Some fields (e.g. TM,TN or TX) have non-continuous data, i.e. in most time-steps they only contain a "-" and no data. Only once or twice a day there is a number.
The node will always return NaN for those fields, because it tries to interpolate between two "-" or between one "-" and one number, which both doesn't work.
The most backwards compatible way of solving that would be to check for "-" before interpolating and returning either the value before or after (or any of those) instead of trying to interpolate. And only return NaN if both values are "-".
Obvious drawback: It forces the user to find the right look ahead time to get the value she wants.
I would prefer it to search for either the last or the next valid data point. Probably the last would be better.
Some fields (e.g. TM,TN or TX) have non-continuous data, i.e. in most time-steps they only contain a "-" and no data. Only once or twice a day there is a number. The node will always return NaN for those fields, because it tries to interpolate between two "-" or between one "-" and one number, which both doesn't work.
The most backwards compatible way of solving that would be to check for "-" before interpolating and returning either the value before or after (or any of those) instead of trying to interpolate. And only return NaN if both values are "-". Obvious drawback: It forces the user to find the right look ahead time to get the value she wants. I would prefer it to search for either the last or the next valid data point. Probably the last would be better.