Tolerence of one second is too low when passing times to plot_fan

[billetd]

BUG

364 introduced a bug(?) when making fan plots by passing a datetime to the scan_index keyword. As tolerance works by searching records, not scans, the default of dt.timedelta(seconds=1) causes maybe only one beam to be selected for plotting, or non at all depending on the time you pick. Setting tolerence = dt.timedelta(seconds=60) fixes this, but this might mess with the plotting if you use an actual index for scan_index.

What should the default be? Should we have a different keyword for selecting a time?

Priority

• [x] Major - misrepresentation of data, failure in a common situation

[RemingtonRohel]

I don't agree that this is a bug, this was the intended design of the function find_records_by_datetime(). I chose a small tolerance as default for my use case of wide-beam experiments, but it is still fully functional for narrow beam experiments as you mentioned.

The function actually matches any records within tolerance of your scan_index datetime value, so with tolerance = 60 seconds you will get two minutes worth of records (60 seconds on either side), centered around the scan_index time.

I agree though that it would be cleaner to have two separate kwargs, since passing a datetime as scan_index means that you haven't exactly passed in an index, you've passed in a time. Definitely could be much clearer. This would mean handling some kwargs more carefully, but that's not too difficult.

[billetd]

Yes I agree find_records_by_datetime() is working as intended, though maybe a default shouldn't be for wide-beam at this time as the majority of data is narrow beam, 1-min scans. Two keywords would be best but I don't think its a priority right now, at least not as much as changing the default tolerance.

With tolerance=60, I just realised it was because I was looking at twofsound data, so you need 60 seconds there because of alternating channels.

[RemingtonRohel]

Fair enough, I agree that changing the default is a good idea.