Re: How can I start urxvt as fullscreen ?

[Jeremy Hankins <nowan nowan org>]

Teika Kazura <teika lavabit com> writes:
> On Thu, 16 Jul 2009 10:22:28 -0500, Jeremy Hankins wrote:
>> But I'm not sure how to do the new-viewport one because I'm not sure
>> how to check to see if a given viewport is empty.

Ok, first, I've written a match setter for new-viewport, since at the
very least it completes the symmetry with new-workspace.  This also
includes a minor fix to new-workspace, and puts both in the placement
section of the ui.

-------------------------------
diff --git a/lisp/sawfish/wm/ext/match-window.jl b/lisp/sawfish/wm/ext/match-window.jl
index 4a7ea50..9986d9b 100644
--- a/lisp/sawfish/wm/ext/match-window.jl
+++ b/lisp/sawfish/wm/ext/match-window.jl
@@ -77,7 +77,9 @@
        (depth (number -16 16))
        (placement-weight number)
        (fixed-position boolean)
-       (maximized (choice all vertical horizontal)))
+       (maximized (choice all vertical horizontal))
+       (new-workspace boolean)
+       (new-viewport boolean))
       (focus ,(_ "Focus")
        (raise-on-focus boolean)
        (focus-when-mapped boolean)
@@ -109,8 +111,7 @@
        (auto-gravity boolean)
        (shade-hover boolean)
        (transients-above (choice all parents none))
-       (ignore-stacking-requests boolean)
-       (new-workspace boolean))))
+       (ignore-stacking-requests boolean))))
 
   ;; alist of (PROPERTY . FEATURE) mapping properties to the lisp
   ;; libraries implementing them
@@ -395,10 +396,32 @@
    (lambda (w prop value)
      (declare (unused prop))
      (when value
-       (let ((space (car (workspace-limits))))
-         (while (not (workspace-empty-p space))
-          (setq space (1+ space)))
-            (set-window-workspaces w (list space))))))
+       (unless (window-get w 'placed)
+        (let ((space (car (workspace-limits))))
+          (while (not (workspace-empty-p space))
+            (setq space (1+ space)))
+          (set-window-workspaces w (list space)))))))
+
+  (define-match-window-setter 'new-viewport
+    (lambda (w prop value)
+      (declare (unused prop))
+      (when value
+       (unless (window-get w 'placed)
+         (let ((row 0)
+               (col 0)
+               (nomatch t))
+           (while (and nomatch (< row (cdr viewport-dimensions)))
+             (setq col 0)
+             (while (and nomatch (< col (car viewport-dimensions)))
+               (if (null (viewport-windows col row))
+                   (setq nomatch nil)
+                 (setq col (1+ col))))
+             (if nomatch
+                 (setq row (1+ row))))
+           ;; TODO: handle the nomatch case more intelligently?
+           (unless nomatch
+             (set-screen-viewport col row)
+             (set-window-viewport w col row)))))))
 
   (define-match-window-setter 'fullscreen-xinerama
    (lambda (w prop value)
-------------------------------

This depends on viewport-windows (which I guess should be put in
viewport.jl?).  It's a bit heavier than I'd like, especially as used in
new-viewport since it's called for each viewport (until and empty one is
found) and each call checks each window.  It's not a problem for my
setup, but I don't use all that many windows.  Perhaps someone more
skilled than I could improve it's efficiency, or improve the way it's
used in new-viewport.  Or maybe I'm just a worry-wart.

(define (viewport-windows col row #!optional workspace)
  "Provide a list of windows that can be seen (completely or
partially) from the specified viewport."
  (let* ((ws (or workspace current-workspace))
	 (width (screen-width))
	 (height (screen-height))
	 (left (+ (- viewport-x-offset) (* col width)))
	 (right (+ left (1- width)))
	 (top (+ (- viewport-y-offset) (* row height)))
	 (bottom (+ top (1- height))))
    (filter-windows (lambda (w)
		      (let ((pos (window-position w))
			    (dims (window-frame-dimensions w)))
			(and (window-in-workspace-p w ws)
			     (not (or (<= (+ (car pos) (car dims)) left)
				      (<= (+ (cdr pos) (cdr dims)) top)
				      (>= (car pos) right)
				      (>= (cdr pos) bottom)))))))))

Also, there's a bug in infinite-desktop: the cursor isn't repositioned
properly when moving up or to the left; here's a fix:

-------------------------------
diff --git a/lisp/sawfish/wm/ext/infinite-desktop.jl b/lisp/sawfish/wm/ext/infinite-desktop.jl
index d97d9d8..549a557 100644
--- a/lisp/sawfish/wm/ext/infinite-desktop.jl
+++ b/lisp/sawfish/wm/ext/infinite-desktop.jl
@@ -91,7 +91,7 @@
             (< (+ viewport-x-offset dist) minx))
        (setq dist (- minx viewport-x-offset)))
     (set-viewport (+ viewport-x-offset dist) viewport-y-offset)
-    (move-cursor (max dist cdist) 0)))
+    (move-cursor (- (max dist cdist)) 0)))
 
 (define (infinite-desktop.move-top)
   (let ((dist (- infinite-desktop.move-distance))
@@ -102,7 +102,7 @@
             (< (+ viewport-y-offset dist) miny))
        (setq dist (- miny viewport-y-offset)))
     (set-viewport viewport-x-offset (+ viewport-y-offset dist))
-    (move-cursor 0 (max dist cdist))))
+    (move-cursor 0 (- (max dist cdist)))))
 
 (define (infinite-desktop.move-bottom)
   (let ((dist infinite-desktop.move-distance)
-------------------------------

> Glad to see new-workspace (superb!), but I doubt that viewport is not
> for such aim. The meaning of viewport as "enlarged screen with scroll
> by screen-size" is getting vaguer with infinite-desktop which is
> shipped in Sawfish-1.5. Read my other email of today on this issue.

The point of infinite-desktop, as I understand it, is (a) to be flexible
about viewport boundaries (i.e., the "infinite" part), and (b) to use
the mouse to move around.  I'm personally not interested in the mousey
part, but dynamic viewport boundaries make sense to me.  But to do it
right it should update viewport-dimensions, etc, and interact properly
with the pager.

> But if someone insists new-viewport, one resort I propose is to expand
> the viewport dimension. :P

The new-viewport above just punts; if there's no empty viewport it does
nothing.  That's probably not ideal, but I'm unclear enough on the big
picture that I'm not sure what the best behavior would be.

-- 
Jeremy Hankins <nowan nowan org>
PGP fingerprint: 748F 4D16 538E 75D6 8333  9E10 D212 B5ED 37D0 0A03