CSV Spec 0.9.0-draft.2


CSV is not a file format, it is a loose set of guidelines of how to structure tabular data into a plain text string. As such there’s an endless amount of *.csv files floating around which are highly incompatible with each other. The closest thing there is to a specification is RFC 4180.


This project is an attempt to summarize RFC 4180 and the information in the Comma-separated values (CSV) Wikipedia article into a easy to understand format. The spec will also take into account that the comma (,) character is not the only character used as a field delimiter. Semi-colons (;), tabs (\t), and more are popular field delimiter characters. As such the specification will more accurately be describing a CSV-like structured data format.

We will also provide input/output test files that CSV parser/writer software libraries can use to validate if they properly adhere to the rules laid out in this specification. And if possible we will even try to provide code snippets in various languages that attempts to automatically determine the delimiter character used in any given input CSV-like formatted file/data.


  1. Write up core specification rules. [in-progress]
  2. ~Create website for csv-spec.org.~ [done]
  3. Create input/output test files covering all rules in the specification.
  4. Create linting tool as a NPM module, allowing easy validation of CSV data both client-side in a web browser, and server side via a command line tool.
  5. Create automatic delimiter character detection code snippets in various programming languages which CSV parser developers can freely use to enhance their libraries.


  • Field — A singular String value within a record.
  • Record (or Row) — A collection of fields. This is often referred to as a “line”, but a single record can span multiple text lines if a field within it contains one or more line breaks.
  • Delimiter — The character used to separate fields withing a row. Commonly this will be a comma (,), but semi-colons (;) or tabs (\t) are two other popular delimiter characters.
  • Header — The first row is often used to contain the column names for all remaining rows. Header names would be used as key names when CSV data is converted to JSON for example.
  • Line Break — Line breaks in CSV files can be CRLF (\r\n), LF (\n), and even in rare cases CR (\r).
  • LF, CR, and CRLF — Different types of line breaks, typically determined by the OS. Linux, macOS, and other *NIX operating systems generally use a line feed (LF or \n) character. Windows uses a carriage return (CR or \r) and a line feed character, effectively “CRLF” (\r\n).

CSV Format Specification

The key words “MUST”, “MUST NOT”, “REQUIRED”, “SHALL”, “SHALL NOT”, “SHOULD”, “SHOULD NOT”, “RECOMMENDED”, “MAY”, and “OPTIONAL” in this document are to be interpreted as described in RFC 2119.

These rules are mostly based on the corresponding section from RFC 4180, with minor changes, clarifications and improved examples.

  1. Each record starts at the beginning of its own line, and ends with a line break (shown as ¬).




    [ ["aaa", "bbb", "ccc"],
      ["xxx", "yyy", "zzz"] ]
  2. Though it is RECOMMENDED, the last record in a file is not required to have a ending line break.




    [ ["aaa", "bbb", "ccc"],
      ["xxx", "yyy", "zzz"] ]
  3. There may be an OPTIONAL header line appearing as the first line of the file with the same format as normal records. This header will contain names corresponding to the fields in the file, and MUST contain the same number of fields as the records in the rest of the file.



    JSON (ignoring headers):

    [ ["field_1", "field_2", "field_3"],
      ["aaa", "bbb", "ccc"],
      ["xxx", "yyy", "zzz"] ]

    JSON (using headers):

    [ {"field_1": "aaa", "field_2": "bbb", "field_3": "ccc"},
      {"field_1": "xxx", "field_2": "yyy", "field_3": "zzz"} ]
  4. Within each record and the OPTIONAL header, there may be one or more fields, separated by a delimiter (normally a comma). Each record MUST contain the same number of fields throughout the file.

    CSV (invalid):

  5. The last field in a record MUST NOT be followed by a comma. This results in a additional field with nothing in it.




    [ ["aaa", "bbb", "ccc", ""],
      ["xxx", "yyy", "zzz", ""] ]
  6. Spaces are considered part of a field and MUST NOT be ignored.


    aaa ,  bbb , ccc¬
     xxx, yyy  ,zzz ¬


    [ ["aaa ", "  bbb ", " ccc"],
      [" xxx", " yyy  ", "zzz "] ]
  7. Fields containing line breaks (CRLF, LF, or CR), double quotes, or the delimiter character (normally a comma) MUST be enclosed in double-quotes.


    xxx,"y, yy",zzz¬


    [ ["aaa", "b\r\nbb", "ccc"],
      ["xxx", "y, yy", "zzz"] ]
  8. A double-quote appearing inside a field MUST be escaped by preceding it with another double quote, and the field itself MUST be enclosed in double quotes.




    [ ["aaa", "b\"bb", "ccc"] ]
  9. When a field enclosed in double quotes has spaces before and/or after the double quotes, the spaces MUST be ignored, as the field starts and ends with the double quotes. However this is considered invalid formatting and the CSV parser SHOULD report some form of warning message.


    xxx,  "y, yy" ,zzz¬


    [ ["aaa", "bbb", "ccc"],
      ["xxx", "y, yy", "zzz"] ]
  10. It is possible to enclose every field in double quotes even if they don’t need to be enclosed. However it is RECOMMENDED to only enclose fields in double quotes that requires it.




    [ ["aaa", "bbb", "ccc"],
      ["xxx", "yyy", "zzz"] ]
  11. All fields are always strings. CSV itself does not support type casting to integers, floats, booleans, or anything else. It is not a CSV library’s responsibility to type cast input CSV data.

    If type casting is required, it is up to the developer using a specific CSV library to ensure types are correctly dealt with.

    Input JSON:

    [ [10, true, 0.3, null, "aaa"],
      [11, false, 2.13, "", "bbb"] ]

    Output CSV:


    Output CSV parsed back to JSON:

    [ ["10", "true", "0.3", "", "aaa"],
      ["11", "false", "2.13", "", "bbb"] ]

    At this point it is up to the developer themselves to type cast the above output data from the CSV parser.

  12. However, when rendering type cast input data to CSV text, non-string types MUST be converted to a string in such a way that minimal information is lost.
    • Integers and floats MUST be rendered as a string version of themselves.
    • Booleans true and false MUST be rendered as true and false strings, not as 1 or 0 numbers. If numbers are used the resulting CSV data is indistinguishable from actual integer numbers.
    • Null/nil values MUST be rendered as empty strings.
  13. When parsing input CSV data all forms of line breaks (CRLF, LF, and CR) MUST be supported.

  14. When rendering output CSV data, CRLF MUST be used for line breaks to ensure maximum cross-platform compatibility.


This CSV specification is authored by Jim Myhrberg.

If you’d like to leave feedback, please open an issue on GitHub.


CC0 1.0 Universal