Rect-based event targeting in views
Author: Terry Anderson (tdanderson at chromium)
Views (tabs, tab close buttons, bookmark star, etc.) formerly only supported point-based event targeting, which meant that targeting for gestures only considered the center point of the touch region and not its size. Treating gesture events in the same manner as mouse events when determining the event's target can lead to a poor experience for touchscreen users because a touch region can overlap many more possible targets than a mouse cursor and it is difficult to ensure that the center of the touch region is over the intended target (especially if that target is small, such as the tab close button). The goal of this feature is to add support for rect-based event targeting ("fuzzing") by using a heuristic to determine the most probable target of a gesture having its touch region represented by a rectangle.
This heuristic is implemented in View::GetEventHandlerForRect() with the idea that small targets are more difficult to touch reliably than large targets, so small targets should get priority when determining the most probable target of a gesture. The heuristic works by recursively looking for candidate targets among the descendants of this which the touch region overlaps by at least 60%. Among all such candidates, we return the target which has its center point closest to the center point of the touch region. If no candidates for rect-based targeting exist, we instead return the target that would have been returned if point-based event targeting were used with the center point of the touch region.
Point-based targeting still uses View::GetEventHandlerForPoint(gfx::Point), but this function has been made non-virtual and simply calls into the virtual function GetEventHandlerForRect(gfx::Rect) with a unit rect (i.e., having size 1x1). The implementation of View::GetEventHandlerForRect() and its overrides has preserved the existing behavior of point-based targeting in this case. Otherwise, if called with a non-unit rect, the overrides of GetEventHandlerForRect() call directly into View::GetEventHandlerForRect(), with two exceptions mentioned in the "Future work" section below.
See r232891 for the implementation details.
The algorithm does not seem to perform as well for touchscreens which do not report their own radius values (as compared to touchscreens which do); see issue 315795. Furthermore, rect-based event targeting is not yet implemented for AutofillDialogViews or NotificationView.
UMA metrics have been added to the Ash.GestureTarget histogram to give a rough estimation of how difficult it is for users to successfully tap on views elements (see r209263). These metrics specifically focus on successful taps (where the tap resulted in a user-visible effect) and unsuccessful taps (where the tap had no visible effect) on the tabstrip and surrounding regions.
Successful taps counted:
GESTURE_TABSWITCH_TAP: A tap on a currently unselected tab (i.e., the tap was responsible for selecting the tab). GESTURE_TABCLOSE_TAP: A tap on a tab's close button. GESTURE_NEWTAB_TAP: A tap on the new tab button. GESTURE_FRAMEMAXIMIZE_TAP: A tap on the frame maximize button. GESTURE_MAXIMIZE_DOUBLETAP: A double-tap on the browser window resulting in a maximize or restore action.
Unsuccessful taps counted:
GESTURE_TABNOSWITCH_TAP: A tap on a currently selected tab. GESTURE_FRAMEVIEW_TAP: A tap on the space to the right of the tabstrip, left of the tabstrip, or in between tabs\[2\]. GESTURE_ROOTVIEWTOP_TAP: A tap on the top edge of a browser window.
The effectiveness is computed as the ratio of successful taps to unsuccesful taps.
This measurement is not perfect because it does not take into account cases where a tap was classified as successful but had an unintended result (for example, tapping with the intent of closing a tab but instead having a new tab open). Measuring these types of cases would require one or both of the following:
Adding new metrics that track pairs of actions which occur with a very small delay in between. For example, a tap on a tab close button followed almost immediately by a Ctrl-Alt-T to restore the most recently closed tab may indicate that closing the tab was not the user’s intent.
- Conducting a user study where participants execute a set of tasks (e.g., “close the second tab”, “open a new tab”) and we keep track of how many taps were required to accomplish each one.
 Originally the distance to the center line (the line that remains after repeatedly removing 1px borders from the rectangle) of the view was used in order to prevent bias against wide/tall rectangular targets. But in r241955 this was changed to instead use the center point to fix an edge case and also because wide/tall targets would never actually be candidates for rect-based targeting due to their size relative to a touch region.
 Each double-tap to maximize or restore the window contributes 2 to the value of this count, and each single tap also contributes 2 to the value of this count. Thus when interpreting the results of the data, the value of GESTURE_FRAMEVIEW_TAP first needs to be adjusted as (GESTURE_FRAMEVIEW_TAP - (2 * GESTURE_MAXIMIZE_DOUBLETAP)) / 2.