add bool to NX_CHAR_OR_NUMBER

[rettigl]

Nexus states "Any valid character string or NeXus number representation", and thus I think it makes sense to also allow bool here.

[lukaspie]

I am not sure this is correct, since neither NX_CHAR nor NX_NUMBER only booleans. Then why would their union support it?

[rettigl]

My reasoning was that currently, there is no way in nexus that a field can accept any kind of data. I would expect that NX_CHAR_OR_NUMBER should be precisely that.

[rettigl]

I am not sure this is correct, since neither NX_CHAR nor NX_NUMBER only booleans. Then why would their union support it?

Technically, we could include booleans also in NX_INT, then they would be contained in NX_NUMBER

[lukaspie]

My reasoning was that currently, there is no way in nexus that a field can accept any kind of data. I would expect that NX_CHAR_OR_NUMBER should be precisely that.

nexusformat/definitions#1246 (comment) introduced NX_CHAR_OR_NUMBER as the union of NX_CHAR and NX_NUMBER. IMO, this does not include NX_BOOLEAN.

I see that it would be nice to be able to define a field can accept any kind of data. However, with the way the types are currently implemented, I don't think it is really possible. I think the best solution would actually be to be able to write unions of types, like type="NX_NUMBER|NX_CHAR" instead of inventing workarounds like NX_CHAR_OR_NUMBER. But this requires again a lengthy NIAC discussion.

The question is, what we do we do for now in pynxtools? Do we go with the literal interpretation of NX_CHAR_OR_NUMBER as it was defined or do we allow booleans as well to be more flexible? I think either one is fine, but other validators will of course make difference choices.

[rettigl]

We can keep it as it is, I'll close this. It was just for the conversion parameters and a bool flag there, but I can also use NXcollection there.

[rettigl]

@lukaspie Interestingly enough, bool is a subclass of int:

isinstance(True, int)
True

I am thus wondering why I got a warning in the first place, I will check that.

[rettigl]

This is a bit inconsistent:

print(isinstance(value.item(), int))
print(isinstance(np.bool_, int))
print(isinstance(np.bool_, np.integer))

True
False
False

[rettigl]

From https://joergdietrich.github.io/python-numpy-bool-types.html:

The bool_ type is not a subclass of the int_ type (the bool_ is not even a number type). This is different than Python’s default implementation of bool as a sub-class of int.

[lukaspie]

From joergdietrich.github.io/python-numpy-bool-types.html:

The bool_ type is not a subclass of the int_ type (the bool_ is not even a number type). This is different than Python’s default implementation of bool as a sub-class of int.

If bool is a subclass of int in Python, I think it's fine if we allow it,

[lukaspie]

LGTM, but linting needs to be fixed