530203 - improve heuristic used for complicated cases when closing a tab that spawned links in new tabs

Status: NEW

(Firefox :: Tabbed Browser, defect)

Description

[Paul O'Shannessy [:zpao] (not bugmail, email directly)]

Bug 465673 changed behavior so that tabs opened next to the current tab.

Per a discussion on d.a.f (http://groups.google.com/group/mozilla.dev.apps.firefox/browse_thread/thread/30b8bb749e061da0), we can improve the heuristic we used so that we handle the complicated cases better.

Comment 1

[Mike Beltzner [:beltzner, not reading bugmail]]

Most of the complicated cases surround links that are forced to open in new tabs as opposed to those that do on their own, correct?

Presently, if I have a tab "B" with a link that opens in a new window using target=blank or window.open (caps indicates focused tab):

a B c d
a b B1 c d    (link opens new window with focus)
a B c d       (closing the tab returns to parent)

Our current heuristic fails, however, when I use the link in tab "B" and that page has a link which also opens in a new window:

a B c d
a b B1 c d      (link opens new window with focus)
a b b1 B11 c d  (link opens new window with focus)
a b B1 c d      (close returns to parent)
a b C d         (close to right)

Here I would have expected to close back to my parent, but we've disposed that information (perhaps prematurely) and we close to right.

(You can use http://beltzner.ca/test/testA.html to try this out - it's a page with a link that uses target="_blank")

This could be solved by either always closing to the left (which is unlike other browsers) or maintaining the parenting information.

When a user forces a link to open multiple links in new tabs (without focus) things get more complicated. For a tab B where three such links are forced into tabs:

a B c d
a B b1 c d
a B b1 b2 c d
a B b1 b2 b3 c d

Now the close behaviour is pretty dependent on where the user places focus, whether they re-order tabs, etc. Let's consider the set {b1,b2,b3} to all understand that their parent was "b".

// Heuristic 1
If focus is switch to any of {b1,b2,b3}, closing a tab should switch to:
  - tab b* to the right, if available
  - otherwise tab b* to the left
  - otherwise tab b (the parent)

// Heuristic 2
If the user opens a new tab from any of the tabs in {b1, b2, b3}, those should be added to the set of tabs for which heuristic 1 applies.

// Heuristic 3
If the user re-orders tabs such that a new tab is inserted into the set {b1,b2,b3} (so either moves b* outside beyond another tab, or moves a tab inside the set) then we stop remembering parenting information.

// Heuristic 4
If the user changes tab focus to a tab outside the set, then we stop remembering parenting information. This includes creating a new tab.

I think that all should work. The only thing that might be wrong is the behaviour around creating a new tab.

Comment 2

[Jennifer Morrow [:Boriss] (UX)]

Whether or not this behavior is the right option, it's worth noting that it would create a zig-zag effect with enough tabs that could seem unpredictable.

If the user begins at the end state above, switches focus to one of the b* tabs, and creates two more spawning tabs, would you consider the b** tabs to be a family within the b* family, and close tabs within one family completely before moving to the next?

Behavior would be:

a (B) b1 b2 b3 c d
a b b1 (B2) b3 c d (switch focus)
a b b1 b2 (B2a) b3 c d (new tab from b2)
a b b1 b2 b2a (B2b) b3 c d (new tab from b2a)
a b b1 b2 (B2a) b2b b3 c d (focus switched back to b2a)

User now presses "close tab" many times in a row.  This would be the behavior described in Comment 1

a b b1 b2 (B2a) b2b b3 c d
a b b1 b2 (B2b) b3 c d (focus moves right)
a b b1 (B2) b3 c d (focus moves left)
a b b1 (B3) c d (focus moves right)
a b (B1) c d (focus moves left)
a (B) c d (focus moves left)
a (C) d (focus moves right)

Comment 3

[Mike Beltzner [:beltzner, not reading bugmail]]

(In reply to comment #2)
> Whether or not this behavior is the right option, it's worth noting that it
> would create a zig-zag effect with enough tabs that could seem unpredictable.

Yeah, that might be true. We can (and should!) identify edge-y cases. The bit I worry about most is the case where a bunch of nested sets have been opened, and then someone forgets the parentage, and closing tabs seems random.

> a family within the b* family, and close tabs within one family completely
> before moving to the next?

That makes most sense to me, yes. Perhaps we should nest the heuristic, as well. So if you open a set within a set, the heuristic only holds if you then close the inner set before the outer set. (ugh)

Comment 4

[Vova Olar]

Any news here?