curs_mouse(3X)

curs_mouse(3X) πŸ’– β€” get mouse events in ncurses

okay bestie, sit DOWN, because this is the page where ncurses learns that pointing devices exist πŸ–±οΈβœ¨. Manual name is curs_mouse(3X), library calls section, and we're about to make your terminal clickable fr fr.


✨ SYNOPSIS ✨

#include <curses.h>

/* data types */
typedef unsigned long mmask_t;

typedef struct {
    short id;       /* ID to distinguish multiple devices */
    int x, y, z;    /* event coordinates */
    mmask_t bstate; /* button state bits */
} MEVENT;

/* functions */
bool has_mouse(void);

mmask_t mousemask(mmask_t new-mask, mmask_t * old-mask);

int getmouse(MEVENT * event);
int ungetmouse(MEVENT * event);

bool wenclose(const WINDOW * win, int y, int x);

bool mouse_trafo(int * pY, int * pX, bool to-screen);
bool wmouse_trafo(const WINDOW * win,
      int * pY, int * pX, bool to-screen);

int mouseinterval(int erval);

🌸 DESCRIPTION 🌸

so ncurses gives you an interface to the mouse (or any other pointing device, no gatekeeping πŸ’…). Your app can register that it's interested in those events, and then the library tells you when a mouse event shows up β€” but it does it through an input character reading function. In the non-wide character curses API that's wgetch(3X), and in the wide character API it's wget_wch(3X). iykyk.

There's a whole separate queue for this β€” distinct from the keyboard queue β€” and it hoards all the deets of mouse events. When there's a mouse event ready to be collected, the input character reading function reports the KEY_MOUSE key code. And here's the thing: a single mouse event queue serves all windows associated with the screen. One queue, all the windows, no drama.

The MEVENT structure describes a mouse event. Its y and x coordinates are screen-relative, not window-relative β€” remember that, it matters. And the bstate member has exactly one bit set</>, indicating the event type. Just one. Periodt.

⚠️ ncurses ignores mouse events when input is in canonical (β€œcooked”) mode, and it produces an error beep when they occur while the library simulates canonical mode in a window β€” like with getstr(3X) (wide-character API girlies: that's get_wstr(3X)), which expects a line feed to terminate its input loop. So yeah, don't expect clicks to do anything cute in there. 😭

πŸ¦‹ has_mouse

For any of this to work, the terminal type or the operating system interface has to support the encoding of mouse events. has_mouse returns TRUE if ncurses's mouse driver initialized successfully, and FALSE otherwise. That's it. That's the whole function. a vibe check, basically.


πŸŽ€ mousemask

Use mousemask to pick which varieties of mouse event your application wants to receive. By default, ncurses reports no mouse events β€” you gotta opt in, bestie.

  • The function returns an updated copy of new-mask indicating which event types of interest are reportable by the terminal's mouse protocol.

If the screen is not initialized, or the terminal interface does not report mouse events, mousemask returns 0. πŸ’€

  • If old-mask is not a null pointer, mousemask stores the previous value of the screen's mouse event mask there. Free history lesson, you're welcome.

As a side effect, setting a zero mouse mask may turn off the mouse cursor, and setting a nonzero mask may turn it on. Whether this happens is device-dependent β€” so don't come for me if your cursor doesn't cooperate. πŸ™ˆ


🌈 Mouse Events

Several mouse event types may be selected. You construct a mask by logically β€œor”-ing their values together. Here's the whole menu, no skips:

  • BUTTON1_PRESSED β€” mouse button 1 down
  • BUTTON1_RELEASED β€” mouse button 1 up
  • BUTTON1_CLICKED β€” mouse button 1 clicked
  • BUTTON1_DOUBLE_CLICKED β€” mouse button 1 double clicked
  • BUTTON1_TRIPLE_CLICKED β€” mouse button 1 triple clicked
  • BUTTON2_PRESSED β€” mouse button 2 down
  • BUTTON2_RELEASED β€” mouse button 2 up
  • BUTTON2_CLICKED β€” mouse button 2 clicked
  • BUTTON2_DOUBLE_CLICKED β€” mouse button 2 double clicked
  • BUTTON2_TRIPLE_CLICKED β€” mouse button 2 triple clicked
  • BUTTON3_PRESSED β€” mouse button 3 down
  • BUTTON3_RELEASED β€” mouse button 3 up
  • BUTTON3_CLICKED β€” mouse button 3 clicked
  • BUTTON3_DOUBLE_CLICKED β€” mouse button 3 double clicked
  • BUTTON3_TRIPLE_CLICKED β€” mouse button 3 triple clicked
  • BUTTON4_PRESSED β€” mouse button 4 down
  • BUTTON4_RELEASED β€” mouse button 4 up
  • BUTTON4_CLICKED β€” mouse button 4 clicked
  • BUTTON4_DOUBLE_CLICKED β€” mouse button 4 double clicked
  • BUTTON4_TRIPLE_CLICKED β€” mouse button 4 triple clicked
  • BUTTON5_PRESSED β€” mouse button 5 down
  • BUTTON5_RELEASED β€” mouse button 5 up
  • BUTTON5_CLICKED β€” mouse button 5 clicked
  • BUTTON5_DOUBLE_CLICKED β€” mouse button 5 double clicked
  • BUTTON5_TRIPLE_CLICKED β€” mouse button 5 triple clicked
  • BUTTON_SHIFT β€” a shift key was down during button state change
  • BUTTON_CTRL β€” a control key was down during button state change
  • BUTTON_ALT β€” an alt key was down during button state change
  • ALL_MOUSE_EVENTS β€” report all button state changes
  • REPORT_MOUSE_POSITION β€” report mouse movement

that's 30 whole event types like hello?? the mouse is THRIVING. 🫢


πŸ“ getmouse

When a window is configured to report a non-empty set of event types, calling the input character reading function on that window may return KEY_MOUSE to indicate that an enqueued mouse event is available. To read the event data and remove it from the queue, you call getmouse. It returns OK if a mouse event is visible in the given window, and ERR otherwise.

When getmouse returns OK, it deposits data describing the mouse event in the event pointer you supply. And a subsequent getmouse call retrieves the next older event from the queue β€” so you're walking backwards through history, cute.


🧚 ungetmouse

ungetmouse behaves analogously to ungetch(3X). It pushes a KEY_MOUSE event onto the screen's input queue, and event onto the mouse event queue. Put it back, bestie. 🫳


πŸ’œ wenclose

wenclose returns TRUE if the pair of screen-relative coordinates (y, x) is enclosed by the given window win, and FALSE otherwise. If win is a pad, wenclose uses its most recent screen coordinates as specified in a prefresh(3X) or pnoutrefresh(3X) call.

wenclose is useful for determining what subset of the screen's windows encloses the location of a mouse event. It is otherwise independent of the ncurses mouse API β€” so it's just out here doing its own thing. iconic.


⭐ wmouse_trafo

wmouse_trafo transforms the given pair of coordinate pointers (pY, pX) from a win-relative basis to a screen-relative one, or vice versa β€” as to-screen is TRUE or FALSE, respectively.

stdscr-relative coordinates are not always identical to screen coordinates: curses supports reservation of screen lines at the top and/or bottom for other purposes; see ripoffline(3X) and slk_init(3X).

  • If to-screen is TRUE and the pointers (pY, pX) reference coordinates inside win, ncurses updates their values to stdscr-relative coordinates and returns TRUE. If either pY or pX is a null pointer, or (pY, pX) is not inside win, wmouse_trafo returns FALSE.
  • If to-screen is FALSE and the pointers (pY, pX) reference coordinates inside stdscr, ncurses updates their values to win-relative coordinates and returns TRUE. If either pY or pX is a null pointer, or (pY, pX) is not inside stdscr, wmouse_trafo returns FALSE.

🌸 mouse_trafo

mouse_trafo applies the wmouse_trafo translation to stdscr. If no screen lines are reserved by ripoffline(3X) or slk_init(3X), this is the identity transformation β€” i.e. it does nothing and that's correct. no notes.


πŸ”₯ mouseinterval

mouseinterval sets the maximum time (in thousandths of a second) that can elapse between press and release events for them to be resolved as a click. An application might interpret button press and release events separated by more than the mouse interval as a β€œlong press”, or, with motion, as a β€œdrag”.

When ncurses detects a mouse event, it awaits further input activity up to this interval, and then checks for a subsequent mouse event which can be combined with the first event. If the timeout expires without input activity, then no click resolution occurs. Calling mouseinterval(0) disables click resolution.

mouseinterval returns the previous interval value. Use mouseinterval(-1) to obtain the interval without altering it. sneaky but valid. πŸ’…

The mouse interval is set to one sixth of a second when the corresponding screen is initialized, e.g., in initscr(3X) or setupterm(3X).


πŸ‘‘ RETURN VALUE πŸ‘‘

  • has_mouse, wenclose, mouse_trafo, and wmouse_trafo return TRUE or FALSE as noted above.
  • getmouse and ungetmouse return ERR upon failure and OK upon success.

getmouse fails if:

  1. no mouse driver was initialized,
  2. the mask of reportable events is zero,
  3. a mouse event was detected that does not match the mask, or
  4. no more events remain in the queue.

ungetmouse returns ERR if the event queue is full. 😭

mousemask returns the mask of reportable events.

mouseinterval returns the previous interval value, unless the terminal was not initialized. In that case, it returns the maximum interval value (166).


okay and that's the first half bestie, we're not even done yet β€” there's more coming, don't go anywhere 🌈✨

πŸ“ NOTES 🫢

okay bestie, real talk first: the order of the MEVENT structure members is not guaranteed. Like, at all. Additional fields may be added to the structure in the future. So don't go assuming things about the layout, mwah πŸ’–

Under ncurses, these calls are implemented using either xterm's built-in mouse-tracking API or platform-specific drivers including

  • Alessandro Rubini's gpm server
  • FreeBSD sysmouse
  • OS/2 EMX

If you are using an unsupported configuration, mouse events are not visible to ncurses (and the mousemask function always returns 0). Tragic, I know 😭

If the terminal type possesses the (nonstandard) terminfo string capability XM, ncurses's xterm mouse driver uses it when initializing the terminal for mouse operation. The default, if XM is not found, corresponds to private mode 1000 of xterm.

\E[?1000%?%p1%{1}%=%th%el%;

ncurses also recognizes xterm's newer private mode 1006.

\E[?1006;1000%?%p1%{1}%=%th%el%;

The id member of the mouse event structure is not presently used; no terminal type or operating system interface supports reporting events from distinguishable pointing devices. If you synthesize an MEVENT, use an id of 0. slay responsibly ✨

The z member of the mouse event structure is not presently used. It is intended for use with touch screens (which may be pressure-sensitive) or with 3D-mice/trackballs/power gloves. future tech vibes fr fr πŸ¦‹

The ALL_MOUSE_EVENTS class does not include REPORT_MOUSE_POSITION. They are distinct. For example, in xterm, wheel/scrolling mice send position reports as a sequence of presses of buttons 4 or 5 without matching button-releases. it's giving chaos but okay πŸ’€


✨ EXTENSIONS ✨

These functions are ncurses extensions, and are not found in SVr4 curses, 4.4BSD curses, or any other previous curses implementation. (SVr4 curses did have a getmouse function, which took no argument and returned an unsigned long.) so like, don't go looking for them elsewhere, they're ours πŸ’…


🌈 PORTABILITY 🌈

Applications employing the ncurses mouse extension should condition its use on the visibility of the NCURSES_MOUSE_VERSION preprocessor macro. When the interface changes, the macro's value increments. Multiple versions are available when ncurses is configured; see section β€œALTERNATE CONFIGURATIONS” of ncurses(3X). The following values may be specified.

  • 1 β€” has definitions for reserved events. The mask uses 28 bits.
  • 2 β€” adds definitions for button 5, removes the definitions for reserved events. The mask uses 29 bits.
check the macro before you commit, bestie. no cap πŸ‘‘

πŸ‘‘ HISTORY πŸ‘‘

SVr4 (1989) added mouse support to its variant of xterm(1). It is mentioned in a few places, with little supporting documentation. lowkey mysterious of them πŸ™ˆ

  • Its β€œlibcurses” manual page lists functions for this feature prototyped in curses.h.
extern int mouse_set(long int);
extern int mouse_on(long int);
extern int mouse_off(long int);
extern int request_mouse_pos(void);
extern int map_button(unsigned long);
extern void wmouse_position(WINDOW *, int *, int *);
extern unsigned long getmouse(void), getbmap(void);
  • Its β€œterminfo” manual page lists capabilities for the feature.
buttons         btns    BT   Number of buttons on the mouse
get_mouse       getm    Gm   Curses should get button events
key_mouse       kmous   Km   0631, Mouse event has occurred
mouse_info      minfo   Mi   Mouse status information
req_mouse_pos   reqmp   RQ   Request mouse position report
  • The interface made assumptions (as does ncurses) about the escape sequences sent to and received from the terminal.

For instance, the SVr4 curses library used the get_mouse (getm) capability to tell the terminal which mouse button events it should send, passing the mouse-button bit mask to the terminal. Also, it could ask the terminal where the mouse was using the req_mouse_pos (reqmp) capability.

Those features required a terminal program that had been modified to work with SVr4 curses. They were not part of the X Consortium's xterm.

When developing the xterm mouse support for ncurses in September 1995, Eric Raymond was uninterested in using the same interface due to its lack of documentation. Later, in 1998, Mark Hesseling provided support in PDCurses 2.3 using the SVr4 interface. PDCurses, however, does not use video terminals, making it unnecessary to be concerned about compatibility with the escape sequences. the way they just… moved on 😭 iconic honestly


🚨 BUGS 🚨

Mouse events from xterm are not ignored in canonical (β€œcooked”) mode if they have been enabled by mousemask. Instead, the xterm mouse report sequence appears in the string read. yeah. that's a whole thing πŸ’€

An ncurses window must enable keypad(3X) to correctly receive mouse event reports from xterm since they are encoded like function keys. Set the terminal's terminfo capability key_mouse (kmous) to β€œ\E[M” (the beginning of the response from xterm for mouse clicks). Other values of key_mouse are permitted under the same assumption β€” that is, a mouse report begins with the value of the key_mouse (kmous) string capability.

Because there are no standard response sequences that serve to identify terminals supporting the xterm mouse protocol, ncurses assumes that if key_mouse (kmous) is defined in the terminal description, or if the terminal type's primary name or aliases contain the string β€œxterm”, then the terminal may send mouse events. ncurses checks the kmous cap-code first, allowing use of newer xterm mouse protocols, such as its private mode 1006. bestie really said "I'll check myself" and honestly? valid πŸŽ€


πŸ¦‹ SEE ALSO πŸ¦‹

curses(3X), curs_inopts(3X), curs_kernel(3X), curs_pad(3X), curs_slk(3X), curs_variables(3X)


and that's the whole page, bestie!! πŸ’– this one's from ncurses 6.6, dated 2025-11-11 β€” go forth and track those mouse events, you absolute icon 🌸✨🫢

πŸ“¦ where this page comes from