35 handle null values

[aadityaamehrotra17]

1. Manually specifying fill values is now possible. This can be done using the fill_values dict in get_constants().
2. If fill value is specified during variable creation as an attribute (using fill_value = X), the output data array doesn't seem to make use of this fill value. NCView interprets it as 0.
3. For the PMFINE group, just one fill value is specified for all species. Similarly, for NOX, one value is used for both NO and NO2

[douglowe]

What happens to the fill value if we apply the multiplying_factor after we replace NaN's / Nulls with the fill value? Does the fill value remain the same, or is it modified by the multiplying factor?

[douglowe]

I think it's probably safer to add the fill value after all the maths operations have been carried out.

[aadityaamehrotra17]

What happens to the fill value if we apply the multiplying_factor after we replace NaN's / Nulls with the fill value? Does the fill value remain the same, or is it modified by the multiplying factor?

According to me:

Fill Value first, maths later:
[None] -> [10] # fill value
[10] * 3 -> [30] # modified
Here * 3 was the maths op, so clearly fill value changes/gets treated as a normal value

Fill Value later, maths first:
[None] * 3 -> TYPEERROR?
Here * 3 was the maths op

(The error can be avoided with NAN, but there is a risk of propagation)

[aadityaamehrotra17]

With reference to this commit:
I have added the fill_value attribute for variables to show up in metadata. This is different from the actual _FillValue attribute, which can only be added during variable creation.