encoding/csv: Provide a mean to change the quote character

[gopherbot]

by fuzxxl:

The package encoding/csv recognizes quoted fields. Sadly, it's not possible to change
the quote character to something different, as required for some use cases [1].

I request that encoding/csv be expanded to allow users to select a custom quote
character.

[1]: http://stackoverflow.com/questions/25062281/golang-enclosure-rule-for-csv-parsing

[ianlancetaylor]

Comment 1:

Labels changed: added repo-main, release-none.

[adg]

Comment 2:

This could be done by adding a SingleQuote boolean field to Reader and Writer. False,
its default, would indicate the usual double-quoted behaviour.

Labels changed: added suggested.

Status changed to Accepted.

[gopherbot]

Comment 3 by fuzxxl:

I think this would be a bad fix. The next time someone with an entirely different quote
character appears, this approach doesn't work anymore.
Why not add a field
    QuoteCharacter rune
Which defaults to double-quotes if set to 0?

[minux]

Comment 4:

What if people then want different quote characters for start and end
quotes?

[satran]

I've submitted a fix: https://go-review.googlesource.com/#/c/1576/
The fix is only for the reader. I was wondering if the writer should also have an option to specify the quote character.

[bradfitz]

Why?

Is there a standard that says the quote character can differ?

If you have a file format that uses a different quote character, is it really CSV?

I don't think so.

I'm compelled to close this without action. The encoding/csv package is small and easily forkable elsewhere for PQCSV ("pipe-quoted CSV") or whatever.

[satran]

I believe the standard does not specify it. You are right in pointing out that the package is small and easily forkable. Just saw this issue open and thought of fixing it :)

[bradfitz]

The standard specifies it as double quote. It's not a tweakable parameter:

https://tools.ietf.org/html/rfc4180 says

   escaped = DQUOTE *(TEXTDATA / COMMA / CR / LF / 2DQUOTE) DQUOTE
...
   DQUOTE =  %x22 ;as per section 6.1 of RFC 2234 [2]

We'll follow the standard.

People needing wacky formats can use wacky packages.

[clausecker]

CSV is well known for being a format ever program implements somewhat differently, like HTML in the early years of the internet. Yes, the standard says that the quote character is ", but there are many programs out there that expect differently formatted data. Having a CSV package that is flexible in the way it generates output is a very useful thing. I am not sure if suggesting that every program that tries to generate CSV files which do not exactly match the standard shall just fork the encoding/csv package is a good idea both in terms of code-reuse and reliability.

[bradfitz]

Can you provide a few examples of such programs, ideally popular ones?

Absent a compelling reason, I see no reason to introduce complexity for theoretical uses.

Even with examples, I'm tempted to say no just as a minor encouragement to those program's authors and users to do something more normal.