Hi All, There is a JIRA feature request ( https://issues.apache.org/jira/browse/CSV-293) for supporting multiple null strings, not just one.
To be honest, I did not really know much about the mentioned CSVW definitions, but based on the comment in the ticket "Feel free to provide a PR on GitHub" and the requested change seemed not a big deal to me. So I created the PR (https://github.com/apache/commons-csv/pull/456). After I created it, things have changed. I got a comment on github from Gary to check the comment in JIRA. He explained two concerns about the feature. So I started to dig into the topic and learn a bit about the CSVW. I think Gary has a point. This particular feature is far too little to support CSVW. As I understand it, CSVW introduces a lot of (let's say) validation rules (in a json metadata file) that are not really needed for a "simple" CSV parser. E.g.: - it introduces a feature to represent data like relational databases with multiple CSV files and calls them TableGroup. - the TableGroup contains multiple tables (CSV files) which are in relation. There are keys (ForeignKey, PrimaryKey). - there are Data Types (string as a base, but there are multiple number integer, non negative integer, float, boolean, date, datetime, etc.) - fields can be required And much more. So the feature to support multiple null values is not enough for supporting CSVW at all. And I agree with Gary. I can't see the real life use case where multiple null values were used in a simple single CSV file. Maybe in a tableGroup with multiple CSV files in a CSVW dataset it is useful (because the different files could contain different null values), but not in the level of a CSV file parser handles a single file. What do you think? Regards, David
