user_caps(5)
user_caps(5) โ File formats ๐ฆ
okay bestie, this is the manual page for the user-defined terminfo capability format, and it's giving deep Unix lore with a side of โwhy is the binary format like this ๐ญโ. Buckle up.
โจ NAME โจ
user_caps โ user-defined terminfo capability format ๐
๐ SYNOPSIS ๐
infocmp -x
tic -x
๐ DESCRIPTION ๐
๐ง Background ๐ง
Prior to ncurses 5.0 (1999), terminfo databases used a fixed repertoire of terminal capabilities designed for the SVr2 terminal database in 1984, added to in stages through SVr4 (1989), and standardized in X/Open Curses starting in 1995.
Most such additions to that fixed repertoire supplemented the tables of Boolean, numeric, and string capabilities. Rather than changing the meaning of an existing capability, a new name was added. The terminfo database uses a binary format; binary compatibility was ensured by using a header that counted the number of items in the tables for each type of capability. Because each curses vendor extended the standard capability lists in distinct ways, a library could be programmed to recognize only compiled terminfo entries that it was prepared to interpret. Standardization was incomplete.
- X/Open Curses describes only the source format, not its binary representation on disk.
- Library developers rely upon
SVr4documentation and reverse engineering of compiled terminfo files to match the binary format.
- Library developers rely upon
- Lacking a standard for the binary format, most implementations copy the
SVr2binary format, which uses 16-bit signed integers, and is limited to 4096-byte entries.- The
SVr2format cannot represent very large numeric capability values, nor can it represent large numbers of key definitions, as are required to distinguish multiple modifier keys used in combination with a function key.
- The
- The tables of capability names differ between implementations.
- Although they may provide all of the standard capability names, each arranges its table entries differently because some features were added as needed, while others were added โ out of order โ for X/Open Curses conformance.
- While
ncurses's capability repertoire is closest to that of Solaris, the set of capabilities supported by each vendor's terminfo database differs from the list published by X/Open Curses.ncursescan be configured with tables that match the terminal databases for AIX, HP-UX, or OSF/1, rather than the default Solaris-like configuration.
- In
SVr4curses andncurses, the terminal database is defined at compile time by interpolating a text file that lists the different terminal capabilities.- In principle, the text file can be extended, but doing so requires recompiling and reinstalling the library. The text file used by
ncursesfor terminal capabilities includes details of extensions to X/Open Curses made by various systems. For example,ncursessupports the following nonstandard capabilities in each configuration.memory_lock(meml) โ lock memory above cursormemory_unlock(memu) โ unlock memorybox_chars_1(box1) โ box characters primary set
- The memory lock/unlock capabilities were included because they were used in the X11R6 terminal description for
xterm(1).ticuses thebox1capability to cope with terminal descriptions written for AIX.
- In principle, the text file can be extended, but doing so requires recompiling and reinstalling the library. The text file used by
During the 1990s, some application developers were reluctant to use terminfo in spite of its performance (and other) advantages over termcap.
- The fixed repertoire prevented users from adding support for terminal features unanticipated by X/Open Curses (or required them to reuse existing capabilities as a workaround).
- The limitation to 16-bit signed integers was also mentioned. Because termcap stores everything as a string, it could represent larger numbers.
Although termcap's extensibility was rarely used โ the claimant was never an implementor who had actually exercised it โ the criticism had a point. ncurses 5.0 provided a way to detect nonstandard capabilities, to determine their type, and to optionally store and retrieve them in a way that did not interfere with other applications. ncurses terms these user-defined capabilities because no modifications to the standard capability list are needed.
The ncurses utilities tic and infocmp have a command-line option -x to control whether the nonstandard capabilities are stored or retrieved. ncurses provides use_extended_names(3X) to programs for the same purpose.
When compiling a terminal database, if -x is used, tic stores a user-defined capability if the capability name is not standard.
Because ncurses provides a termcap library interface, these user-defined capabilities may be visible to termcap applications.
- The termcap interface (like all implementations of termcap) restricts capability names to two characters.
- When the capability is simple enough for use in a termcap application, it is provided as a two-character name.
- Other user-defined capabilities employ features not usable in termcap, such as parameterized strings that use more than two parameters or require more powerful expressions than termcap supports. Such capabilities should, in the terminfo database, have names at least three characters in length.
- Some terminals can send distinct strings for special keys (cursor-, keypad- or function-keys) depending on modifier keys (shift, control, etc.). While terminfo and termcap define a set of sixty function key names, to which a series of keys can be assigned, that is insufficient for more than a dozen keys multiplied by more than a couple of modifier combinations. The
ncursesdatabase uses a convention based onxterm(1)to provide extended special-key names.- Fitting that into termcap's limitation of 2-character names would be pointless. These extended keys are available only with terminfo.
๐ Recognized Capabilities ๐
The ncurses library employs user-definable capabilities. While the terminfo database may have other extensions, ncurses makes explicit checks for the following.
AX(Boolean) โ asserts that the terminal interpretsSGR 39andSGR 49by resetting the foreground and background colors, respectively, to the default.screen(1)recognizes this capability as well.E3(string) โ tells an application how to clear the terminal's scrollback buffer. When present, theclear(1)program sends this before clearing the terminal. The commandtput cleardoes the same thing.NQ(Boolean) โ suppresses a consistency check inticfor thencursesstring capabilitiesuser6(u6) throughuser9(u9), which tell an application how to query the terminal's cursor position and its device attributes.RGB(Boolean, numeric, or string) โ asserts that theset_a_foreground(setaf) andset_a_background(setab) capabilities employ direct colors, using an RGB (red/green/blue) convention. This capability allowscolor_content(3X)to return appropriate values without requiring the application to initialize colors usinginit_color(3X). The capability type determines the valuesncursessees.- Boolean โ implies that the number of bits for red, green, and blue are the same. Starting with the value of the capability
max_colors(colors; termcap:co),ncursesadds two, divides the sum by three, and assigns the result to red, green, and blue, in that order. If the number of bits needed for the number of colors is not a multiple of three, the blue (and green) color channels lose in comparison to red. - numeric โ tells
ncurseswhat result to add to red, green, and blue. Ifncursesruns out of bits, blue (and green) lose just as in the Boolean case. - string โ specify the quantity of bits used for red, green, and blue color channels as a slash-separated list of decimal integers.
- Because there are several RGB encodings in use, applications that make assumptions about the number of bits per color channel are unlikely to work reliably. As a trivial case, one could define
RGB#1to represent the standard eight ANSI X3.64/ECMA-48/ISO 6429 colors using one bit per color channel.
- Boolean โ implies that the number of bits for red, green, and blue are the same. Starting with the value of the capability
U8(numeric) โ asserts whetherncursesmust use Unicode values for line-drawing characters, and that it should ignore the alternate character set (ACS) capabilities when the locale uses UTF-8 encoding. See the discussion ofNCURSES_NO_UTF8_ACSin section โENVIRONMENTโ ofncurses(3X). Set this capability to a nonzero value to enable it.XM(string) โ overridencurses's built-in string that directsxterm(1)to enable or disable mouse mode.ncursessends a character sequence to the terminal to initialize mouse mode, and when the user clicks the mouse buttons or (in certain modes) moves the mouse, handles the characters sent back by the terminal to tell the application what was done with the mouse.- The mouse protocol is enabled when the mask argument to the
mousemask(3X)function is nonzero. By default,ncurseshandles the responses for the X11 xterm mouse protocol. It also knows about the SGR 1006 xterm mouse protocol, but has to be told to look for it specifically.ncursescan't guess which of the two modes is used, because the responses are enough alike that only confusion would result. - The
XMcapability has a single numeric parameter. If nonzero, the mouse protocol should be enabled. If zero, the mouse protocol should be disabled.ncursesinspects this capability if it is present, to see whether the 1006 protocol is used. If so, it expects the responses to use the SGR 1006 xterm mouse protocol. - The xterm mouse protocol is used by other terminal emulators. The terminal database uses building blocks for the various xterm mouse protocols usable in customized terminal descriptions.
- The terminal database building blocks for this mouse feature also have an experimental capability,
xm, that describes the mouse response. No known interpreter uses this information, which could make mouse support completely data-driven. xmshows the format of the mouse responses. In this experimental capability, the parameters are as follows.p1โ y-ordinatep2โ x-ordinatep3โ buttonp4โ state, e.g., pressed or releasedp5โ y-ordinate starting region
p6โ x-ordinate starting regionp7โ y-ordinate ending regionp8โ x-ordinate ending region
Okay so real talk, here's what these look like in the actual terminal database for the mouse protocols everyone actually uses ๐ this is the good stuff:
xterm+x11mouse|X11 xterm mouse protocol,
kmous=\E[M, XM=\E[?1000%?%p1%{1}%=%th%el%;,
xm=\E[M
%?%p4%t%p3%e%{3}%;%' '%+%c
%p2%'!'%+%c
%p1%'!'%+%c,
xterm+sm+1006|xterm SGR-mouse,
kmous=\E[<, XM=\E[?1006;1000%?%p1%{1}%=%th%el%;,
xm=\E[<%i%p3%d;
%p1%d;
%p2%d;
%?%p4%tM%em%;,
Some of these extended capabilities have parameters that ncurses might pass into tparm, tiparm, or tiparm_s. ๐ซถ And like, the ncurses library checks โ it makes sure both standard and extended capabilities have the expected number and type of parameters. No sneaky mismatches allowed, bestie. The terminfo compiler tic does similar checks for these extensions, plus for the ones other programs use, which include mintty, screen, tack, tmux, vim, and xterm. Iconic crossover episode honestly โจ
๐น Extended Key Definitions โจ
Several terminals can send you distinct strings for combos of modified special keys. And there'sโฆ no standard at all for what those keys can send. ๐ Literally none. Wild.
Since 1999, xterm(1) has supported shift, control, alt, and meta modifiers, and those produce distinct special-key strings. But here's the thing โ in a terminal description, ncurses has no special knowledge of the modifiers being used. So applications can use the naming convention that was established for xterm to find these special keys in the terminal description. iykyk ๐ง
Starting with the curses convention that capability codes describing the input generated by a terminal's key caps begin with โkโ, and that shifted special keys use uppercase letters in their names, ncurses's terminal database defines the following names and codes, to which a suffix gets added:
kDCโ shiftedkdch1(delete character)kDNโ shiftedkcud1(cursor down)kENDโ shiftedkend(end)kHOMโ shiftedkhome(home)kLFTโ shiftedkcub1(cursor back)kNXTโ shiftedknext(next)kPRVโ shiftedkprev(previous)kRITโ shiftedkcuf1(cursor forward)kUPโ shiftedkcuu1(cursor up)
Okay so keycap nomenclature on the Unix systems where curses was actually developed is not the same as the IBM PC/AT keyboard layout descendants that everyone has today, and that trips people up. So in all of the above: read โbackwardโ as โleftโ, โforwardโ as โrightโ, โnextโ as โpage downโ, and โprev(ious)โ as โpage upโ. ๐ฅ Translation service, free of charge.
These are the suffixes that denote the modifiers, bestie:
2โ Shift3โ Alt4โ Shift + Alt5โ Control6โ Shift + Control7โ Alt + Control8โ Shift + Alt + Control9โ Meta10โ Meta + Shift11โ Meta + Alt12โ Meta + Alt + Shift13โ Meta + Ctrl14โ Meta + Ctrl + Shift15โ Meta + Ctrl + Alt16โ Meta + Ctrl + Alt + Shift
And now the sad part ๐ญ ncurses defines no capabilities for modified F-keys. Terminal descriptions can refer to names that ncurses allocates at runtime to key codes. So if you actually want to use these keys in an ncurses program, an application could do this:
- using a list of extended key names, ask
tigetstr(3X)for their values, and - given that list of values, ask
key_defined(3X)for the key-code which would be returned for those keys bywgetch(3X).
That's the whole pipeline. Go forth and vibe check your function keys ๐
๐ PORTABILITY ๐
The โ-xโ extension feature of tic and infocmp has been adopted in NetBSD curses. That implementation stores user-defined capabilities, but makes no use of these capabilities itself. So likeโฆ it's holding them for you, not doing anything with them. ๐
๐ AUTHORS ๐
Thomas E. Dickey
beginning with ncurses 5.0 (1999)
๐ฆ SEE ALSO ๐ฆ
infocmp(1M), tic(1M)
In the source form of the terminal database, terminfo.src, the section โNCURSES USER-DEFINABLE CAPABILITIESโ. summarizes commonly-used user-defined capabilities employed in the terminal descriptions. Some of those features are mentioned in screen(1) or tmux(1). ๐ธ
XTerm Control Sequences provides further information on the xterm(1) features that are used in these extended capabilities. โจ
aaaaand that's the whole page, bestie, you made it to the end ๐ซถ๐ this absolute banger of a manpage is ncurses 6.6, dated 2026-01-11, and it's the user_caps(5) page โ go build something iconic with it ๐
๐โญ
๐ฆ where this page comes from
- alpine 3.24: ncurses-doc 6.6_p20260516-r0 ยท X11
- Copyright 2018-2025,2026 Thomas E. Dickey
- Copyright 2017 Free Software Foundation, Inc
- ๐ girlified by deepseek-ai/DeepSeek-V4.1-Flash-1017 on : a language model's rewrite of the original, under its terms