Window placement in floating mode

[bklaase]

Current strategy seems to be to place windows in the top left of the current monitor. While it was the easiest thing in the world to create a hook, and subsequently a script that gets triggered on windows with certain conditions (i.e. no size hints for location), that centres the window, this has a disadvantage. When I do this there is an almost unnoticeable movement before the window moves to the centre of the monitor. With a compton fade effect not so much noticable, without it, it looks quite clunky. Would it be possible (maybe in next release) to implement placement strategy natively in the WM ?

Possible values could be:

• Centre (or center, I never know.. [1])
• Corner (all of them)
• Smart (I never use this, and never will :) )

[blairdrummond]

I found that this centers the windows in floating mode, and they behave appropriately when dragged. I don't know this code very well though, so perhaps this will cause other issues. If someone implemented a way to specify "floating_mode (center|corner)" as you suggest, then I think that this might be a really simple fix.

// clientlist.cpp
void client_resize_floating(HSClient* client, HSMonitor* m) {
    // ...
    Rectangle rect = client->float_size;

    + rect.x += (m->rect.width - rect.width)/2;
    + rect.y += (m->rect.height - rect.height)/2;

    rect.x += m->rect.x;
    rect.x += m->rect.y;
    // ...

I have tested this only on a basic one monitor setup.

[t-wissmann]

Right now, hlwm positions a new floating window the way the window itselfs suggests it. That's why terminal windows are placed in the top right corner and other windows in the monitor center.

But it's a todo for me to have this overwritable by rules.

[bklaase]

Alright, thank you both for replying. :) I might have a go at this myself and make a pull request.

[bklaase]

@blairdrummond, because of your post I was motivated to try this solution. It does work somewhat in a one monitor scenario, but resizing starts behaving very weirdly (resizing from window center, instead of upper left corner, etc). In a multi monitor setup, it only functions on the primary monitor.

I did however find what I think is the way to fix this. Perhaps @t-wissmann, can spot some suble mistake, but I run this now on a multi-monitor setup, without any problems

I place it here, so that people annoyed by this behaviour can patch the 7.0 release and move on with their lives, whilst waiting for WinterBreeze

--- a/src/clientlist.cpp
+++ b/src/clientlist.cpp
@@ -239,9 +239,27 @@ HSClient* manage_client(Window win) {
     int x, y;
     unsigned int w, h;
     XGetGeometry(g_display, win, &root_win, &x, &y, &w, &h, &border, &depth);
-    // treat wanted coordinates as floating coords
-    client->float_size.x = x;
-    client->float_size.y = y;
+    // treat wanted coordinates as floating coords but center otherwise
+
+       //monitor dims
+       int mw = m->rect.width;
+       int mh = m->rect.height;
+       //padding values
+       int pl = m->pad_left;
+       int pr = m->pad_right;
+       int pu = m->pad_up;
+       int pd = m->pad_down;
+
+       if(x == 0 && w <= mw)
+               client->float_size.x = (mw-(pl+pr)-w)/2;
+       else
+               client->float_size.x = x;
+
+       if(y == 0 && h <= mh)
+               client->float_size.y = (mh-(pu+pd)-h)/2;
+       else
+               client->float_size.y = y;
+
     client->float_size.width = w;
     client->float_size.height = h;
     client->last_size = client->float_size;

[imtbl]

But it's a todo for me to have this overwritable by rules.

This issue is quite old and I can't find anything in the docs. Are there any updates on this?

I open quite a few CLI programs via click actions in my Polybar:

I've set a rule: hc rule class='float' floating=on
And the Polybar click action: click-left = kitty --class float -e htop

Right now they all open in the top left corner (due to them being launched inside a terminal emulator, I assume).

So I also think being able to set rules for the placement (e.g., hc rule class='float' placement=center) would be great.

[t-wissmann]

But it's a todo for me to have this overwritable by rules.

This issue is quite old and I can't find anything in the docs. Are there any updates on this?

No, sorry for letting you wait!

So I also think being able to set rules for the placement (e.g., hc rule class='float' placement=center) would be great.

Yes, placement=center is a perfect name for this. Please keep reminding me of this because this is really an issue/missing feature that should be fixed soon.

[blarz]

This can be closed with #957 being merged, right?

[bew]

#957 only implements none & center afaik, corners are not done yet (topleft, topright, bottomleft, bottomright), so I think this issue should be kept open until we're happy with the floatplacement values.

I think another useful one could be framecenter to place the window centered in the current frame.
(or even: frame:center, so we can have frame:topleft, etc...)

[t-wissmann]

Do you really need the corners? I'm hesitating to implement (and to write tests for) something that only might be useful at some point in the future.

[bew]

Do you really need the corners?

I don't need the corners myself, it's just that it was mentionned by @bklaase in his first post of the issue, so I wanted to remind you about it!

However the framecenter option (or whatever the name) could be quite helpful in addition to center.

[bklaase]

yeah, no to be honest I am very grateful for the feature as is, and the rest of my initial post is (in hindsight) just rambling ;)

I'll close this, for bookkeeping, sorry to have left it open :)

[t-wissmann]

yeah, no to be honest I am very grateful for the feature as is, and the rest of my initial post is (in hindsight) just rambling ;)

I'll close this, for bookkeeping, sorry to have left it open :)

I see! No Worries! :)
Btw I find that 'smart' is reasonable (and I wanted this feature for a very long time, now it's in #978), even though you didn't explicitly asked for this.

[Piping]

is this feature available in v0.8.3?

[t-wissmann]

is this feature available in v0.8.3?

No, you need the current git-version for this feature, @Piping