Update ulab

[jepler]

This gets us

• the new "approx" submodule
• the new "equal / not_equal" predicates (since we can't do array == array due to technical limitations)
• vectorise() for vectori(s/z)ing arbitrary Python functions with reasonably good performance

The specific submodule ref is in a pending PR v923z/micropython-ulab#122 so it should not be merged while that PR is not resolved.

[v923z]

This gets us

* the new "approx" submodule

* the new "equal / not_equal" predicates (since we can't do `array == array` due to technical limitations)

* vectorise() for vectori(s/z)ing arbitrary Python functions with reasonably good performance

This is definitely vectorize(). Since it was an original numpy function, I stuck with the American spelling:)

The specific submodule ref is in a pending PR v923z/micropython-ulab#122 so it should not be merged while that PR is not resolved.

The sub-module itself has been in master for approx. (pun intended) two weeks now.

[jepler]

The specific submodule ref is in a pending PR v923z/micropython-ulab#122 so it should not be merged while that PR is not resolved.

The sub-module itself has been in master for approx. (pun intended) two weeks now.

Overloaded terms. I mean to say, the git submodule ref for extmod/ulab is in the non-merged PR v923z/micropython-ulab#122.

[tannewt]

Why can't we do == and !=? We shouldn't introduce a new API just because we can't match MicroPython or CPython.

[jepler]

equal and not_equal are in numpy. https://numpy.org/doc/stable/reference/generated/numpy.not_equal.html

However, on some more full implementations of Python, == and != can be used for this functionality as well (I think there may be a minor difference in how non-finite values are handled)

[tannewt]

Thanks for the update!

[tannewt]

Good! Thank you!

[v923z]

Why can't we do == and !=? We shouldn't introduce a new API just because we can't match MicroPython or CPython.

@tannewt Here is what you need for the ==, and != operators to work: micropython#5593

I am not emotionally attached to the equal, and not_equal functions, so if you want, we can remove them. It was more like a temporary measure for circuitpython. As far as I remember, it was specifically a user request.

[tannewt]

Thanks @v923z. I opened #2999 for reasons to merge upstream changes.