Support for sep. metadata in CSV files

[DestBro]

As a developer, I want support for the "sep" metadata in CSV files. It's used to indicate the CSV separator symbol.

I need this feature so that the CSVs files generated by our system are easily read by the library without modifying them each time.

[Jolanrensen]

To give some context: (I think) this is a csv-dialect of Excel: https://www.ablebits.com/office-addins-blog/change-excel-csv-delimiter/ though it may be used in other programs to a lesser extent as well.

[kosak]

Hi, thanks for opening this issue. I wasn't aware of this special Excel dialect.

I think we could support this, but it will have to go through triage first compared to other priorities.

However I would also point out that since CsvReader.read takes an InputStream, you could easily support this yourself:

• take your input file
• read the first line
• parse the sep= information
• pass the InputStream and configured CsvSpecs to CsvReader for processing.

The deluxe version of the above would be made to work whether or not there was a sep= line with a custom InputStream that either allowed you to peek at the first line without consuming it, or one that allows you to consume a line but then push it back if you don't want to consume it.

(edit: let me know if this doesn't make sense or doesn't work for your use case)

[DestBro]

What you say makes sense, the implementation of this feature can be done on our end as well - it's easy, but it's not the problem;

Having a centralized lib specialized in parsing different kinds of CSVs is the goal, I like the idea of single source of truth, rather than handling new parsing scenarios as side effects outside of the lib each time - we can do it in a single place, so other people can use it as well

[DestBro]

The point you made: “to be written in a way that works whether or not there is a sep= line” - is true; this behavior is desired, especially for backward compatibility and ease of use