Problems accessing scipp.constants

[arm61]

The following raises an error for me:

import scipp as sc
sc.constants

But works if I do

from scipp import constants

Scipp version: 24.9.1

[SimonHeybrock]

Hi! This is by design. Similar to, say SciPy submodules, it is a submodule that needs to be imported explicitly. One reason for this is that it relies on Scipy, which is an optional dependency of Scipp.

We will consider adding lazy loading, please see #3567.

For now, if you would like to use sc.constants in code, you can

import scipp as sc
import scipp.constants

sc.constants.h  # works now

[SimonHeybrock]

Seemed to be all: https://github.com/scipy/scipy/blob/4d7413563cfa477c08717daa61e5994e968ee79f/scipy/__init__.py#LL15C1-L16C26

[jl-wynen]

Scipy imports modules lazily when you access them: https://github.com/scipy/scipy/blob/1adc8cb7a37e09444632a86afea5ee1d07f3e3c5/scipy/__init__.py#L134
We could do the same but I am not sure it is worth the complication.

[SimonHeybrock]

There is actually a SPEC about it: https://scientific-python.org/specs/spec-0001/

[arm61]

Does a SPEC mean that you will change in the future?

[SimonHeybrock]

It is something the "Scientific Python" community agreed on. So there is nothing forcing us to do so, but it would make a lot of sense. Please subscribe to #3567 for more updates on the topic.