This patch implements a generalization of the tile layout which adds two attributes (direction and fact) to three areas (global, master, stack). The global area is the entire allocatable visual space and it’s subdivided into the master and stack subareas.

The direction of the global area controls the position of the master area relatively to the stack area and it can be one of DirHor (traditional right stack), DirVer (bottom stack), DirRotHor (left stack) and DirRotVer (top stack). The direction of the master and of the stack areas are independently set and can be one of DirHor and DirVer. This combines to a total of 4*2*2=16 layouts.

The fact numbers indicate the relative size of the first subarea/client along the direction of the considered area (i.e. width for DirHor and DirRotHor and height for DirVer and DirRotVer). A fact of 1 means that the first subarea/client is on par the rest, while a fact of 2 means that its size must double the size of each of the remaining subareas/clients, etc. So the fact for the global area is similar to the traditional mfact in the sense that it manages the relative allocation of visual space between the master and stack subareas, while the fact for the master area stands for the relative importance of the first master client against the rest of masters and, similarly, the fact for the stack area stands for the importance of the first slave client in relation to the rest of slaves.

xtile adds two new commands to dwm: setdir and setfact (which supersedes setmfact). Both commands take an array of three values (of type int for setdir and float for setfact), one value for each area (the first one for the global area, the second one for the master area and the third one for the stack area). If you pass the value v as INC(v) it will be taken as a relative increment to be added to the current value, otherwise it will be taken as an absolute value. Usually the resulting value will be truncated to the valid range of values for each area/attribute combination, but relative increments for directions wrap around the limits of the valid range. Notice that INC(0) means “do nothing here”, so it gives you a way to easily modify the value for some area while leaving the rest untouched.

Default key bindings

The areas are selected by modifiers as follows:

Modifier Area
MODKEY+Shift Master
MODKEY+Control Stack
MODKEY+Shift+Control All three areas simultaneously

Each of the modifiers then combines with each of the following keys up to a total of 4*3=12 key bindings:

Key Function
r Rotate direction
h Decrement fact by 10%.
l Increment fact by 10%.

There are two provided default “presets” or “schemas” also:

Modifier Key Preset
MODKEY+Shift t Right stack
MODKEY+Control t Bottom stack

These presets allow to quickly switch between different no-nonsense tilings avoiding the need to rotate through all the nonsense combinations in-between. But notice that MODKEY+Shift+Control+r (i.e. simultaneously rotate all three areas) usually produces sensible layouts (due to the way directions were designed to rotate).

You can also easily define your own presets by calling setdir and setfact as needed. For example, here is the configuration code for the default presets described above:

{ MODKEY|ShiftMask,   XK_t, setdirs, {.v = (int[]){ DirHor, DirVer, DirVer } } },
{ MODKEY|ControlMask, XK_t, setdirs, {.v = (int[]){ DirVer, DirHor, DirHor } } },

Layout symbol

The layout symbol will probably look cryptic at first sight but it’s very easily decoded. It consists of three characters, one for the direction of each area:

For example, ‘<||’ stands for the default right stack tile provided by dwm and ‘^–’ stands for bstack (as defined by the bottom stack patch).


Why facts per area?

There is some arbitrariness in the way facts are defined by xtile: why facts for the first master and the first slave and not, say, for the first two clients instead? Considering that most real life layouts will have one or two masters and a variable number of slaves, the road xtile took will enable the user to effectively control the relative size of the three/four most important clients in a very intuitive way that built on his previous understanding of the mfact and the master and stack area concepts. OTOH it’s not clear to me how to allow the specification of facts for the first two clients in an intuitive way:

Why not deck area?

One obvious additional generalization would have been to extrapolate the nmaster idea to all three areas, or at least to the stack area. So if you allowed only m masters and n slaves you would end up with m+n tiled windows and with the rest of the clients in the current tagset stacked or decked “below” the last tiled client. flextile, clients-per-tag and deck patches provide variations on this kind of layout. I’ve also implemented a version of xtile that supports it and even subsumes monocle, but I think this promotes a bad pattern of usage. Coupled with stack manipulation operations as the ones provided by the stacker or push patches, there is the temptation to manage visibility by moving the desired clients in the current tagset to the first n+m visible positions of the focus stack (not to be confused with the stack area). There are a number of problems with this approach:

Fortunately, dwm provides a much better mechanism to restrict visibility: tags. IMO there is no need to provide a half-assed alternative to one of dwm’s strongest selling points.

Other patches

Recommended complementary patches:

Mandatory dependencies:

Related patches: bottom stack, flextile, cfacts, stackmfact.