• this avoids frameshift bugs when columns are added/removed
  • this is especially a problem when adding source-specific derived columns to a table that is then left-joined to another table
• it also avoids needing to confusingly use output column names in the input column of a left-joined table (to avoid inter-table collisions when reverting the column names to the original names)

fix¶

1. change util.set_col_names_with_metadata() to store the original column name in the column comment, and use this to match columns by name

alternative procedure (not as optimal)¶

1. reset the column names to the names in header.csv (which are positional, but always correspond to the actual columns)
2. apply the renames, matching columns by name