ANSI escape code

Although termcap/terminfo-style libraries were primarily developed on and for Unix, by the mid-1980s programs running on Unix-like operating systems could almost always assume they were using a terminal or emulator that supported ANSI sequences; this led to widespread use of ANSI by programs running on those platforms. For instance, many games and shell scripts, and utilities such as color directory listings, directly write the ANSI sequences and thus cannot be used on a terminal that does not interpret them. Many programs, including text editors such as vi and GNU Emacs, use termcap or terminfo, or use libraries such as curses that use termcap or terminfo, and thus in theory support non-ANSI terminals, but this is so rarely tested nowadays that they are unlikely to work with those terminals.

Terminal emulators for communicating with local programs as well as remote machines and the text system console almost always support ANSI escape codes. This includes terminal emulators such as xterm, rxvt, GNOME Terminal, and Konsole on systems with X11-based or Wayland-based window systems, and Terminal.app and common third-party terminals such as iTerm2 on macOS.

MS-DOS 1.x did not support the ANSI or any other escape sequences. Only a few control characters (BEL, CR, LF, BS) were interpreted by the underlying BIOS, making it almost[lower-alpha 1] impossible to do any kind of full-screen application. Any display effects had to be done with BIOS calls, which were notoriously slow, or by directly manipulating the IBM PC hardware.

DOS 2.0 introduced the ability to add a device driver for the ANSI escape sequences – the de facto standard being ANSI.SYS, but others like ANSI.COM,[6] NANSI.SYS[7] and ANSIPLUS.EXE are used as well (these are considerably faster as they bypass the BIOS). Slowness and the fact that it was not installed by default made software rarely take advantage of it; instead, applications continued to directly manipulate the hardware to get the text display needed. ANSI.SYS and similar drivers continued to work in Windows 9x up to Windows Me, and in NT-derived systems for 16-bit legacy programs executing under the NTVDM.

Many clones of DOS were able to interpret the sequences and do not require a separate ANSI driver to be loaded. PTS-DOS[8][9] as well as Concurrent DOS, Multiuser DOS[10] and REAL/32 have built-in support (plus a number of extensions). OS/2 had an ANSI command that enabled the sequences.

The Windows Console did not support ANSI escape sequences, nor did Microsoft provide any method to enable them. Some replacements or additions for the console window such as JP Software's TCC (formerly 4NT), Michael J. Mefford's ANSI.COM, Jason Hood's ANSICON[11] and Maximus5's ConEmu interpreted ANSI escape sequences printed by programs. A Python package[12] internally interpreted ANSI escape sequences in text being printed, translating them to calls to manipulate the color and cursor position, to make it easier to port Python code using ANSI to Windows. Cygwin performs similar translation to all output written to the console using Cygwin file descriptors, the filtering is done by the output functions of cygwin1.dll, to allow porting of POSIX C code to Windows.

In 2016, Microsoft released the Windows 10 version 1511 update which unexpectedly implemented support for ANSI escape sequences, over two decades after the debut of Windows NT.[13] The change was designed to complement the Windows Subsystem for Linux, adding to the Windows Console Host used by Command Prompt support for character escape codes used by terminal-based software for Unix-like systems. This is not the default behavior and must be enabled programmatically with the Win32 API via SetConsoleMode(handle, ENABLE_VIRTUAL_TERMINAL_PROCESSING).[14] This was enabled by CMD.EXE but not initially by PowerShell;[15] however, Windows PowerShell 5.1 now enables this by default. Starting with PowerShell 6, it is possible to embed a ESC character into a string via the `e special sequence; For example, "`e[32m".[16] Earlier versions have to use the [char]27) object; for example, "$([char]27)[32m".

Windows Terminal, introduced in 2019, supports the sequences by default, and it appears Microsoft intends to merge or replace Windows Console with it.

The Atari ST used the command system adapted from the VT52 with some expansions for color support,[17] rather than supporting ANSI escape codes.

AmigaOS not only interprets ANSI code sequences for text output to the screen, the AmigaOS printer driver also interprets them (with extensions proprietary to AmigaOS) and translates them into the codes required for the particular printer that is actually attached.[18]

VMS was designed to be managed interactively using Digital's text-based video terminals such as the aforementioned VT100; later with graphical terminal emulators such as the VWS Terminal, DECTerm, and xterm.[19]

Though not technically part of the standard, almost all users assume some functions of some single-byte characters. This is used to shorten the amount of data transmitted, or to perform some functions that are unavailable from escape sequences:

Escape sequences vary in length. The general format for an ANSI-compliant escape sequence is defined by ANSI X3.41 (equivalent to ECMA-35 or ISO/IEC 2022).[20]^:13.1 The escape sequences consist only of bytes in the range 0x20—0x7F (all the non-control ASCII characters), and can be parsed without looking ahead. Behavor when a control character, a byte with the high bit set, or a byte that is not part of any valid sequence, is encountered before the end is undefined.

If the ESC is followed by a byte in the range 0x40 to 0x5F it is of this type. Delegated to the applicable C1 control code standard.[20]^:13.2.1 Accordingly, all escape sequences corresponding to C1 control codes from ANSI X3.64 / ECMA-48 follow this format.[5]^:5.3.a

The standard says that, in 8-bit environments, the control functions corresponding to type Fe escape sequences (those from the set of C1 control codes) can be represented as single bytes in the 0x80–0x9F range.[5]^:5.3.b However, on modern devices those codes are often used for other purposes, such as parts of UTF-8 or for CP-1252 characters, so only the 2-byte sequence is typically used. In the case of UTF-8 a C1 control code can be encoded as two bytes (e.g. 0xC2,0x8E for U+008E), but no space is saved this way.

For Control Sequence Introducer, or CSI, commands, the ESC [ is followed by any number (including none) of "parameter bytes" in the range 0x30–0x3F (ASCII 0–9:;<=>?), then by any number of "intermediate bytes" in the range 0x20–0x2F (ASCII space and !"#$%&'()*+,-./), then finally by a single "final byte" in the range 0x40–0x7E (ASCII @A–Z[\]^_`a–z{|}~).[5]^:5.4

All common sequences just use the parameters as a series of semicolon-separated numbers such as 1;2;3. Missing numbers are treated as 0 (1;;3 acts like the middle number is 0, and no parameters at all in ESC[m acts like a 0 reset code). Some sequences (such as CUU) treat 0 as 1 in order to make missing parameters useful.[5]^:F.4.2

A subset of arrangements was declared "private" so that terminal manufacturers could insert their own sequences without conflicting with the standard. Sequences containing the parameter bytes <=>? or the final bytes 0x70–0x7E (p–z{|}~) are private.

The behavior of the terminal is undefined in the case where a CSI sequence contains any character outside of the range 0x20–0x7E. These illegal characters are either C0 control characters (the range 0–0x1F), DEL (0x7F), or bytes with the high bit set. Possible responses are to ignore the byte, to process it immediately, and furthermore whether to continue with the CSI sequence, to abort it immediately, or to ignore the rest of it.

Select Graphic Rendition (SGR) sets display attributes. Several attributes can be set in the same sequence, separated by semicolons.[27] Each display attribute remains in effect until a following occurrence of SGR resets it.[5] If no codes are given, CSI m is treated as CSI 0 m (reset / normal).

In ECMA-48 SGR is called "Select Graphic Rendition".[5]^:8.3.117 In Linux manual pages the term "Set Graphics Rendition" is used.[27]

The original specification only had 8 colors, and just gave them names. The SGR parameters 30–37 selected the foreground color, while 40–47 selected the background. Quite a few terminals implemented "bold" (SGR code 1) as a brighter color rather than a different font, thus providing 8 additional foreground colors. Usually you could not get these as background colors, though sometimes inverse video (SGR code 7) would allow that. Examples: to get black letters on white background use ESC[30;47m, to get red use ESC[31m, to get bright red use ESC[1;31m. To reset colors to their defaults, use ESC[39;49m (not supported on some terminals), or reset all attributes with ESC[0m. Later terminals added the ability to directly specify the "bright" colors with 90–97 and 100–107.

When hardware started using 8-bit digital-to-analog converters (DACs) several pieces of software assigned 24-bit color numbers to these names. The chart below shows values sent to the DAC for some common hardware and software.

As 256-color lookup tables became common on graphic cards, escape sequences were added to select from a pre-defined set of 256 colors:

ESC[ 38;5;⟨n⟩ m Select foreground color
ESC[ 48;5;⟨n⟩ m Select background color
  0-  7:  standard colors (as in ESC [ 30–37 m)
  8- 15:  high intensity colors (as in ESC [ 90–97 m)
 16-231:  6 × 6 × 6 cube (216 colors): 16 + 36 × r + 6 × g + b (0 ≤ r, g, b ≤ 5)
232-255:  grayscale from black to white in 24 steps

The ITU's T.416 Information technology - Open Document Architecture (ODA) and interchange format: Character content architectures[37] uses ':' as separator characters instead:

ESC[ 38:5:⟨n⟩ m Select foreground color
ESC[ 48:5:⟨n⟩ m Select background color

As "true color" graphic cards with 16 to 24 bits of color became common, Xterm,[21] KDE's Konsole,[38] iTerm, as well as all libvte based terminals[39] (including GNOME Terminal) support 24-bit foreground and background color setting.[40]

ESC[ 38;2;⟨r⟩;⟨g⟩;⟨b⟩ m Select RGB foreground color
ESC[ 48;2;⟨r⟩;⟨g⟩;⟨b⟩ m Select RGB background color

The ITU's T.416 Information technology - Open Document Architecture (ODA) and interchange format: Character content architectures[37] which was adopted as ISO/IEC International Standard 8613-6 gives an alternative version that seems to be less supported. The parameters after the '2', i.e. even the r,g,b are optional. Note that this is not just the above sequence with semicolon replaced by colon, there is a leading "colorspace ID" (this fact was missed by many terminal emulators, this omission seems to have come from KDE Konsole).[21] The definition of the colorspace ID is not included in that document so it may be blank to represent the unspecified default. As well as the '2' value to specify a Red-Green-Blue format (and the '5' above for a 0-255 indexed color) there are alternatives of '0' for implementation defined and '1' for transparent - neither of what have any further parameters; '3' specifies colors using a Cyan-Magenta-Yellow scheme, and '4' for a Cyan-Magenta-Yellow-Black one, the latter using the position marked as "unused" for the Black component:

ESC[ 38:2:⟨Color-Space-ID⟩:⟨r⟩:⟨g⟩:⟨b⟩:⟨unused⟩:⟨CS tolerance⟩:⟨Color-Space associated with tolerance: 0 for "CIELUV"; 1 for "CIELAB"⟩; m Select RGB foreground color
ESC[ 48:2:⟨Color-Space-ID⟩:⟨r⟩:⟨g⟩:⟨b⟩:⟨unused⟩:⟨CS tolerance⟩:⟨Color-Space associated with tolerance: 0 for "CIELUV"; 1 for "CIELAB"⟩; m Select RGB background color

Mostly these were defined by Xterm. For historical reasons xterm can end the command with BEL as well as ST.[21]

Xterm allows the window title to be set by ESC ]0;this is the window title BEL.

A non-xterm extension is the hyperlink, ESC ]8;link ST from 2017, used by VTE and iTerm2.[41][42]

The Linux console uses ESC ] P n rr gg bb to change the palette, which, if hard-coded into an application, may hang other terminals.[43] However, appending ST will be ignored by Linux and form a proper, ignorable sequence for other terminals.

If the ESC is followed by a byte in the range 0x60—0x7E it is of this type. Used for control functions individually registered with the ISO-IR registry and, consequently, available even in contexts where a different C1 control code set is used. Specifically, they correspond to single control functions approved by ISO/IEC JTC 1/SC 2 and standardized by ISO or an ISO-recognised body.[20]^:6.5.1 Some of these are specified in ECMA-35 (ISO 2022 / ANSI X3.41), others in ECMA-48 (ISO 6429 / ANSI X3.64).[20]^:6.5.4 ECMA-48 refers to these as "independent control functions".[5]^:5.5

If the ESC is followed by a byte in the range 0x30—0x3F it is of this type. Set apart for private-use control functions.[20]^:6.5.3

If the ESC is followed by a byte in the range 0x20—0x2F it is of this type. Followed by any number of additional bytes in this range, and then a byte in the range 0x30-0x7E. They are further subcategorised by the low four bits of the first byte, e.g. "type 2F" for sequences where the first byte is 0x22; and by whether the final byte is in the private use range 0x30—0x3F (e.g. "type 2Fp") or not (e.g. "type 2Ft").[20]^:13.2.1 They are mostly used for ANSI/ISO code-switching mechanisms such as those used by ISO-2022-JP, except for type 3F sequences (those where the first intermediate byte is 0x23), which are used for individual control functions. Type 3Ft sequences are reserved for additional ISO-IR registered individual control functions,[20]^:6.5.2 while type 3Fp sequences are available for private-use control functions.[20]^:6.5.3

CSI 2 J — This clears the screen and, on some devices, locates the cursor to the y,x position 1,1 (upper left corner).

CSI 32 m — This makes text green. The green may be a dark, dull green, so you may wish to enable Bold with the sequence CSI 1 m which would make it bright green, or combined as CSI 32 ; 1 m. Some implementations use the Bold state to make the character Bright.

CSI 0 ; 6 8 ; "DIR" ; 13 p — This reassigns the key F10 to send to the keyboard buffer the string "DIR" and ENTER, which in the DOS command line would display the contents of the current directory. (MS-DOS ANSI.SYS only) This was sometimes used for ANSI bombs. This is a private-use code (as indicated by the letter p), using a non-standard extension to include a string-valued parameter. Following the letter of the standard would consider the sequence to end at the letter D.

CSI s — This saves the cursor position. Using the sequence CSI u will restore it to the position. Say the current cursor position is 7(y) and 10(x). The sequence CSI s will save those two numbers. Now you can move to a different cursor position, such as 20(y) and 3(x), using the sequence CSI 20 ; 3 H or CSI 20 ; 3 f. Now if you use the sequence CSI u the cursor position will return to 7(y) and 10(x). Some terminals require the DEC sequences ESC 7 / ESC 8 instead which is more widely supported.

ANSI escape codes are often used in UNIX and UNIX-like terminals to provide syntax highlighting. For example, on compatible terminals, the following list command color-codes file and directory names by type.

ls --color

Users can employ escape codes in their scripts by including them as part of standard output or standard error. For example, the following GNU sed command embellishes the output of the make command by displaying lines containing words starting with "WARN" in reverse video and words starting with "ERR" in bright yellow on a dark red background (letter case is ignored). The representations of the codes are highlighted.[51]

make 2>&1 | sed -e 's/.*\bWARN.*/\x1b[7m&\x1b[0m/i' -e 's/.*\bERR.*/\x1b[93;41m&\x1b[0m/i'

The following Bash function flashes the terminal (by alternately sending reverse and normal video mode codes) until the user presses a key.[52]

flasher () { while true; do printf \\e[?5h; sleep 0.1; printf \\e[?5l; read -s -n1 -t1 && break; done; }

This can be used to alert a programmer when a lengthy command terminates, such as with make ; flasher .[53]

printf \\033c

This will reset the console, similar to the command reset on modern Linux systems; however it should work even on older Linux systems and on other (non-Linux) UNIX variants.

Pressing special keys on the keyboard, as well as outputting many xterm CSI, DCS, or OSC sequences, often produces a CSI, DCS, or OSC sequence, sent from the terminal to the computer as though the user typed it.

When typing input on a terminal keypresses outside the normal main alphanumeric keyboard area can be sent to the host as ANSI sequences. For keys that have an equivalent output function, such as the cursor keys, these often mirror the output sequences. However, for most keypresses there isn't an equivalent output sequence to use.

There are several encoding schemes, and unfortunately most terminals mix sequences from different schemes, so host software has to be able to deal with input sequences using any scheme. To complicate the matter, the VT terminals themselves have two schemes of input, normal mode and application mode that can be switched by the application.

(draft section)

<char>                                -> char
<esc> <nochar>                        -> esc
<esc> <esc>                           -> esc
<esc> <char>                          -> Alt-keypress or keycode sequence
<esc> '[' <nochar>                    -> Alt-[
<esc> '[' (<num>) (';'<num>) '~'      -> keycode sequence, <num> defaults to 1

If the terminating character is '~', the first number must be present and is a keycode number, the second number is an optional modifier value. If the terminating character is a letter, the letter is the keycode value, and the optional number is the modifier value.

The modifier value defaults to 1, and after subtracting 1 is a bitmap of modifier keys being pressed: Meta-Ctrl-Alt-Shift. So, for example, <esc>[4;2~ is Shift-End, <esc>[20~ is function key 9, <esc>[5C is Ctrl-Right.

vt sequences:
<esc>[1~    - Home        <esc>[16~   -             <esc>[31~   - F17
<esc>[2~    - Insert      <esc>[17~   - F6          <esc>[32~   - F18
<esc>[3~    - Delete      <esc>[18~   - F7          <esc>[33~   - F19
<esc>[4~    - End         <esc>[19~   - F8          <esc>[34~   - F20
<esc>[5~    - PgUp        <esc>[20~   - F9          <esc>[35~   - 
<esc>[6~    - PgDn        <esc>[21~   - F10         
<esc>[7~    - Home        <esc>[22~   -             
<esc>[8~    - End         <esc>[23~   - F11         
<esc>[9~    -             <esc>[24~   - F12         
<esc>[10~   - F0          <esc>[25~   - F13         
<esc>[11~   - F1          <esc>[26~   - F14         
<esc>[12~   - F2          <esc>[27~   -             
<esc>[13~   - F3          <esc>[28~   - F15         
<esc>[14~   - F4          <esc>[29~   - F16         
<esc>[15~   - F5          <esc>[30~   -             

xterm sequences:
<esc>[A     - Up          <esc>[K     -             <esc>[U     -
<esc>[B     - Down        <esc>[L     -             <esc>[V     -
<esc>[C     - Right       <esc>[M     -             <esc>[W     -
<esc>[D     - Left        <esc>[N     -             <esc>[X     -
<esc>[E     -             <esc>[O     -             <esc>[Y     -
<esc>[F     - End         <esc>[1P    - F1          <esc>[Z     -
<esc>[G     - Keypad 5    <esc>[1Q    - F2       
<esc>[H     - Home        <esc>[1R    - F3       
<esc>[I     -             <esc>[1S    - F4       
<esc>[J     -             <esc>[T     - 

<esc>[A to <esc>[D are the same as the ANSI output sequences. The <num> is normally omitted if no modifier keys are pressed, but most implementations always emit the <num> for F1-F4. (draft section)

Xterm has a comprehensive documentation page on the various function-key and mouse input sequence schemes from DEC's VT terminals and various other terminals it emulates.[21] Thomas Dickey has added a lot of support to it over time;[54] he also maintains a list of default keys used by other terminal emulators for comparison.[55]

• On the Linux console, certain function keys generate sequences of the form CSI [ char. The CSI sequence should terminate on the [.
• Old versions of Terminator generate SS3 1; modifiers char when F1–F4 are pressed with modifiers. The faulty behavior was copied from GNOME Terminal.
• xterm replies CSI row ; column R if asked for cursor position and CSI 1 ; modifiers R if the F3 key is pressed with modifiers, which collide in the case of row == 1. This can be avoided by using the ? private modifier as CSI ? 6 n, which will be reflected in the response as CSI ? row ; column R.
• many terminals prepend ESC to any character that is typed with the alt key down. This creates ambiguity for uppercase letters and symbols @[\]^_, which would form C1 codes.
• Konsole generates SS3 modifiers char when F1–F4 are pressed with modifiers.