Unexpected quotations when dumping

[lbauret]

When dumping the following JSON, it seems there are some specific situations where strings are not held between quotation marks. For example when a certain string starts with 0, or "08" and "09" strings. Also, strings are not properly indented when containing \t or \r.

Input:

{
  "start-with-0" : {
    "fails": "015678",
    "ok": "115678"
  },
  "number-list" : [
    "01",
    "02",
    "03",
    "04",
    "05",
    "06",
    "07",
    "08",
    "09",
    "10",
    "11"
  ],
  "white-spaces": {
    "string1": "**Hello World**\n\n**Sample Text.** \t End of text.",
    "string2": "**Hello World**\n\n**Sample Text.** \r End of text.",
    "string3": "**Hello World**\n\n**Sample Text.**  End of text.",
  }
}

Output:

start-with-0:
  fails: 015678
  ok: '115678'
number-list:
  - '01'
  - '02'
  - '03'
  - '04'
  - '05'
  - '06'
  - '07'
  - 08
  - 09
  - '10'
  - '11'
white-spaces:
  string1: "**Hello World**\n\n**Sample Text.** \t End of text."
  string2: "**Hello World**\n\n**Sample Text.** \r End of text."
  string3: |-
    **Hello World**

    **Sample Text.**  End of text.

[puzrin]

Did you tried online demo with your sample?

Dumper keeps minimal possible quoting, and tries to beautify strings when possible for readability. As far as i see, loading dumed yaml (from your message) is equal to original. Did i missed something?

[aepsilon]

As the author of the current scalar dumper, I hope this helps:

• Numbers beginning with 0 are octals under YAML 1.1. By default the dumper tries to remain backward-compatible, so it enquotes strings that could be misinterpreted. In your example the strings that aren't in quotes are exactly the ones that aren't valid octals: 015678, 08, and 09 contain the digits 8 and 9. Cf. http://yaml.org/type/int.html. It looks like Ruby's YAML module does something similar: https://stackoverflow.com/questions/33097353.

• Escaping tabs was a simplification to avoid potential problems like issue 80. It may have been overly cautious though. I could try supporting literal tabs if there's demand. (For future reference, some changes would be needed in block indent indicators and line folding.)

• The primary goal of the dumper is to preserve content. To preserve a carriage return \r in YAML, it has to be escaped, which necessitates double-quoted style. Otherwise it will be normalized into a \n the next time it's parsed.

  Line breaks inside scalar content must be normalized by the YAML processor. Each such line break must be parsed into a single line feed character.
  http://yaml.org/spec/1.2/spec.html#id2774608

  You could use something like stringvalue.replace(/\r\n?/g, '\n') to normalize each string first if you would rather have a nicely-formatted output and don't mind losing the distinction between \r, \r\n, and \n.

[lbauret]

Thank you very much for the quick response.
About the octal numbers, you said:

By default the dumper tries to remain backward-compatible, so it enquotes strings that could be misinterpreted.

Is there a way to avoid this default behavior then? If there is not, would you consider this to be a bug?

[puzrin]

I don't see enougth added value to change current behaviour.