essoca

Many thanks @jpivarski for your quick response! Your same procedure (grabbing the `sig` after JIT-compilation and using it to type) doesn't work for me. It throws the error above. I'm...

Thanks for trying @ianna. You're almost there. Up to that point, it also works for me. Now test it with a different array with the same nested structure: ```python array2...

Thanks @ianna. Could you be a bit more explicit about what should be done for that to work? When `array` is initially created, it has a `numba_type`, its `_numbaview` field...

@ianna: it **partially** works. But executing `array2._numbaview` before passing `array2` to the kernel has no effect at all.  After putting a breakpoint at line 518, the...

@ianna - Thanks for the PR! Things now work as expected. If the signature type does not correspond to the array type passed to the kernel, the Numba dispatcher will...

Thanks for the info @leofang. What I don't get is that, if array-API compliant routines are implemented as shortcuts to `numpy.unique()`, then why not choosing the most efficient `unique()` implementation...

> That's an interesting error. How did you install `awkward`? Did you `pip` install from the Git repo perchance? Yes @agoose77, I pip-installed it a couple of weeks ago from...

@jpivarski: many thanks for the instructions :+1: Being able to slice large data in CUDA (following numpy syntax) is a very important operation, in my opinion. It would be cool...

@ianna - I'm currently using version 2.7.1 and confirm that there is no such issue anymore! Many thanks.