Rows starting with comment char should be automatically quoted

[adutra]

I ran into a corner case when exporting/importing CSV data. Consider the following format and parse operations:

StringWriter sw = new StringWriter();
{
  CsvWriterSettings writerSettings = new CsvWriterSettings();
  CsvWriter writer = new CsvWriter(sw, writerSettings);
  writer.writeRow(new String[] {"#field1", "field2", "field3"});
  writer.close();
}
StringReader sr = new StringReader(sw.toString());
{
  CsvParserSettings parserSettings = new CsvParserSettings();
  CsvParser parser = new CsvParser(parserSettings);
  List<String[]> rows = parser.parseAll(sr);
  System.out.println(Arrays.toString(rows.get(0)));
}

The above will fail because the written row happens to start with the configured comment character #.

I am currently using the following workaround when exporting data:

writerSettings.setQuotationTriggers(writerSettings.getFormat().getComment());

This makes the code above work as expected.

But I think it would be a good idea to make CsvWriter automatically quote the first field of a row if its value starts with the configured comment character. Because otherwise the written row is likely to be ignored when read with the same settings.

[jbax]

Agree 100% and not even sure how we got this far forcing people to do that.
It will be less bad if you do this for writing:

writerSettings.getFormat().setComment('\0');

and for parsing:

parserSettings.getFormat().setComment('\0');

But I'll adjust to allow writing the first value without having to set the comment character to \0

Thanks for the suggestion

[adutra]

parserSettings.getFormat().setComment('\0');

This won't always work. Believe me or not, the initial issue I ran into was precisely with a row starting with \0, as in "\0foo,bar,qix" and my parser was configured with \0 as the comment char (or rather, was configured to consider no comment char). That row got written nicely, but couldn't be read.

If it's feasible for you, I would avoid mixing these two notions: \0 as the comment char, and no comment char at all. But that's another issue.

[jbax]

You can also use the 0xFF character, which is probably more appropriate but less common.

…

On Fri, 22 Nov 2019 at 19:12, Alexandre Dutra ***@***.***> wrote: parserSettings.getFormat().setComment('\0'); This won't always work. Believe me or not, the initial issue I ran into was precisely with a row starting with \0, as in "\0foo,bar,qix" and my parser was configured with \0 as the comment char (or rather, was configured to consider no comment char). That row got written nicely, but couldn't be read. If it's feasible for you, I would avoid mixing these two notions: \0 as the comment char, and no comment char at all. But that's another issue. — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub <#362?email_source=notifications&email_token=ABWFQPRYAVRDMPT4ITIM2QDQU6LNTA5CNFSM4JPHCSMKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEE46YJA#issuecomment-557444132>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABWFQPR4GXJMFRYC6ZU2SS3QU6LNTANCNFSM4JPHCSMA> .

[adutra]

That's an option too, but I sometimes deal with data encrypted with weird algorithms, and 0xFF is unfortunately likely to appear as well. That's why I came up quoting whichever character is set to be the comment character. At least it's bullet-proof :-)

[jbax]

Implemented in the latest 2.8.4-SNAPSHOT. Thanks!