Alexander Clausen

Seems to be a duplicate of https://github.com/scikit-learn-contrib/hdbscan/issues/457 - which is fixed in master, but not yet released on PyPI

It looks like you are missing the header files for your Python installation. How to install them can vary depending on the version of Python you are using, and on...

I would also be grateful for a new release, especially #457 is a bit painful, if the workaround of installing the current numpy version is not possible in a certain...

> are there negative consequences? One consequence of having one test per cell is that `pytest --ff` doesn't work properly - it tries to execute the cells out-of-order. Other than...

> Or only on demand? For our use case, restart on-demand would work best: a user can decide to stop a web service they don't need anymore, and it won't...

> Is the automatic restart already working in case a process is missing or got killed? I'm using jupyter-server-proxy to execute code-server but if the process gets killed I keep...

Quickly testing with `read_direct` reveals that still a `memmove` is done:  This is a flamegraph acquired with linux `perf` of just using `read_direct` to read a whole 8GB dataset...

Thank you for your response! @stavrospapadopoulos disabling the cache already gave a nice speedup, thanks for the tip! Looks like you are on the right track efficiency-wise! Two remarks: *...

There are binary wheels for h5py [available](https://pypi.org/project/h5py/#files) if that helps, and the nexusformat package only depends on numpy and h5py as far as I can see. Otherwise the question would...

cc @uellue