Proposal: New option to distinguish null from empty string

[t-suwa]

It is critical to distinguish null from empty string for accomplishing data migration between different RDBMS using csv files.

Unfortunately, CSVParser does not support this functionality and returns ["", "", "", ""] as a result of parsing ",,\"\",". Thus, it impossible to make nullable columns have null instead of empty string when importing data from csv files.

So I'd like to add a new option to distinguish them.

• CSVFormat.withMissingColumnValuesAreNull()

With this option on, you can get [null, null, "", null] as a result of above input data. Its default value is off and CSVParser keeps backward compatibility.

[coveralls]

Coverage increased (+0.1%) to 93.168% when pulling 9412e00 on t-suwa:missing-column-values-are-null into b1e5d93 on apache:master.

[garydgregory]

What about:

Open Declaration CSVFormat org.apache.commons.csv.CSVFormat.withNullString(String nullString)

Returns a new CSVFormat with conversions to and from null for strings on input and output.
•Reading: Converts strings equal to the given nullString to null when reading records.
•Writing: Writes null as the given nullString when writing records.
Parameters:nullString the String to convert to and from null. No substitution occurs if nullReturns:A new CSVFormat that is equal to this but with the specified null conversion string.

?

[t-suwa]

Not sufficient. It does not provide a way to detect unquoted empty strings only. Take a look at the test case.

https://github.com/t-suwa/commons-csv/commit/9412e001d2e3d598edb6f51e051d2ef0116ad514

or am I missing something ?

[garydgregory]

I think this is now obsolete in git master, see commit ac280e7.