Change the join to LEFT JOIN

[mgdenno]

This join includes unmatched rows. This is a feature that has been requested. I don't know the ramifications of doing this, so it needs testing. At a minimum we would need to filter and null value rows before calculating metrics i suspect.

joined_df = ev.sql("""
            SELECT
                sf.reference_time
                , sf.value_time as value_time
                --, pf.location_id as primary_location_id
                , cf.primary_location_id as primary_location_id
                , sf.location_id as secondary_location_id
                , pf.value as primary_value
                , sf.value as secondary_value
                , sf.configuration_name
                , sf.unit_name
                , sf.variable_name
                , sf.member
            FROM secondary_timeseries sf
            JOIN location_crosswalks cf
                on cf.secondary_location_id = sf.location_id
            LEFT JOIN primary_timeseries pf
                on cf.primary_location_id = pf.location_id
                and sf.value_time = pf.value_time
                and sf.unit_name = pf.unit_name
                and sf.variable_name = pf.variable_name
        """)

[samlamont]

This also impacts the joined timeseries table when adding attributes

from Katie:

When I join timeseries with add_attrs = True, it only includes locations in the joined table that have attribute values defined (only 198 of 315 gauges included and no MAP data included). All timeseries are included in the joined table when I set add_attrs to False, but no attributes. I need all timeseries joined and attributes added where they exist.