of the combined PDF.

I had a similar issue with dask jupyter dashboard. Downgrading to tornado 5.1.1 fixed it. On tornado 6.0.1, the browser's console showed websocket connection errors. This was on jupyter-server-proxy 1.0.0...

I had a similar issue with dask jupyter dashboard. Downgrading to tornado 5.1.1 fixed it. On tornado 6.0.1, the browser's console showed websocket connection errors. This was on jupyter-server-proxy 1.0.0...

I had a similar issue with dask jupyter dashboard. Downgrading to tornado 5.1.1 fixed it. On tornado 6.0.1, the browser's console showed websocket connection errors. This was on jupyter-server-proxy 1.0.0...

You can check whether it computes using the following ```python from xarray.tests import raise_if_dask_computes with raise_if_dask_computes(): ds.to_dask_dataframe() ``` Does that raise an error?

Well then it isn't computing. Depending on the shape of your DataArray, there's potentially a reshape involved which can be expensive in parallel: https://docs.dask.org/en/stable/array-chunks.html#reshaping . It does manifest as large...

> I thought we could avoid the problem they discuss by making the chunks take up the entire width of the array using chunks=(100, -1) This is surprising. The reshape...

Cool experiment Tom. > Generalise xarray to allow for variable-length dimensions This is somewhat similar to supporting nan-shaped dask arrays (https://github.com/pydata/xarray/issues/5168, https://github.com/pydata/xarray/issues/2801).

Thanks for trying out our pre-release @lukasbindreiter ! This is an intentional change. Can you tell us more about why this breaks your code?