Hacking i3: How To ================== Michael Stapelberg May 2009 This document is intended to be the first thing you read before looking and/or touching i3’s source code. It should contain all important information to help you understand why things are like they are. If it does not mention something you find necessary, please do not hesitate to contact me. == Window Managers A window manager is not necessarily needed to run X, but it is usually used in combination to facilitate some things. The window manager's job is to take care of the placement of windows, to provide the user some mechanisms to change the position/size of windows and to communicate with clients to a certain extent (for example handle fullscreen requests of clients such as MPlayer). There are no different contexts in which X11 clients run, so a window manager is just another client, like all other X11 applications. However, it handles some events which normal clients usually don’t handle. In the case of i3, the tasks (and order of them) are the following: . Grab the key bindings (events will be sent upon keypress/keyrelease) . Iterate through all existing windows (if the window manager is not started as the first client of X) and manage them (= reparent them, create window decorations) . When new windows are created, manage them . Handle the client’s `_WM_STATE` property, but only the `_WM_STATE_FULLSCREEN` . Handle the client’s `WM_NAME` property . Handle the client’s size hints to display them proportionally . Handle enter notifications (focus follows mouse) . Handle button (as in mouse buttons) presses for focus/raise on click . Handle expose events to re-draw own windows such as decorations . React to the user’s commands: Change focus, Move windows, Switch workspaces, Change the layout mode of a container (default/stacking), Start a new application, Restart the window manager In the following chapters, each of these tasks and their implementation details will be discussed. === Tiling window managers Traditionally, there are two approaches to managing windows: The most common one nowadays is floating, which means the user can freely move/resize the windows. The other approach is called tiling, which means that your window manager distributing windows to use as much space as possible while not overlapping. The idea behind tiling is that you should not need to waste your time moving/resizing windows while you usually want to get some work done. After all, most users sooner or later tend to lay out their windows in a way which corresponds to tiling or stacking mode in i3. Therefore, why not let i3 do this for you? Certainly, it’s faster than you could ever do it. The problem with most tiling window managers is that they are too unflexible. In my opinion, a window manager is just another tool, and similar to vim which can edit all kinds of text files (like source code, HTML, …) and is not limited to a specific file type, a window manager should not limit itself to a certain layout (like dwm, awesome, …) but provide mechanisms for you to easily create the layout you need at the moment. === The layout table To accomplish flexible layouts, we decided to simply use a table. The table grows and shrinks as you need it. Each cell holds a container which then holds windows (see picture below). You can use different layouts for each container (default layout and stacking layout). So, when you open a terminal and immediately open another one, they reside in the same container, in default layout. The layout table has exactly one column, one row and therefore one cell. When you move one of the terminals to the right, the table needs to grow. It will be expanded to two columns and one row. This enables you to have different layouts for each container. The table then looks like this: [width="15%",cols="^,^"] |======== | T1 | T2 |======== When moving terminal 2 to the bottom, the table will be expanded again. [width="15%",cols="^,^"] |======== | T1 | | | T2 |======== You can really think of the layout table like a traditional HTML table, if you’ve ever designed one. Especially col- and rowspan work equally. Below you see an example of colspan=2 for the first container (which has T1 as window). [width="15%",cols="^asciidoc"] |======== | T1 | [cols="^,^",frame="none"] !======== ! T2 ! T3 !======== |======== Furthermore, you can freely resize table cells. == Files include/data.h:: Contains data definitions used by nearly all files. You really need to read this first. include/*.h:: Contains forward definitions for all public functions, aswell as doxygen-compatible comments (so if you want to get a bit more of the big picture, either browse all header files or use doxygen if you prefer that). src/client.c:: Contains all functions which are specific to a certain client (make it fullscreen, see if its class/name matches a pattern, kill it, …). src/commands.c:: Parsing commands and actually execute them (focussing, moving, …). src/config.c:: Parses the configuration file. src/debug.c:: Contains debugging functions to print unhandled X events. src/floating.c:: Contains functions for floating mode (mostly resizing/dragging). src/handlers.c:: Contains all handlers for all kind of X events (new window title, new hints, unmapping, key presses, button presses, …). src/layout.c:: Renders your layout (screens, workspaces, containers). src/mainx.c:: Initializes the window manager. src/manage.c:: Looks at existing or new windows and decides whether to manage them. If so, it reparents the window and inserts it into our data structures. src/resize.c:: Contains the functions to resize columns/rows in the table. src/resize.c:: Contains the functions to resize columns/rows in the table. src/table.c:: Manages the most important internal data structure, the design table. src/util.c:: Contains useful functions which are not really dependant on anything. src/xcb.c:: Contains wrappers to use xcb more easily. src/xinerama.c:: (Re-)initializes the available screens and converts them to virtual screens (see below). == Data structures See include/data.h for documented data structures. The most important ones are explained right here. image:bigpicture.png[The Big Picture] So, the hierarchy is: . *Virtual screens* (Screen 0 in this example) . *Workspaces* (Workspace 1 in this example) . *Table* (There can only be one table per Workspace) . *Container* (left and right in this example) . *Client* (The two clients in the left container) === Virtual screens A virtual screen (type `i3Screen`) is generated from the connected screens obtained through Xinerama. The difference to the raw Xinerama monitors as seen when using +xrandr(1)+ is that it falls back to the lowest common resolution of the logical screens. For example, if your notebook has 1280x800 and you connect a video projector with 1024x768, set up in clone mode (+xrandr \--output VGA \--mode 1024x768 \--same-as LVDS+), i3 will have one virtual screen. However, if you configure it using +xrandr \--output VGA \--mode 1024x768 \--right-of LVDS+, i3 will generate two virtual screens. For each virtual screen, a new workspace will be assigned. New workspaces are created on the screen you are currently on. === Workspace A workspace is identified by its number. Basically, you could think of workspaces as different desks in your bureau, if you like the desktop methaphor. They just contain different sets of windows and are completely separate of each other. Other window managers also call this ``Virtual desktops''. === The layout table Each workspace has a table, which is just a two-dimensional dynamic array containing Containers (see below). This table grows and shrinks as you need it (by moving windows to the right you can create a new column in the table, by moving them to the bottom you create a new row). === Container A container is the content of a table’s cell. It holds an arbitrary amount of windows and has a specific layout (default layout or stack layout). Containers can consume multiple table cells by modifying their colspan/rowspan attribute. === Client A client is x11-speak for a window. == List/queue macros i3 makes heavy use of the list macros defined in BSD operating systems. To ensure that the operating system on which i3 is compiled has all the awaited features, i3 comes with `include/queue.h`. On BSD systems, you can use man `queue(3)`. On Linux, you have to use google. The lists used are `SLIST` (single linked lists) and `CIRCLEQ` (circular queues). Usually, only forward traversal is necessary, so an `SLIST` works fine. However, for the windows inside a container, a `CIRCLEQ` is necessary to go from the currently selected window to the window above/below. == Naming conventions There is a row of standard variables used in many events. The following names should be chosen for those: * ``conn'' is the xcb_connection_t * ``event'' is the event of the particular type * ``container'' names a container * ``client'' names a client, for example when using a +CIRCLEQ_FOREACH+ == Startup (src/mainx.c, main()) * Establish the xcb connection * Check for XKB extension on the separate X connection * Check for Xinerama screens * Grab the keycodes for which bindings exist * Manage all existing windows * Enter the event loop == Keybindings === Grabbing the bindings Grabbing the bindings is quite straight-forward. You pass X your combination of modifiers and the keycode you want to grab and whether you want to grab them actively or passively. Most bindings (everything except for bindings using Mode_switch) are grabbed passively, that is, just the window manager gets the event and cannot replay it. We need to grab bindings that use Mode_switch actively because of a bug in X. When the window manager receives the keypress/keyrelease event for an actively grabbed keycode, it has to decide what to do with this event: It can either replay it so that other applications get it or it can prevent other applications from receiving it. So, why do we need to grab keycodes actively? Because X does not set the state-property of keypress/keyrelease events properly. The Mode_switch bit is not set and we need to get it using XkbGetState. This means we cannot pass X our combination of modifiers containing Mode_switch when grabbing the key and therefore need to grab the keycode itself without any modiffiers. This means, if you bind Mode_switch + keycode 38 ("a"), i3 will grab keycode 38 ("a") and check on each press of "a" if the Mode_switch bit is set using XKB. If yes, it will handle the event, if not, it will replay the event. === Handling a keypress As mentioned in "Grabbing the bindings", upon a keypress event, i3 first gets the correct state. Then, it looks through all bindings and gets the one which matches the received event. The bound command is parsed directly in command mode. == Manage windows (src/mainx.c, manage_window() and reparent_window()) `manage_window()` does some checks to decide whether the window should be managed at all: * Windows have to be mapped, that is, visible on screen * The override_redirect must not be set. Windows with override_redirect shall not be managed by a window manager Afterwards, i3 gets the intial geometry and reparents the window if it wasn’t already managed. Reparenting means that for each window which is reparented, a new window, slightly larger than the original one, is created. The original window is then reparented to the bigger one (called "frame"). After reparenting, the window type (`_NET_WM_WINDOW_TYPE`) is checked to see whether this window is a dock (`_NET_WM_WINDOW_TYPE_DOCK`), like dzen2 for example. Docks are handled differently, they don’t have decorations and are not assigned to a specific container. Instead, they are positioned at the bottom of the screen. To get the height which needsd to be reserved for the window, the `_NET_WM_STRUT_PARTIAL` property is used. == What happens when an application is started? i3 does not care for applications. All it notices is when new windows are mapped (see `src/handlers.c`, `handle_map_notify_event()`). The window is then reparented (see section "Manage windows"). After reparenting the window, `render_layout()` is called which renders the internal layout table. The window was placed in the currently focused container and therefore the new window and the old windows (if any) need te be moved/resized so that the currently active layout (default mode/stacking mode) is rendered correctly. To move/resize windows, a window is ``configured'' in X11-speak. Some applications, such as MPlayer obivously assume the window manager is stupid and therefore configure their windows by themselves. This generates an event called configurenotify. i3 handles these events and pushes the window back to its position/size. == _NET_WM_STATE Only the _NET_WM_STATE_FULLSCREEN atom is handled. It calls ``toggle_fullscreen()'' for the specific client which just configures the client to use the whole screen on which it currently is. Also, it is set as fullscreen_client for the i3Screen. == WM_NAME When the WM_NAME property of a window changes, its decoration (containing the title) is re-rendered. == Size hints Size hints specify the minimum/maximum size for a given window aswell as its aspect ratio. At the moment, as i3 does not have a floating mode yet, only the aspect ratio is parsed. This is important for clients like mplayer, who only set the aspect ratio and resize their window to be as small as possible (but only with some video outputs, for example in Xv, while when using x11, mplayer does the necessary centering for itself). So, when an aspect ratio was specified, i3 adjusts the height of the window until the size maintains the correct aspect ratio. For the code to do this, see src/layout.c, function resize_client(). == Rendering (src/layout.c, render_layout() and render_container()) There are two entry points to rendering: render_layout() and render_container(). The former one renders all virtual screens, the currently active workspace of each virtual screen and all containers (inside the table cells) of these workspaces using render_container(). Therefore, if you need to render only a single container, for example because a window was removed, added or changed its title, you should directly call render_container(). Rendering consists of two steps: In the first one, in render_layout(), each container gets its position (screen offset + offset in the table) and size (container's width times colspan/rowspan). Then, render_container() is called: render_container() then takes different approaches, depending on the mode the container is in. === Common parts On the frame (the window which was created around the client’s window for the decorations), a black rectangle is drawn as a background for windows like MPlayer, which don’t completely fit into the frame. === Default mode Each clients gets the container’s width and an equal amount of height. === Stack mode In stack mode, a window containing the decorations of all windows inside the container is placed at the top. The currently focused window is then given the whole remaining space. === Window decorations The window decorations consist of a rectangle in the appropriate color (depends on whether this window is the currently focused one or the last focused one in a not focused container or not focused at all) forming the background. Afterwards, two lighter lines are drawn and the last step is drawing the window’s title (see WM_NAME) onto it. === Fullscreen windows For fullscreen windows, the `rect` (x, y, width, height) is not changed to allow the client to easily go back to its previous position. Instead, fullscreen windows are skipped when rendering. === Resizing containers By clicking and dragging the border of a container, you can resize the whole column (respectively row) which this container is in. This is necessary to keep the table layout working and consistent. Currently, only vertical resizing is implemented. The resizing works similarly to the resizing of floating windows or movement of floating windows: * A new, invisible window with the size of the root window is created (+grabwin+) * Another window, 2px width and as high as your screen (or vice versa for horizontal resizing) is created. Its background color is the border color and it is only there to signalize the user how big the container will be (it creates the impression of dragging the border out of the container). * The +drag_pointer+ function of +src/floating.c+ is called to grab the pointer and enter an own event loop which will pass all events (expose events) but motion notify events. This function then calls the specified callback (+resize_callback+) which does some boundary checking and moves the helper window. As soon as the mouse button is released, this loop will be terminated. * The new width_factor for each involved column (respectively row) will be calculated. == User commands / commandmode (src/commands.c) Like in vim, you can control i3 using commands. They are intended to be a powerful alternative to lots of shortcuts, because they can be combined. There are a few special commands, which are the following: exec :: Starts the given command by passing it to `/bin/sh`. restart:: Restarts i3 by executing `argv[0]` (the path with which you started i3) without forking. w:: "With". This is used to select a bunch of windows. Currently, only selecting the whole container in which the window is in, is supported by specifying "w". f, s, d:: Toggle fullscreen, stacking, default mode for the current window/container. The other commands are to be combined with a direction. The directions are h, j, k and l, like in vim (h = left, j = down, k = up, l = right). When you just specify the direction keys, i3 will move the focus in that direction. You can provide "m" or "s" before the direction to move a window respectively or snap. == Gotchas * Forgetting to call `xcb_flush(conn);` after sending a request. This usually leads to code which looks like it works fine but which does not work under certain conditions. == Using git / sending patches For a short introduction into using git, see http://www.spheredev.org/wiki/Git_for_the_lazy or, for more documentation, see http://git-scm.com/documentation When you want to send a patch because you fixed a bug or implemented a cool feature (please talk to us before working on features to see whether they are maybe already implemented, not possible because of some reason or don’t fit into the concept), please use git to create a patchfile. First of all, update your working copy to the latest version of the master branch: -------- git pull -------- Afterwards, make the necessary changes for your bugfix/feature. Then, review the changes using +git diff+ (you might want to enable colors in the diff using +git config diff.color auto+). When you are definitely done, use +git commit -a+ to commit all changes you’ve made. Then, use the following command to generate a patchfile which we can directly apply to the branch, preserving your commit message and name: ----------------------- git format-patch origin ----------------------- Just send us the generated file via mail.