Unity – Script to Open Application Window on Specific Viewport and Position

unitywindowwindow-managementwindow-managerworkspaces

So, I have 8 virtual Desktops in Unity (with Compiz) because I have many projects I'm working on simultaneously.

Problem is, every time I need to reboot or accidentally close Chrome (which constitutes a large part of the windows I need for work) I have to manually go an open those windows again and then set them up (open files, go to the correct urls etc.).

How would you go about writing a script that will do all that for me? That is: 1) Open windows 2) Put them into the correct coordinates at correct virtual screens

(1) is obvious, for Google Chrome you just run 'google-chrome'. But then how do you put it in the right place? (2)

Or is there a script/software that already exists that would do it for me?

Best Answer

It can be done very well, but you need some understanding on Unity/viewports. I hope the story below is understandable, if not, please leave a comment.

The script below can be used to open a window of any application on any of your viewports, on any position, if you run it with the right arguments. The script is an edited version of this one, but now prepared to place windows on the spanning virtual desktop.

1. Understanding viewports and window coordinates

Workspaces in Unity

In Unity, unlike other window managers, you actually only have one spanning workspace, which is divided into viewports. In your case, your workspace is divided into eight viewports.

How the position of the windows is defined

The window position, as the output of the command:

wmctrl -lG
(you need to have wmctrl installed to run the command)

is described as the position, relative to the upper left corner of the current viewport:

So if you are on viewport 1:
a window on viewport 2 to could be positioned on e.g. 1700 (x-wise) x 500 (y-wise)
(my screen is 1680x1050)

enter image description here

However, if you are on viewport 6:
the same window would be positioned on 20 (x), -550 (y) enter image description here

Using these coordinates correctly is important to run the script with the right arguments, as described below:

2. How to use the script

The script below can be used to place a new window of an application on your virtual (spanning) workspace.

Make sure wmctrl is installed:
```
sudo apt-get install wmctrl
```
Copy the script below into an empty file, save it as setwindow (no extension) in ~/bin. Create the directory if it doesn't exist yet. Make the script executable.
If you just created ~/bin, either run the command source ~/.profile or log out/in to make the directory available in $PATH.

Test run the command:

setwindow <application> <x_position> <y_position> <horizontal_size> <vertical_size>

e.g.

setwindow gedit 100 100 200 200

A gedit window should show up on the current viewport.

Notes:

Keep in mind that not all applications allow window sizes below a certain width or height. The minimum width of a gedit window on my system is e.g. appr. 470 px.
The script only works fine if the whole window fits on the targeted viewport, choose your coordinates/sizes accordingly. Also mind that the Unity Launcher and the panel use some space (!) which can influence the position of the window.
Use negative <x_position> to place windows on the left of the current viewport(s)
Use negative <y_position> to place windows above the current viewport(s)
To open new windows on different viewports at once, you can simply chain commands. Looking at the viewport setup in the "Long story" example, If I am on viewport 1, I can open gedit windows on viewport 1, 2, 3 and 4 with the command:
```
setwindow gedit 100 100 200 200&&setwindow gedit 1780 100 200 200&&setwindow gedit 3460 100 200 200&&setwindow gedit 5140 100 200 200
```

The script

#!/usr/bin/env python3
import subprocess
import time
import sys

app = sys.argv[1]

get = lambda x: subprocess.check_output(["/bin/bash", "-c", x]).decode("utf-8")
ws1 = get("wmctrl -lp"); t = 0
subprocess.Popen(["/bin/bash", "-c", app])
# fix exception for Chrome (command = google-chrome-stable, but processname = chrome)
app = "chrome" if "chrome" in app else app
while t < 30:      
    ws2 = [w.split()[0:3] for w in get("wmctrl -lp").splitlines() if not w in ws1]
    procs = [[(p, w[0]) for p in get("ps -e ww").splitlines() \
              if app in p and w[2] in p] for w in ws2]
    if len(procs) > 0:
        w_id = procs[0][0][1]
        cmd1 = "wmctrl -ir "+w_id+" -b remove,maximized_horz"
        cmd2 = "wmctrl -ir "+w_id+" -b remove,maximized_vert"
        cmd3 = "wmctrl -ir "+procs[0][0][1]+" -e 0,"+sys.argv[2]+","+sys.argv[3]+","+sys.argv[4]+","+sys.argv[5]
        for cmd in [cmd1, cmd2, cmd3]:   
            subprocess.call(["/bin/bash", "-c", cmd])
        break
    time.sleep(0.5)
    t = t+1

EDIT: the lazy version

In case you'd prefer to just enter coordinates and size, simply as if you would open a window on the current viewport, and give the targeted viewport as an argument (without having to calculate anything), then use the version below...

If you set it up like the first version of the script, you can run it with the command:

setwindow <application> <x_position> <y_position> <horizontal_size> <vertical_size> <targeted_viewport>

An example: to open a Google-Chrome window positioned on 20, 20, size 300x300, on viewport 5:

setwindow google-chrome 20 20 300 300 5

The setup is pretty much the same as the first version of the script.
Note that also this script only works correctly if the defined window (position/size) fits completely within the targeted viewport.

The script:

#!/usr/bin/env python3
import subprocess
import time
import sys

app = sys.argv[1]
target_vp = int(sys.argv[6])

def get_res():
    # get resolution
    xr = subprocess.check_output(["xrandr"]).decode("utf-8").split()
    pos = xr.index("current")
    return [int(xr[pos+1]), int(xr[pos+3].replace(",", "") )]

res = get_res()

def current(set_vp):
    # get the current viewport
    vp_data = subprocess.check_output(
        ["wmctrl", "-d"]
        ).decode("utf-8").split()
    dt = [int(n) for n in vp_data[3].split("x")]
    cols = int(dt[0]/res[0])
    rows = int(dt[1]/res[1])    
    curr_vpdata = [int(n) for n in vp_data[5].split(",")]
    curr_col = int(curr_vpdata[0]/res[0])
    curr_row = int(curr_vpdata[1]/res[1])
    curr_vp = curr_col+curr_row*cols+1
    # calculate the vector to the origin from the current viewport (in resolution units)
    vec_curr = vector(curr_vp, cols)
    # calculate the vector to the origin from the targeted viewport
    vec_set = vector(set_vp, cols)
    # calculate the vector between current and targeted viewport
    vec_relative = [vec_set[0] - vec_curr[0],
                    vec_set[1] - vec_curr[1]]
    # calculate needed correction (absolute)
    relative = [vec_relative[0]*res[0],
                vec_relative[1]*res[1]]
    return relative

def vector(vp, cols):
    rem = vp%cols
    vec_x = rem-1 if rem != 0 else cols-1
    vec_y = int((vp-1)/cols)
    return [vec_x, vec_y]

res = get_res() # nieuw
get = lambda x: subprocess.check_output(["/bin/bash", "-c", x]).decode("utf-8")
ws1 = get("wmctrl -lp"); t = 0
# check for additional arguments to run the application
try:
    subprocess.Popen(["/bin/bash", "-c", app+" "+sys.argv[7]])  
except IndexError:
    subprocess.Popen(["/bin/bash", "-c", app])

# fix exception for Chrome (command = google-chrome-stable, but processname = chrome)
app = "chrome" if "chrome" in app else app
while t < 30:      
    ws2 = [w.split()[0:3] for w in get("wmctrl -lp").splitlines() if not w in ws1]
    procs = [[(p, w[0]) for p in get("ps -e ww").splitlines() \
              if app in p and w[2] in p] for w in ws2]
    if len(procs) > 0:
        w_id = procs[0][0][1]
        cmd1 = "wmctrl -ir "+w_id+" -b remove,maximized_horz"
        cmd2 = "wmctrl -ir "+w_id+" -b remove,maximized_vert"
        # calculate the correction, related to the current workspace, marge for launcher and panel
        pos_x = int(sys.argv[2]); pos_y = int(sys.argv[3]); x_marge = 65; y_marge = 35
        pos_x = pos_x if pos_x > x_marge else x_marge; pos_y = pos_y if pos_y > y_marge else y_marge
        x_relative = pos_x+current(target_vp)[0]
        y_relative = pos_y+current(target_vp)[1]
        # correct possible inaccurately set width / height
        x_size = res[0]; y_size = res[1]
        set_width = int(sys.argv[4]); set_height = int(sys.argv[5])
        width = set_width if set_width+x_marge+pos_x < x_size else x_size - pos_x - x_marge
        height = set_height if set_height+y_marge+pos_y < y_size else y_size - pos_y - y_marge
        cmd3 = "wmctrl -ir "+w_id+" -e 0,"+str(x_relative)+","+str(y_relative)+","+str(width)+","+str(height)
        for cmd in [cmd1, cmd2, cmd3]:   
            subprocess.call(["/bin/bash", "-c", cmd])
        break
    time.sleep(0.5)
    t = t+1

Opening application windows with arguments

To finish the job, answering your question completely:

If you run the script as e.g.:

setwindow google-chrome 20 20 300 300 5

it will open a default window on the targeted desktop(s).
With the latest version of the script however, you can add an additional argument to open the application window, for example a url:

setwindow <application> <x_position> <y_position> <horizontal_size> <vertical_size> <targeted_viewport> <(optional)_argument>

e.g.:

setwindow google-chrome 0 0 600 600 3 "--new-window http://askubuntu.com"

If the (extra) argument contains spaces, use quotes. The above example will open agoogle-chrome window on viewport 3, opening the url http://askubuntu.com.

You can chain commands to open multiple windows/urls on different workspaces in one command, e.g.:

setwindow google-chrome 0 0 600 600 8 "--new-window http://askubuntu.com"&&setwindow google-chrome 0 0 600 600 7 "--new-window www.google.com"

Using the wmctrl command on Unity

The wmctrl command shows some peculiarities when used to move windows to- or very nearby the launcher or the panel. Therefore the margins:

left_margin = 70; top_margin = 30

cannot be set to zero. You have to keep at least a few px distance to both the panel and the launcher. I'd suggest leaving both margins- values as they are. All other values, padding, columns and rows you can play around with and set it as you like.

The script

#!/usr/bin/env python3
import subprocess
import getpass
import sys

#--- set your preferences below: columns, rows, padding between windows, margin(s)
cols = 2; rows = 2; padding = 10; left_margin = 70; top_margin = 30
#---

get = lambda cmd: subprocess.check_output(["/bin/bash", "-c", cmd]).decode("utf-8")
def get_res():
    xr = get("xrandr").split(); pos = xr.index("current")
    return [int(xr[pos+1]), int(xr[pos+3].replace(",", "") )]

# get resolution
res = get_res()
# define (calculate) the area to divide
area_h = res[0] - left_margin; area_v = res[1] - top_margin
# create a list of calculated coordinates
x_coords = [int(left_margin+area_h/cols*n) for n in range(cols)]
y_coords = [int(top_margin+area_v/rows*n) for n in range(rows)]
coords = sum([[(cx, cy) for cx in x_coords] for cy in y_coords], [])
# calculate the corresponding window size, given the padding, margins, columns and rows
w_size = [str(int(area_h/cols - padding)), str(int(area_v/rows - padding))]
# find windows of the application, identified by their pid
pids = [p.split()[0] for p in get("ps -e").splitlines() if sys.argv[1] in p]
w_list = sum([[w.split()[0] for w in get("wmctrl -lp").splitlines() if p in w] for p in pids], [])
print(pids, w_list, coords)
# remove possibly maximization, move the windows
for n, w in enumerate(w_list):
    data = (",").join([str(item) for item in coords[n]])+","+(",").join(w_size)
    cmd1 = "wmctrl -ir "+w+" -b remove,maximized_horz"
    cmd2 = "wmctrl -ir "+w+" -b remove,maximized_vert"
    cmd3 = "wmctrl -ir "+w+" -e 0,"+data
    for cmd in [cmd1, cmd2, cmd3]:
        subprocess.Popen(["/bin/bash", "-c", cmd])

How to use

Make sure wmctrl is installed :)
Copy thye script into an empty file, save it as rearrange_windows.py
In the head section of the script, set your preferences

Run it by the command:

python3 /path/to/rearrange_windows.py <application>

example: to rearrange chromium windows:

python3 /path/to/rearrange_windows.py chromium

to rearrange chrome windows

python3 /path/to/rearrange_windows.py chrome

Note

The script can be used to put windows of any application into a grid, since the process name of the application is used as an argument.

EDIT

Dynamic version

below a dynamic version of the script, as requested in a comment. This version of the script calculates the number of columns and rows, depending on the number of windows. The proportions of the rearranged window(s) is similar to the proportions of the screen.

The setup and the use is pretty much the same as the version above, only the number of columns and rows is now set automatically.

#!/usr/bin/env python3
import subprocess
import getpass
import sys
import math

#--- set your preferences below: padding between windows, margin(s)
padding = 10; left_margin = 70; top_margin = 30
#---

get = lambda cmd: subprocess.check_output(["/bin/bash", "-c", cmd]).decode("utf-8")
def get_res():
    xr = get("xrandr").split(); pos = xr.index("current")
    return [int(xr[pos+1]), int(xr[pos+3].replace(",", "") )]

# find windows of the application, identified by their pid
pids = [p.split()[0] for p in get("ps -e").splitlines() if sys.argv[1] in p]
w_list = sum([[w.split()[0] for w in get("wmctrl -lp").splitlines() if p in w] for p in pids], [])
# calculate the columns/rows, depending on the number of windows
cols = math.ceil(math.sqrt(len(w_list))); rows = cols
# define (calculate) the area to divide
res = get_res()
area_h = res[0] - left_margin; area_v = res[1] - top_margin
# create a list of calculated coordinates
x_coords = [int(left_margin+area_h/cols*n) for n in range(cols)]
y_coords = [int(top_margin+area_v/rows*n) for n in range(rows)]
coords = sum([[(cx, cy) for cx in x_coords] for cy in y_coords], [])
# calculate the corresponding window size, given the padding, margins, columns and rows
if cols != 0:
    w_size = [str(int(area_h/cols - padding)), str(int(area_v/rows - padding))]
# remove possibly maximization, move the windows
for n, w in enumerate(w_list):
    data = (",").join([str(item) for item in coords[n]])+","+(",").join(w_size)
    cmd1 = "wmctrl -ir "+w+" -b remove,maximized_horz"
    cmd2 = "wmctrl -ir "+w+" -b remove,maximized_vert"
    cmd3 = "wmctrl -ir "+w+" -e 0,"+data
    for cmd in [cmd1, cmd2, cmd3]:
        subprocess.call(["/bin/bash", "-c", cmd])

See below the examples with a varying number of opened windows:

enter image description here

Explanation (second script)

Finding the specific windows

The command:
```
wmctrl -lp
```
lists all windows, in the format:
```
0x19c00085  0 14838  jacob-System-Product-Name *Niet-opgeslagen document 1 - gedit
```
where the first column is the window's unique id, and the third column is the pid of the application that owns the window.
The command:
```
ps -e
```
lists all processes, in the format:
```
14838 ?        00:00:02 gedit
```
where the first column is the application's pid, the last one is the process name.

By comparing these two lists, we can find all windows (id of-) which belong to a specific application (called w_list in the script, as the result of line 17/18 in the script):

pids = [p.split()[0] for p in get("ps -e").splitlines() if sys.argv[1] in p]
w_list = sum([[w.split()[0] for w in get("wmctrl -lp").splitlines() if p in w] for p in pids], [])

Calculating the number of rows/columns

If we make the windows of the same proportions as the screen, that means the number of columns is equal to the number of rows.
That implies that both the number of colums and rows are equal to the rounded up square root of the number of windows to rearrange. That is done in line 20:
```
cols = math.ceil(math.sqrt(len(w_list))); rows = cols
```

Calculating the window size and position

Once we have the number of columns, all we need to do is divide the available area (screen resolution - left margin/top margin) in the columns/rows and we have the targeted window size, which is then diminished by the padding, as set in the head of the script:
```
w_size = [str(int(area_h/cols - padding)), str(int(area_v/rows - padding))]
```
The horizontal (x) positions are the result of the product(s) of the horizontal window size (including padding) times the column number, in a range of the number of columns. for example: if I have 3 colums of 300 px, the resulting x-positions are:
```
[0, 300, 600]
```

The vertical (y) positions are calculated likewise. Both lists are then combined into a list of coordinates, in which the windows will be rearranged.

This is done in line 26-28 of the script:

x_coords = [int(left_margin+area_h/cols*n) for n in range(cols)]
y_coords = [int(top_margin+area_v/rows*n) for n in range(rows)]
coords = sum([[(cx, cy) for cx in x_coords] for cy in y_coords], [])

The actual rearranging finally (after unmaximizing possibly maximized windows) is done from line 33 and further.

Best Answer

1. Understanding viewports and window coordinates

Workspaces in Unity

How the position of the windows is defined

2. How to use the script

The script

EDIT: the lazy version

The script:

Opening application windows with arguments

Related Solutions

Window Management – Open Application in Specific Workspace

How to Find and Grid Windows of a Specific Application – Bash Scripting

Using the wmctrl command on Unity

The script

How to use

Note

EDIT

Dynamic version

Explanation (second script)

Related Question