Tom de Geus

Able to reproduce in https://github.com/xtensor-stack/xtensor/pull/2692 . I was scanning the code, it seems that it does not use `typeid`. Using it might make things safer?

Thanks. At first glance I do not spot a typo in the tests. Is there a typo in the code? Do you have time to debug?

Indeed, I think that there might be a 'typo' in the actual library. Indeed it seems that [`XTENSOR_CHECK_DIMENSION`](https://xtensor.readthedocs.io/en/latest/dev-build-options.html#build) is stricter than `XTENSOR_ENABLE_ASSERT`. Yet is might incorrectly identify https://github.com/xtensor-stack/xtensor/blob/807aa88e005925f27031b885038e595e2c14117a/include/xtensor/xhistogram.hpp#L591 as problematic....

It is probably `xt::view` that make it slow. What about doing the same as in NumPy? Probably ```cpp data -= xt::mean(data, 1); ``` (unchecked)

That's a bit surprising. How did you compile? How much slower?

As a reference, I am experimenting with running benchmarking to get on top of this in the future : https://github.com/xtensor-stack/xtensor-python/pull/288 . At the moment it is not completely obvious how...

A priori this is not something we support. If this is something that is important to you please consider a PR!

I think mostly that we 'configure' actually almost nothing. As such the `.in` extension that seems to be causing you problems is misleading and can be removed. But there may...

I think that what happens is that the return forces the evaluation of the expression. (It is a bit tricky to see from you error snippet)

I concur. I too have had my annoying small bugs or missing templates: - https://github.com/xtensor-stack/xtensor/issues/2643 - https://github.com/xtensor-stack/xtensor/pull/2569 - https://github.com/xtensor-stack/xtensor/pull/2561 That is also why I proposed - https://github.com/xtensor-stack/xtensor/issues/2511 What I was...