If a column only has null values, wrangler will drop that column from its schema.
1. Create a workspace on the attached input file
2. Use the following directives:
parse-as-csv :body ',' false
set-column body_3 "null".equals(body_3) ? null : body_3
3. Create a pipeline
Note that the body_3 column gets removed from the output schema. Also note that the type for body_3 shows up as 'unknown' in the wrangler UI.
This is likely because wrangler derives schema from data. This clearly is not a good strategy for nulls.