process.py: filtering of observations by time requires a guard

[rod-glover]

process.py, lines 130ff are

    observations = list(
        # Note: filter(None, <collection>) removes falsy values from <collection>,
        # in this case possible None values returned by align.
        filter(
            None,
            (
                align(sesh, row, is_diagnostic)
                for row in rows
                if start_date <= row.time <= end_date
            ),
        )
    )

and needs a guard on row.time to prevent errors:

    observations = list(
        # Note: filter(None, <collection>) removes falsy values from <collection>,
        # in this case possible None values returned by align.
        filter(
            None,
            (
                align(sesh, row, is_diagnostic)
                for row in rows
                if row.time is not None and (start_date <= row.time <= end_date)
            ),
        )
    )

or else revert to checking time after align has been called on the row. The original idea was to not cause it to do extra work on values that would be filtered out by time. This is still true but may be slightly on the side of over-optimization.

[jameshiebert]

Have you seen this happen in the wild? While it's technically possible to create a Row instance w/o a time value, in practice, I don't think that we have any code that can do so.

[rod-glover]

I haven't. I'm just a worry wart.