.TSV File Format – Rules and conventions

Tab-separated values

TSV’s only formal definition is the IANA registration of the media type text/tab-separated-values, whose contact is the University of Minnesota Internet Gopher Team (registrant Paul Lindner); the registration carries no date. This page follows the IANA registration, the PostgreSQL COPY documentation, the MySQL LOAD DATA documentation, the Python csv module documentation and the W3C Model for Tabular Data and Metadata on the Web, read 2026-11-12.

The registered rules

  • Each record is a single line; fields are separated by a tab character.
  • The first line is special: it holds the name of each field, separated by tabs.
  • Each record must have the same number of fields.
  • Fields that contain tabs are “not allowable”. The registration defines no quoting or escaping.
  • Required parameters are given as “Character Set, Encoding Type”; optional parameters as “None”.

The registration’s own grammar, as written there:

field    ::= [character]+           # multiple characters
name     ::= [character]+
nameline ::= name [TAB name]+ EOL
record   ::= field [TAB field]+ EOL   # at least one field, or more
tsv      ::= nameline record+

Read literally, [TAB name]+ needs at least two fields, although the comment says “at least one field”. The registration does not define EOL or character.

Conventions in practice

ConventionUsed byRule
CSV-style quotingPython’s csv module, whose excel-tab dialect “defines the usual properties of an Excel-generated TAB-delimited file”With the default QUOTE_MINIMAL, fields containing the delimiter, the quote character, \r or \n are quoted, and quotes inside are doubled, as in CSV.
Backslash escapesPostgreSQL COPY text format (tab is its default delimiter)Backslash, newline, carriage return and the delimiter must be preceded by a backslash in a value. COPY FROM reads \b, \f, \n, \r, \t, \v, octal \digits and hex \xdigits; null is \N by default; a line holding only \. marks end of data.
Backslash escapesMySQL LOAD DATA (defaults FIELDS TERMINATED BY '\t' ENCLOSED BY '' ESCAPED BY '\\' LINES TERMINATED BY '\n')\0 NUL, \b backspace, \n line feed, \r carriage return, \t tab, \Z Control+Z, \N null.

PostgreSQL’s COPY TO ends rows with LF (CR LF only for a server file on Windows), and COPY FROM accepts LF, CR or CR LF but complains if they are mixed. The W3C model records that data copied from Excel for Mac 2011 is tab-delimited with LF line breaks.