Please add ulab.numpy.absolute function

[jamesra]

The abs function from numpy would be nice to add to ulab.numpy. There is a lot of numpy functionality missing for good reason but this function feels particularly fundamental and it would be convenient to have it built-in. Thanks.

[github-actions]

Heads up @v923z - the "ulab" label was applied to this issue.

[v923z]

@jamesra The absolute value unary operator was one of the first operators implemented:

https://github.com/v923z/micropython-ulab/blob/3e996d9bd93005edbbb9763c4a0a525fe52dec3d/code/ndarray.c#L1968-L2000
Or do you have something else in mind?

In any case, it's easier to track issues and requests, if you raise them directly in https://github.com/v923z/micropython-ulab/issues.

[jepler]

I think the request is for ulab.numpy.abs to exist. It's true that the builtin abs function works on ulab arrays, though.

>>> ulab.numpy.abs
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'module' object has no attribute 'abs'

[v923z]

OK, but then is it worth the flash space? If there is a good reason, we can easily add the function.

[v923z]

So, if this is the function in question https://numpy.org/doc/stable/reference/generated/numpy.fabs.html, then the difference to abs is that it can support keyword arguments, which are not supported in ulab, anyway. The only useful argument there is out, I believe, which we could add. But beyond that, I don't see any advantages compared to abs.

On the other hand, math functions always return a floatint point array, while the abs unary operator keeps the dtype.

[jamesra]

This may be a documentation issue instead. abs is only listed on the page for complex arrays. I tried np.abs and np.absolute, but I know I did not try calling it directly off my array. Edit: I didn't think to use the interactive console to run dir(myarray) to find available methods.

np.abs is probably still worth having, but not necessary if it is exposed on the array object itself. (My apologies if abs is documented somewhere else and I just missed it.)

[jamesra]

OK, at the risk of sounding dumb, what is the syntax for the unary absolute value operator in numpy? (I am using 8.1.0-beta 2 to obtain analogbufio on the ESP32-S3)

I launched an interactive Circuit Python session and there is no abs function on the array:

Adafruit CircuitPython 8.1.0-beta.2 on 2023-04-26; Adafruit Feather ESP32S3 4MB Flash 2MB PSRAM with ESP32S3
>>> import ulab.numpy as np
>>> a = np.array([1,-2,3,-4])
>>> dir(a)
['__class__', 'copy', 'sort', 'byteswap', 'dtype', 'flat', 'flatten', 'itemsize', 'reshape', 'shape', 'size', 'strides', 'tobytes', 'tolist', 'transpose']
>>> a.abs
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'ndarray' object has no attribute 'abs'
>>> a.absolute
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'ndarray' object has no attribute 'absolute'

[jepler]

MicroPython 8.1.0-beta.2-43-g4901bdceb7-dirty on 2023-05-18; linux version
Use Ctrl-D to exit, Ctrl-E for paste mode
>>> import ulab.numpy as numpy
>>> abs(numpy.array([-1, 2, -3]))
array([1.0, 2.0, 3.0], dtype=float64)

It corresponds to the abs built-in function. In another post v923z referred to it as a "unary operator", but that reflects some implementation detail of micropython, it's not actually a unary operator like - is in a Python statement like x = -y.

[jamesra]

Oh!!! I'm trying to decide how I feel about that. On one level it makes total sense to use the built-in language operator. On another level it is some lateral thinking to step outside of the numpy library to do one operation... My suggested fix would be documentation. Thank you for explaining the intended use.

[v923z]

My suggested fix would be documentation. Thank you for explaining the intended use.

This has actually been documented, unless you have something else in mind: https://micropython-ulab.readthedocs.io/en/latest/ulab-ndarray.html#unary-operators

[dhalbert]

Closing because I think this has been resolved.