bugfix: inputs/VegBank/taxon_observation.**/postprocess.sql: removed row_num (=identificationID), because there is actually more than one row per VegBank taxonobservation_, so this does not properly enumerate the view rows. this is because there is a 1:many left-join to stemcount_, stemlocation_ which adds rows to each taxonobservation_. since the row_num is gone, any row-subsetting of the view using OFFSET will always need to materialize the entire view up to the OFFSET value. this works for smaller datasources like VegBank that fit almost entirely into one column-based import chunk (1 million rows), but not for larger datasources like FIA where it would be much slower to materialize all preceding 16 million rows on the last chunk (which is what OFFSET normally does with left-joins).
bugfix: inputs/VegBank/taxon_observation.**/postprocess.sql: removed row_num (=identificationID), because there is actually more than one row per VegBank taxonobservation_, so this does not properly enumerate the view rows. this is because there is a 1:many left-join to stemcount_, stemlocation_ which adds rows to each taxonobservation_. since the row_num is gone, any row-subsetting of the view using OFFSET will always need to materialize the entire view up to the OFFSET value. this works for smaller datasources like VegBank that fit almost entirely into one column-based import chunk (1 million rows), but not for larger datasources like FIA where it would be much slower to materialize all preceding 16 million rows on the last chunk (which is what OFFSET normally does with left-joins).