Spotted by @divine7022 in #4006 (comment):
we could end up with 1 row segments frame where start_date is set but every other column (end_date, crop_code, site_id, pft) is NA; and that gets returned to write_segment_configs, which then loops over row and tries to build segment configs with NA dates and NA pft. downstream error site is far from the actual problem
Since we already return an empty dataframe if there are zero rows before filtering (row 115 in the version we're discussing today) , it would be consistent to replace this error with a simple return(segments) -- possibly with a warning that no relevant segments were found? Today I lean towards just returning silently (If no segments are relevant to the requested time range, we can simply run the existing unsegmented config and it's not a problem), but am open to counterarguments.
Spotted by @divine7022 in #4006 (comment):
Since we already return an empty dataframe if there are zero rows before filtering (row 115 in the version we're discussing today) , it would be consistent to replace this error with a simple
return(segments)-- possibly with a warning that no relevant segments were found? Today I lean towards just returning silently (If no segments are relevant to the requested time range, we can simply run the existing unsegmented config and it's not a problem), but am open to counterarguments.