data type constraints issue

[X-Fan-Jack]

Dear developer,

I am using Python and trying to calculate the soundscape index.

the code is

ssid, excl = sspy.isd.validate(data, allow_paq_na=False)
ssid = sspy.surveys.add_iso_coords(ssid)

Then I found that some records might have date-type issues, which results in a long number, as shown below.

I'm speculating that the reason for this is that numpy did a big update(v2.1.0) on August 18th. This may lead to an update of some of the underlying methods.
Although users can add their own data type checks, if you can add data type constraints and transformation rules to the library, this will reduce the potential use risk and make the project more robust.

Best regards!

[MitchellAcoustics]

Hi, thanks for submitting!

I don't think the numpy update should be the issue - numpy 2.1 should be outside the allowed range based on other dependencies anyway (specifically pandas and numba). I can't even force my environment to install numpy 2.1.0, since the numba version installed with pandas requires numpy<=2.0: ImportError: Numba needs NumPy 2.0 or less. Got NumPy 2.1.

How are you setting up your environment? A fresh install of soundscapy should by default be using less than numpy 2.1.

That said, this isn't really an issue with the data type. The ISO values are floats (float64), which is correct. These just happen to be essentially zero, which I think due to floating point calculations ends up just 'close to' zero. We could solve it by adding a round to the surveys.processing.calculate_iso_coords function, but I'm not sure I see the necessity.