3 Michael Stapelberg <michael@i3wm.org>
6 This document is intended to be the first thing you read before looking and/or
7 touching i3’s source code. It should contain all important information to help
8 you understand why things are like they are. If it does not mention something
9 you find necessary, please do not hesitate to contact me.
13 A window manager is not necessarily needed to run X, but it is usually used in
14 combination with X to facilitate some things. The window manager's job is to
15 take care of the placement of windows, to provide the user with some mechanisms
16 to change the position/size of windows and to communicate with clients to a
17 certain extent (for example handle fullscreen requests of clients such as
20 There are no different contexts in which X11 clients run, so a window manager
21 is just another client, like all other X11 applications. However, it handles
22 some events which normal clients usually don’t handle.
24 In the case of i3, the tasks (and order of them) are the following:
26 . Grab the key bindings (events will be sent upon keypress/keyrelease)
27 . Iterate through all existing windows (if the window manager is not started as
28 the first client of X) and manage them (reparent them, create window
30 . When new windows are created, manage them
31 . Handle the client’s `_WM_STATE` property, but only `_WM_STATE_FULLSCREEN` and
32 `_NET_WM_STATE_DEMANDS_ATTENTION`
33 . Handle the client’s `WM_NAME` property
34 . Handle the client’s size hints to display them proportionally
35 . Handle the client’s urgency hint
36 . Handle enter notifications (focus follows mouse)
37 . Handle button (as in mouse buttons) presses for focus/raise on click
38 . Handle expose events to re-draw own windows such as decorations
39 . React to the user’s commands: Change focus, Move windows, Switch workspaces,
40 Change the layout mode of a container (default/stacking/tabbed), start a new
41 application, restart the window manager
43 In the following chapters, each of these tasks and their implementation details
46 === Tiling window managers
48 Traditionally, there are two approaches to managing windows: The most common
49 one nowadays is floating, which means the user can freely move/resize the
50 windows. The other approach is called tiling, which means that your window
51 manager distributes windows to use as much space as possible while not
52 overlapping each other.
54 The idea behind tiling is that you should not need to waste your time
55 moving/resizing windows while you usually want to get some work done. After
56 all, most users sooner or later tend to lay out their windows in a way which
57 corresponds to tiling or stacking mode in i3. Therefore, why not let i3 do this
58 for you? Certainly, it’s faster than you could ever do it.
60 The problem with most tiling window managers is that they are too inflexible.
61 In my opinion, a window manager is just another tool, and similar to vim which
62 can edit all kinds of text files (like source code, HTML, …) and is not limited
63 to a specific file type, a window manager should not limit itself to a certain
64 layout (like dwm, awesome, …) but provide mechanisms for you to easily create
65 the layout you need at the moment.
69 The data structure which i3 uses to keep track of your windows is a tree. Every
70 node in the tree is a container (type +Con+). Some containers represent actual
71 windows (every container with a +window != NULL+), some represent split
72 containers and a few have special purposes: they represent workspaces, outputs
73 (like VGA1, LVDS1, …) or the X11 root window.
75 So, when you open a terminal and immediately open another one, they reside in
76 the same split container, which uses the default layout. In case of an empty
77 workspace, the split container we are talking about is the workspace.
79 To get an impression of how different layouts are represented, just play around
80 and look at the data structures -- they are exposed as a JSON hash. See
81 http://i3wm.org/docs/ipc.html#_tree_reply for documentation on that and an
86 include/atoms.xmacro::
87 A file containing all X11 atoms which i3 uses. This file will be included
88 various times (for defining, requesting and receiving the atoms), each time
89 with a different definition of xmacro().
92 Contains data definitions used by nearly all files. You really need to read
96 Contains forward definitions for all public functions, as well as
97 doxygen-compatible comments (so if you want to get a bit more of the big
98 picture, either browse all header files or use doxygen if you prefer that).
101 Contains the lexer for i3’s configuration file, written for +flex(1)+.
104 Contains the parser for i3’s configuration file, written for +bison(1)+.
107 Contains all functions which handle mouse button clicks (right mouse button
108 clicks initiate resizing and thus are relatively complex).
111 Contains the lexer for i3 commands, written for +flex(1)+.
114 Contains the parser for i3 commands, written for +bison(1)+.
117 Contains all functions which deal with containers directly (creating
118 containers, searching containers, getting specific properties from containers,
122 Contains all functions handling the configuration file (calling the parser
123 (src/cfgparse.y) with the correct path, switching key bindings mode).
126 Contains debugging functions to print unhandled X events.
129 Functions to get/set certain EWMH properties easily.
132 Contains functions for floating mode (mostly resizing/dragging).
135 Contains all handlers for all kinds of X events (new window title, new hints,
136 unmapping, key presses, button presses, …).
139 Contains code for the IPC interface.
142 Contains code for loading layouts from JSON files.
145 Contains the logging functions.
148 Initializes the window manager.
151 Looks at existing or new windows and decides whether to manage them. If so, it
152 reparents the window and inserts it into our data structures.
155 A "match" is a data structure which acts like a mask or expression to match
156 certain windows or not. For example, when using commands, you can specify a
157 command like this: [title="*Firefox*"] kill. The title member of the match
158 data structure will then be filled and i3 will check each window using
159 match_matches_window() to find the windows affected by this command.
162 Contains code to move a container in a specific direction.
165 Functions to handle CT_OUTPUT cons.
168 The RandR API is used to get (and re-query) the configured outputs (monitors,
172 Renders the tree data structure by assigning coordinates to every node. These
173 values will later be pushed to X11 in +src/x.c+.
176 Contains the functions to resize containers.
179 Handles +SIGSEGV+, +SIGABRT+ and +SIGFPE+ by showing a dialog that i3 crashed.
180 You can chose to let it dump core, to restart it in-place or to restart it
181 in-place but forget about the layout.
184 Contains functions which open or close containers in the tree, change focus or
185 cleanup ("flatten") the tree. See also +src/move.c+ for another similar
186 function, which was moved into its own file because it is so long.
189 Contains useful functions which are not really dependant on anything.
192 Handlers to update X11 window properties like +WM_CLASS+, +_NET_WM_NAME+,
193 +CLIENT_LEADER+, etc.
196 Contains all functions related to workspaces (displaying, hiding, renaming…)
199 Transfers our in-memory tree (see +src/render.c+) to X11.
202 Contains wrappers to use xcb more easily.
205 XCursor functions (for cursor themes).
208 Legacy support for Xinerama. See +src/randr.c+ for the preferred API.
213 See include/data.h for documented data structures. The most important ones are
214 explained right here.
216 /////////////////////////////////////////////////////////////////////////////////
217 // TODO: update image
219 image:bigpicture.png[The Big Picture]
221 /////////////////////////////////////////////////////////////////////////////////
223 So, the hierarchy is:
225 . *X11 root window*, the root container
226 . *Output container* (LVDS1 in this example)
227 . *Content container* (there are also containers for dock windows)
228 . *Workspaces* (Workspace 1 in this example, with horizontal orientation)
229 . *Split container* (vertically split)
230 . *X11 window containers*
232 The data type is +Con+, in all cases.
236 The X11 root window is a single window per X11 display (a display is identified
237 by +:0+ or +:1+ etc.). The root window is what you draw your background image
238 on. It spans all the available outputs, e.g. +VGA1+ is a specific part of the
239 root window and +LVDS1+ is a specific part of the root window.
243 Every active output obtained through RandR is represented by one output
244 container. Outputs are considered active when a mode is configured (meaning
245 something is actually displayed on the output) and the output is not a clone.
247 For example, if your notebook has a screen resolution of 1280x800 px and you
248 connect a video projector with a resolution of 1024x768 px, set it up in clone
249 mode (+xrandr \--output VGA1 \--mode 1024x768 \--same-as LVDS1+), i3 will
250 reduce the resolution to the lowest common resolution and disable one of the
251 cloned outputs afterwards.
253 However, if you configure it using +xrandr \--output VGA1 \--mode 1024x768
254 \--right-of LVDS1+, i3 will set both outputs active. For each output, a new
255 workspace will be assigned. New workspaces are created on the output you are
258 === Content container
260 Each output has multiple children. Two of them are dock containers which hold
261 dock clients. The other one is the content container, which holds the actual
262 content (workspaces) of this output.
266 A workspace is identified by its name. Basically, you could think of
267 workspaces as different desks in your office, if you like the desktop
268 metaphor. They just contain different sets of windows and are completely
269 separate of each other. Other window managers also call this ``Virtual
274 A split container is a container which holds an arbitrary amount of split
275 containers or X11 window containers. It has an orientation (horizontal or
276 vertical) and a layout.
278 Split containers (and X11 window containers, which are a subtype of split
279 containers) can have different border styles.
281 === X11 window container
283 An X11 window container holds exactly one X11 window. These are the leaf nodes
284 of the layout tree, they cannot have any children.
288 i3 makes heavy use of the list macros defined in BSD operating systems. To
289 ensure that the operating system on which i3 is compiled has all the expected
290 features, i3 comes with `include/queue.h`. On BSD systems, you can use man
291 `queue(3)`. On Linux, you have to use google (or read the source).
293 The lists used are +SLIST+ (single linked lists), +CIRCLEQ+ (circular
294 queues) and +TAILQ+ (tail queues). Usually, only forward traversal is necessary,
295 so an `SLIST` works fine. If inserting elements at arbitrary positions or at
296 the end of a list is necessary, a +TAILQ+ is used instead. However, for the
297 windows inside a container, a +CIRCLEQ+ is necessary to go from the currently
298 selected window to the window above/below.
300 == Naming conventions
302 There is a row of standard variables used in many events. The following names
303 should be chosen for those:
305 * ``conn'' is the xcb_connection_t
306 * ``event'' is the event of the particular type
307 * ``con'' names a container
308 * ``current'' is a loop variable when using +TAILQ_FOREACH+ etc.
310 == Startup (src/mainx.c, main())
312 * Establish the xcb connection
313 * Check for XKB extension on the separate X connection, load Xcursor
314 * Check for RandR screens (with a fall-back to Xinerama)
315 * Grab the keycodes for which bindings exist
316 * Manage all existing windows
317 * Enter the event loop
321 === Grabbing the bindings
323 Grabbing the bindings is quite straight-forward. You pass X your combination of
324 modifiers and the keycode you want to grab and whether you want to grab them
325 actively or passively. Most bindings (everything except for bindings using
326 Mode_switch) are grabbed passively, that is, just the window manager gets the
327 event and cannot replay it.
329 We need to grab bindings that use Mode_switch actively because of a bug in X.
330 When the window manager receives the keypress/keyrelease event for an actively
331 grabbed keycode, it has to decide what to do with this event: It can either
332 replay it so that other applications get it or it can prevent other
333 applications from receiving it.
335 So, why do we need to grab keycodes actively? Because X does not set the
336 state-property of keypress/keyrelease events properly. The Mode_switch bit is
337 not set and we need to get it using XkbGetState. This means we cannot pass X
338 our combination of modifiers containing Mode_switch when grabbing the key and
339 therefore need to grab the keycode itself without any modifiers. This means,
340 if you bind Mode_switch + keycode 38 ("a"), i3 will grab keycode 38 ("a") and
341 check on each press of "a" if the Mode_switch bit is set using XKB. If yes, it
342 will handle the event, if not, it will replay the event.
344 === Handling a keypress
346 As mentioned in "Grabbing the bindings", upon a keypress event, i3 first gets
349 Then, it looks through all bindings and gets the one which matches the received
352 The bound command is parsed by the cmdparse lexer/parser, see +parse_cmd+ in
355 == Manage windows (src/main.c, manage_window() and reparent_window())
357 `manage_window()` does some checks to decide whether the window should be
360 * Windows have to be mapped, that is, visible on screen
361 * The override_redirect must not be set. Windows with override_redirect shall
362 not be managed by a window manager
364 Afterwards, i3 gets the initial geometry and reparents the window (see
365 `reparent_window()`) if it wasn’t already managed.
367 Reparenting means that for each window which is reparented, a new window,
368 slightly larger than the original one, is created. The original window is then
369 reparented to the bigger one (called "frame").
371 After reparenting, the window type (`_NET_WM_WINDOW_TYPE`) is checked to see
372 whether this window is a dock (`_NET_WM_WINDOW_TYPE_DOCK`), like dzen2 for
373 example. Docks are handled differently, they don’t have decorations and are not
374 assigned to a specific container. Instead, they are positioned at the bottom
375 or top of the screen (in the appropriate dock area containers). To get the
376 height which needs to be reserved for the window, the `_NET_WM_STRUT_PARTIAL`
379 Furthermore, the list of assignments (to other workspaces, which may be on
380 other screens) is checked. If the window matches one of the user’s criteria,
381 it may either be put in floating mode or moved to a different workspace. If the
382 target workspace is not visible, the window will not be mapped.
384 == What happens when an application is started?
386 i3 does not care about applications. All it notices is when new windows are
387 mapped (see `src/handlers.c`, `handle_map_request()`). The window is then
388 reparented (see section "Manage windows").
390 After reparenting the window, `render_tree()` is called which renders the
391 internal layout table. The new window has been placed in the currently focused
392 container and therefore the new window and the old windows (if any) need to be
393 moved/resized so that the currently active layout (default/stacking/tabbed mode)
394 is rendered correctly. To move/resize windows, a window is ``configured'' in
397 Some applications, such as MPlayer obviously assume the window manager is
398 stupid and try to configure their windows by themselves. This generates an
399 event called configurerequest. i3 handles these events and tells the window the
400 size it had before the configurerequest (with the exception of not yet mapped
401 windows, which get configured like they want to, and floating windows, which
402 can reconfigure themselves).
406 Only the _NET_WM_STATE_FULLSCREEN atom is handled. It calls
407 ``toggle_fullscreen()'' for the specific client which just configures the
408 client to use the whole screen on which it currently is. Also, it is set as
409 fullscreen_client for the i3Screen.
413 When the WM_NAME property of a window changes, its decoration (containing the
414 title) is re-rendered. Note that WM_NAME is in COMPOUND_TEXT encoding which is
415 totally uncommon and cumbersome. Therefore, the _NET_WM_NAME atom will be used
420 Like WM_NAME, this atom contains the title of a window. However, _NET_WM_NAME
421 is encoded in UTF-8. i3 will recode it to UCS-2 in order to be able to pass it
422 to X. Using an appropriate font (ISO-10646), you can see most special
423 characters (every special character contained in your font).
427 Size hints specify the minimum/maximum size for a given window as well as its
428 aspect ratio. This is important for clients like mplayer, who only set the
429 aspect ratio and resize their window to be as small as possible (but only with
430 some video outputs, for example in Xv, while when using x11, mplayer does the
431 necessary centering for itself).
433 So, when an aspect ratio was specified, i3 adjusts the height of the window
434 until the size maintains the correct aspect ratio. For the code to do this, see
435 src/layout.c, function resize_client().
437 == Rendering (src/layout.c, render_layout() and render_container())
439 Rendering in i3 version 4 is the step which assigns the correct sizes for
440 borders, decoration windows, child windows and the stacking order of all
441 windows. In a separate step (+x_push_changes()+), these changes are pushed to
444 Keep in mind that all these properties (+rect+, +window_rect+ and +deco_rect+)
445 are temporary, meaning they will be overwritten by calling +render_con+.
446 Persistent position/size information is kept in +geometry+.
448 The entry point for every rendering operation (except for the case of moving
449 floating windows around) currently is +tree_render()+ which will re-render
450 everything that’s necessary (for every output, only the currently displayed
451 workspace is rendered). This behavior is expected to change in the future,
452 since for a lot of updates, re-rendering everything is not actually necessary.
453 Focus was on getting it working correct, not getting it work very fast.
455 What +tree_render()+ actually does is calling +render_con()+ on the root
456 container and then pushing the changes to X11. The following sections talk
457 about the different rendering steps, in the order of "top of the tree" (root
458 container) to the bottom.
460 === Rendering the root container
462 The i3 root container (`con->type == CT_ROOT`) represents the X11 root window.
463 It contains one child container for every output (like LVDS1, VGA1, …), which
464 is available on your computer.
466 Rendering the root will first render all tiling windows and then all floating
467 windows. This is necessary because a floating window can be positioned in such
468 a way that it is visible on two different outputs. Therefore, by first
469 rendering all the tiling windows (of all outputs), we make sure that floating
470 windows can never be obscured by tiling windows.
472 Essentially, though, this code path will just call +render_con()+ for every
473 output and +x_raise_con(); render_con()+ for every floating window.
475 In the special case of having a "global fullscreen" window (fullscreen mode
476 spanning all outputs), a shortcut is taken and +x_raise_con(); render_con()+ is
477 only called for the global fullscreen window.
479 === Rendering an output
481 Output containers (`con->layout == L_OUTPUT`) represent a hardware output like
482 LVDS1, VGA1, etc. An output container has three children (at the moment): One
483 content container (having workspaces as children) and the top/bottom dock area
486 The rendering happens in the function +render_l_output()+ in the following
489 1. Find the content container (`con->type == CT_CON`)
490 2. Get the currently visible workspace (+con_get_fullscreen_con(content,
492 3. If there is a fullscreened window on that workspace, directly render it and
493 return, thus ignoring the dock areas.
494 4. Sum up the space used by all the dock windows (they have a variable height
496 5. Set the workspace rects (x/y/width/height) based on the position of the
497 output (stored in `con->rect`) and the usable space
498 (`con->rect.{width,height}` without the space used for dock windows).
499 6. Recursively raise and render the output’s child containers (meaning dock
500 area containers and the content container).
502 === Rendering a workspace or split container
504 From here on, there really is no difference anymore. All containers are of
505 `con->type == CT_CON` (whether workspace or split container) and some of them
506 have a `con->window`, meaning they represent an actual window instead of a
511 In default layout, containers are placed horizontally or vertically next to
512 each other (depending on the `con->orientation`). If a child is a leaf node (as
513 opposed to a split container) and has border style "normal", appropriate space
514 will be reserved for its window decoration.
518 In stacked layout, only the focused window is actually shown (this is achieved
519 by calling +x_raise_con()+ in reverse focus order at the end of +render_con()+).
521 The available space for the focused window is the size of the container minus
522 the height of the window decoration for all windows inside this stacked
525 If border style is "1pixel" or "none", no window decoration height will be
526 reserved (or displayed later on), unless there is more than one window inside
527 the stacked container.
531 Tabbed layout works precisely like stacked layout, but the window decoration
532 position/size is different: They are placed next to each other on a single line
535 ==== Dock area layout
537 This is a special case. Users cannot choose the dock area layout, but it will be
538 set for the dock area containers. In the dockarea layout (at the moment!),
539 windows will be placed above each other.
541 === Rendering a window
543 A window’s size and position will be determined in the following way:
545 1. Subtract the border if border style is not "none" (but "normal" or "1pixel").
546 2. Subtract the X11 border, if the window has an X11 border > 0.
547 3. Obey the aspect ratio of the window (think MPlayer).
548 4. Obey the height- and width-increments of the window (think terminal emulator
549 which can only be resized in one-line or one-character steps).
551 == Pushing updates to X11 / Drawing
553 A big problem with i3 before version 4 was that we just sent requests to X11
554 anywhere in the source code. This was bad because nobody could understand the
555 entirety of our interaction with X11, it lead to subtle bugs and a lot of edge
556 cases which we had to consider all over again.
558 Therefore, since version 4, we have a single file, +src/x.c+, which is
559 responsible for repeatedly transferring parts of our tree datastructure to X11.
561 +src/x.c+ consists of multiple parts:
563 1. The state pushing: +x_push_changes()+, which calls +x_push_node()+.
564 2. State modification functions: +x_con_init+, +x_reinit+,
565 +x_reparent_child+, +x_move_win+, +x_con_kill+, +x_raise_con+, +x_set_name+
567 3. Expose event handling (drawing decorations): +x_deco_recurse()+ and
568 +x_draw_decoration()+.
570 === Pushing state to X11
572 In general, the function +x_push_changes+ should be called to push state
573 changes. Only when the scope of the state change is clearly defined (for
574 example only the title of a window) and its impact is known beforehand, one can
575 optimize this and call +x_push_node+ on the appropriate con directly.
577 +x_push_changes+ works in the following steps:
579 1. Clear the eventmask for all mapped windows. This leads to not getting
580 useless ConfigureNotify or EnterNotify events which are caused by our
581 requests. In general, we only want to handle user input.
582 2. Stack windows above each other, in reverse stack order (starting with the
583 most obscured/bottom window). This is relevant for floating windows which
584 can overlap each other, but also for tiling windows in stacked or tabbed
585 containers. We also update the +_NET_CLIENT_LIST_STACKING+ hint which is
586 necessary for tab drag and drop in Chromium.
587 3. +x_push_node+ will be called for the root container, recursively calling
588 itself for the container’s children. This function actually pushes the
589 state, see the next paragraph.
590 4. If the pointer needs to be warped to a different position (for example when
591 changing focus to a differnt output), it will be warped now.
592 5. The eventmask is restored for all mapped windows.
593 6. Window decorations will be rendered by calling +x_deco_recurse+ on the root
594 container, which then recursively calls itself for the children.
595 7. If the input focus needs to be changed (because the user focused a different
596 window), it will be updated now.
597 8. +x_push_node_unmaps+ will be called for the root container. This function
598 only pushes UnmapWindow requests. Separating the state pushing is necessary
599 to handle fullscreen windows (and workspace switches) in a smooth fashion:
600 The newly visible windows should be visible before the old windows are
603 +x_push_node+ works in the following steps:
605 1. Update the window’s +WM_NAME+, if changed (the +WM_NAME+ is set on i3
606 containers mainly for debugging purposes).
607 2. Reparents a child window into the i3 container if the container was created
608 for a specific managed window.
609 3. If the size/position of the i3 container changed (due to opening a new
610 window or switching layouts for example), the window will be reconfigured.
611 Also, the pixmap which is used to draw the window decoration/border on is
612 reconfigured (pixmaps are size-dependent).
613 4. Size/position for the child window is adjusted.
614 5. The i3 container is mapped if it should be visible and was not yet mapped.
615 When mapping, +WM_STATE+ is set to +WM_STATE_NORMAL+. Also, the eventmask of
616 the child window is updated and the i3 container’s contents are copied from
618 6. +x_push_node+ is called recursively for all children of the current
621 +x_push_node_unmaps+ handles the remaining case of an i3 container being
622 unmapped if it should not be visible anymore. +WM_STATE+ will be set to
623 +WM_STATE_WITHDRAWN+.
626 === Drawing window decorations/borders/backgrounds
628 +x_draw_decoration+ draws window decorations. It is run for every leaf
629 container (representing an actual X11 window) and for every non-leaf container
630 which is in a stacked/tabbed container (because stacked/tabbed containers
631 display a window decoration for split containers, which at the moment just says
632 "another container").
634 Then, parameters are collected to be able to determine whether this decoration
635 drawing is actually necessary or was already done. This saves a substantial
636 number of redraws (depending on your workload, but far over 50%).
638 Assuming that we need to draw this decoration, we start by filling the empty
639 space around the child window (think of MPlayer with a specific aspect ratio)
640 in the user-configured client background color.
642 Afterwards, we draw the appropriate border (in case of border styles "normal"
643 and "1pixel") and the top bar (in case of border style "normal").
645 The last step is drawing the window title on the top bar.
648 /////////////////////////////////////////////////////////////////////////////////
650 == Resizing containers
652 By clicking and dragging the border of a container, you can resize the whole
653 column (respectively row) which this container is in. This is necessary to keep
654 the table layout working and consistent.
656 The resizing works similarly to the resizing of floating windows or movement of
659 * A new, invisible window with the size of the root window is created
661 * Another window, 2px width and as high as your screen (or vice versa for
662 horizontal resizing) is created. Its background color is the border color and
663 it is only there to inform the user how big the container will be (it
664 creates the impression of dragging the border out of the container).
665 * The +drag_pointer+ function of +src/floating.c+ is called to grab the pointer
666 and enter its own event loop which will pass all events (expose events) but
667 motion notify events. This function then calls the specified callback
668 (+resize_callback+) which does some boundary checking and moves the helper
669 window. As soon as the mouse button is released, this loop will be
671 * The new width_factor for each involved column (respectively row) will be
674 /////////////////////////////////////////////////////////////////////////////////
676 == User commands (parser-specs/commands.spec)
678 In the configuration file and when using i3 interactively (with +i3-msg+, for
679 example), you use commands to make i3 do things, like focus a different window,
680 set a window to fullscreen, and so on. An example command is +floating enable+,
681 which enables floating mode for the currently focused window. See the
682 appropriate section in the link:userguide.html[User’s Guide] for a reference of
685 In earlier versions of i3, interpreting these commands was done using lex and
686 yacc, but experience has shown that lex and yacc are not well suited for our
687 command language. Therefore, starting from version 4.2, we use a custom parser
688 for user commands (not yet for the configuration file).
689 The input specification for this parser can be found in the file
690 +parser-specs/commands.spec+. Should you happen to use Vim as an editor, use
691 :source parser-specs/highlighting.vim to get syntax highlighting for this file
692 (highlighting files for other editors are welcome).
694 .Excerpt from commands.spec
695 -----------------------------------------------------------------------
697 '[' -> call cmd_criteria_init(); CRITERIA
700 'workspace' -> WORKSPACE
701 'exit' -> call cmd_exit()
702 'restart' -> call cmd_restart()
703 'reload' -> call cmd_reload()
704 -----------------------------------------------------------------------
706 The input specification is written in an extremely simple format. The
707 specification is then converted into C code by the Perl script
708 generate-commands-parser.pl (the output file names begin with GENERATED and the
709 files are stored in the +include+ directory). The parser implementation
710 +src/commands_parser.c+ includes the generated C code at compile-time.
712 The above excerpt from commands.spec illustrates nearly all features of our
713 specification format: You describe different states and what can happen within
714 each state. State names are all-caps; the state in the above excerpt is called
715 INITIAL. A list of tokens and their actions (separated by an ASCII arrow)
716 follows. In the excerpt, all tokens are literals, that is, simple text strings
717 which will be compared with the input. An action is either the name of a state
718 in which the parser will transition into, or the keyword 'call', followed by
719 the name of a function (and optionally a state).
721 === Example: The WORKSPACE state
723 Let’s have a look at the WORKSPACE state, which is a good example of all
724 features. This is its definition:
726 .WORKSPACE state (commands.spec)
727 ----------------------------------------------------------------
728 # workspace next|prev|next_on_output|prev_on_output
729 # workspace back_and_forth
732 direction = 'next_on_output', 'prev_on_output', 'next', 'prev'
733 -> call cmd_workspace($direction)
735 -> call cmd_workspace_back_and_forth()
737 -> call cmd_workspace_name($workspace)
738 ----------------------------------------------------------------
740 As you can see from the commands, there are multiple different valid variants
741 of the workspace command:
743 workspace <direction>::
744 The word 'workspace' can be followed by any of the tokens 'next',
745 'prev', 'next_on_output' or 'prev_on_output'. This command will
746 switch to the next or previous workspace (optionally on the same
748 There is one function called +cmd_workspace+, which is defined
749 in +src/commands.c+. It will handle this kind of command. To know which
750 direction was specified, the direction token is stored on the stack
751 with the name "direction", which is what the "direction = " means in
754 NOTE: Note that you can specify multiple literals in the same line. This has
755 exactly the same effect as if you specified `direction =
756 'next_on_output' -> call cmd_workspace($direction)` and so forth. +
758 NOTE: Also note that the order of literals is important here: If 'next' were
759 ordered before 'next_on_output', then 'next_on_output' would never
762 workspace back_and_forth::
763 This is a very simple case: When the literal 'back_and_forth' is found
764 in the input, the function +cmd_workspace_back_and_forth+ will be
765 called without parameters and the parser will return to the INITIAL
766 state (since no other state was specified).
768 In this case, the workspace command is followed by an arbitrary string,
769 possibly in quotes, for example "workspace 3" or "workspace bleh". +
770 This is the first time that the token is actually not a literal (not in
771 single quotes), but just called string. Other possible tokens are word
772 (the same as string, but stops matching at a whitespace) and end
773 (matches the end of the input).
775 === Introducing a new command
777 The following steps have to be taken in order to properly introduce a new
778 command (or possibly extend an existing command):
780 1. Define a function beginning with +cmd_+ in the file +src/commands.c+. Copy
781 the prototype of an existing function.
782 2. After adding a comment on what the function does, copy the comment and
783 function definition to +include/commands.h+. Make the comment in the header
784 file use double asterisks to make doxygen pick it up.
785 3. Write a test case (or extend an existing test case) for your feature, see
786 link:testsuite.html[i3 testsuite]. For now, it is sufficient to simply call
787 your command in all the various possible ways.
788 4. Extend the parser specification in +parser-specs/commands.spec+. Run the
789 testsuite and see if your new function gets called with the appropriate
790 arguments for the appropriate input.
791 5. Actually implement the feature.
792 6. Document the feature in the link:userguide.html[User’s Guide].
796 The movement code is pretty delicate. You need to consider all cases before
797 making any changes or before being able to fully understand how it works.
799 === Case 1: Moving inside the same container
801 The reference layout for this case is a single workspace in horizontal
802 orientation with two containers on it. Focus is on the left container (1).
805 [width="15%",cols="^,^"]
810 When moving the left window to the right (command +move right+), tree_move will
811 look for a container with horizontal orientation and finds the parent of the
812 left container, that is, the workspace. Afterwards, it runs the code branch
813 commented with "the easy case": it calls TAILQ_NEXT to get the container right
814 of the current one and swaps both containers.
816 === Case 2: Move a container into a split container
818 The reference layout for this case is a horizontal workspace with two
819 containers. The right container is a v-split with two containers. Focus is on
820 the left container (1).
822 [width="15%",cols="^,^"]
828 When moving to the right (command +move right+), i3 will work like in case 1
829 ("the easy case"). However, as the right container is not a leaf container, but
830 a v-split, the left container (1) will be inserted at the right position (below
831 2, assuming that 2 is focused inside the v-split) by calling +insert_con_into+.
833 +insert_con_into+ detaches the container from its parent and inserts it
834 before/after the given target container. Afterwards, the on_remove_child
835 callback is called on the old parent container which will then be closed, if
838 Afterwards, +con_focus+ will be called to fix the focus stack and the tree will
841 === Case 3: Moving to non-existant top/bottom
843 Like in case 1, the reference layout for this case is a single workspace in
844 horizontal orientation with two containers on it. Focus is on the left
847 [width="15%",cols="^,^"]
852 This time however, the command is +move up+ or +move down+. tree_move will look
853 for a container with vertical orientation. As it will not find any,
854 +same_orientation+ is NULL and therefore i3 will perform a forced orientation
855 change on the workspace by creating a new h-split container, moving the
856 workspace contents into it and then changing the workspace orientation to
857 vertical. Now it will again search for parent containers with vertical
858 orientation and it will find the workspace.
860 This time, the easy case code path will not be run as we are not moving inside
861 the same container. Instead, +insert_con_into+ will be called with the focused
862 container and the container above/below the current one (on the level of
865 Now, +con_focus+ will be called to fix the focus stack and the tree will be
868 === Case 4: Moving to existant top/bottom
870 The reference layout for this case is a vertical workspace with two containers.
871 The bottom one is a h-split containing two containers (1 and 2). Focus is on
872 the bottom left container (1).
874 [width="15%",cols="^,^"]
880 This case is very much like case 3, only this time the forced workspace
881 orientation change does not need to be performed because the workspace already
882 is in vertical orientation.
884 === Case 5: Moving in one-child h-split
886 The reference layout for this case is a horizontal workspace with two
887 containers having a v-split on the left side with a one-child h-split on the
888 bottom. Focus is on the bottom left container (2(h)):
890 [width="15%",cols="^,^"]
896 In this case, +same_orientation+ will be set to the h-split container around
897 the focused container. However, when trying the easy case, the next/previous
898 container +swap+ will be NULL. Therefore, i3 will search again for a
899 +same_orientation+ container, this time starting from the parent of the h-split
902 After determining a new +same_orientation+ container (if it is NULL, the
903 orientation will be force-changed), this case is equivalent to case 2 or case
907 === Case 6: Floating containers
909 The reference layout for this case is a horizontal workspace with two
910 containers plus one floating h-split container. Focus is on the floating
913 TODO: nice illustration. table not possible?
915 When moving up/down, the container needs to leave the floating container and it
916 needs to be placed on the workspace (at workspace level). This is accomplished
917 by calling the function +attach_to_workspace+.
921 Without much ado, here is the list of cases which need to be considered:
923 * click to focus (tiling + floating) and raise (floating)
924 * click to focus/raise when in stacked/tabbed mode
925 * floating_modifier + left mouse button to drag a floating con
926 * floating_modifier + right mouse button to resize a floating con
927 * click on decoration in a floating con to either initiate a resize (if there
928 is more than one child in the floating con) or to drag the
929 floating con (if it’s the one at the top).
930 * click on border in a floating con to resize the floating con
931 * floating_modifier + right mouse button to resize a tiling con
932 * click on border/decoration to resize a tiling con
936 * Forgetting to call `xcb_flush(conn);` after sending a request. This usually
937 leads to code which looks like it works fine but which does not work under
940 * Forgetting to call `floating_fix_coordinates(con, old_rect, new_rect)` after
941 moving workspaces across outputs. Coordinates for floating containers are
942 not relative to workspace boundaries, so you must correct their coordinates
943 or those containers will show up in the wrong workspace or not at all.
945 == Using git / sending patches
949 For a short introduction into using git, see
950 http://web.archive.org/web/20121024222556/http://www.spheredev.org/wiki/Git_for_the_lazy
951 or, for more documentation, see http://git-scm.com/documentation
953 Please talk to us before working on new features to see whether they will be
954 accepted. There are a few things which we don’t want to see in i3, e.g. a
955 command which will focus windows in an alt+tab like way.
957 When working on bugfixes, please make sure you mention that you are working on
958 it in the corresponding bugreport at http://bugs.i3wm.org/. In case there is no
959 bugreport yet, please create one.
961 After you are done, please submit your work for review at http://cr.i3wm.org/
963 Do not send emails to the mailing list or any author directly, and don’t submit
964 them in the bugtracker, since all reviews should be done in public at
965 http://cr.i3wm.org/. In order to make your review go as fast as possible, you
966 could have a look at previous reviews and see what the common mistakes are.
968 === Which branch to use?
970 Work on i3 generally happens in two branches: “master” and “next”. Since
971 “master” is what people get when they check out the git repository, its
972 contents are always stable. That is, it contains the source code of the latest
973 release, plus any bugfixes that were applied since that release.
975 New features are only found in the “next” branch. Therefore, if you are working
976 on a new feature, use the “next” branch. If you are working on a bugfix, use
977 the “next” branch, too, but make sure your code also works on “master”.
979 == Thought experiments
981 In this section, we collect thought experiments, so that we don’t forget our
982 thoughts about specific topics. They are not necessary to get into hacking i3,
983 but if you are interested in one of the topics they cover, you should read them
984 before asking us why things are the way they are or why we don’t implement
987 === Using cgroups per workspace
989 cgroups (control groups) are a linux-only feature which provides the ability to
990 group multiple processes. For each group, you can individually set resource
991 limits, like allowed memory usage. Furthermore, and more importantly for our
992 purposes, they serve as a namespace, a label which you can attach to processes
995 One interesting use for cgroups is having one cgroup per workspace (or
996 container, doesn’t really matter). That way, you could set different priorities
997 and have a workspace for important stuff (say, writing a LaTeX document or
998 programming) and a workspace for unimportant background stuff (say,
999 JDownloader). Both tasks can obviously consume a lot of I/O resources, but in
1000 this example it doesn’t really matter if JDownloader unpacks the download a
1001 minute earlier or not. However, your compiler should work as fast as possible.
1002 Having one cgroup per workspace, you would assign more resources to the
1003 programming workspace.
1005 Another interesting feature is that an inherent problem of the workspace
1006 concept could be solved by using cgroups: When starting an application on
1007 workspace 1, then switching to workspace 2, you will get the application’s
1008 window(s) on workspace 2 instead of the one you started it on. This is because
1009 the window manager does not have any mapping between the process it starts (or
1010 gets started in any way) and the window(s) which appear.
1012 Imagine for example using dmenu: The user starts dmenu by pressing Mod+d, dmenu
1013 gets started with PID 3390. The user then decides to launch Firefox, which
1014 takes a long time. So he enters firefox into dmenu and presses enter. Firefox
1015 gets started with PID 4001. When it finally finishes loading, it creates an X11
1016 window and uses MapWindow to make it visible. This is the first time i3
1017 actually gets in touch with Firefox. It decides to map the window, but it has
1018 no way of knowing that this window (even though it has the _NET_WM_PID property
1019 set to 4001) belongs to the dmenu the user started before.
1021 How do cgroups help with this? Well, when pressing Mod+d to launch dmenu, i3
1022 would create a new cgroup, let’s call it i3-3390-1. It launches dmenu in that
1023 cgroup, which gets PID 3390. As before, the user enters firefox and Firefox
1024 gets launched with PID 4001. This time, though, the Firefox process with PID
1025 4001 is *also* member of the cgroup i3-3390-1 (because fork()ing in a cgroup
1026 retains the cgroup property). Therefore, when mapping the window, i3 can look
1027 up in which cgroup the process is and can establish a mapping between the
1028 workspace and the window.
1030 There are multiple problems with this approach:
1032 . Every application has to properly set +_NET_WM_PID+. This is acceptable and
1033 patches can be written for the few applications which don’t set the hint yet.
1034 . It does only work on Linux, since cgroups are a Linux-only feature. Again,
1036 . The main problem is that some applications create X11 windows completely
1037 independent of UNIX processes. An example for this is Chromium (or
1038 gnome-terminal), which, when being started a second time, communicates with
1039 the first process and lets the first process open a new window. Therefore, if
1040 you have a Chromium window on workspace 2 and you are currently working on
1041 workspace 3, starting +chromium+ does not lead to the desired result (the
1042 window will open on workspace 2).
1044 Therefore, my conclusion is that the only proper way of fixing the "window gets
1045 opened on the wrong workspace" problem is in the application itself. Most
1046 modern applications support freedesktop startup-notifications which can be