New issue
Advanced search Search tips

Issue 685329 link

Starred by 1 user

Issue metadata

Status: Available
Owner:
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Android
Pri: 3
Type: Bug



Sign in to add a comment

Evaluate Whether Clearing Omnibox Input on Android Should Suppress ZeroSuggest

Project Member Reported by mpear...@chromium.org, Jan 25 2017

Issue description

What steps will reproduce the problem?
(1) On an arbitrary web page (not the NTP or a search results page), tap to focus the omnibox.
(2) See the list of zerosuggest omnibox suggestions, which should be the current URL at the top, then a list of highly visited URLs (like those on the NTP) below that.
(3) Tap the "X" to clear the omnibox input, or press backspace to do it.  (The latter should work because the omnibox input is selected by default.)

What is the expected result?
1. all suggestions disappear, i.e., no suggestion for the current URL, and no most visited suggestions.

What happens instead?
1. all suggestions are still there.

It's clear that this form of modification isn't causing the Android omnibox code to send "input in progress".  Instead, it's still pretending to be an "on focus" omnibox call.

 
Cc: rolfe@chromium.org
Status: Available (was: Untriaged)
Summary: Evaluate Whether Clearing Omnibox Input on Android Should Suppress ZeroSuggest (was: Should Clearing Omnibox Input on Android Suppress ZeroSuggest?)
I don't think there's a UI principle that would help us decide whether the expected result is better than what currently happens.  +rolfe@ for her perspective to see if she has a strong opinion.

Instead, I think we should decide based on whether users actually do this behavior (clear input, then select something).  Perhaps we can do this from current UMA logs.

Regardless, I think showing the current URL is wrong.  Filed bug 685332.  Though I do wonder how it will look to have the highly visited suggestions hanging below a blank omnibox...  Maybe rolfe@ has an opinion on this too.

Comment 2 by rolfe@chromium.org, Jan 26 2017

Cc: emilyschechter@chromium.org
(+emilyschechter@ as omnibox pm)

Definitely this is a space UX is considering doing more around, but still undecided. Got a couple of relevant slides in this deck (slides 21-37):
https://docs.google.com/presentation/d/1RS6AAT5J_QZaqF7jk-Eqw7NPXxEB1mqM86anKub_hj0/edit#slide=id.g15fd973d4e_2_0

And I know bbergher@ is considering zero-query as a part of assist work. I'd say we're mostly waiting on how Chrome Home project work pans out from cleer@ and pschaffner@ as well.

But emilyschechter@ - open to whatever you might want to do here in the coming months as you solidify an omnibox plan.

Cc: mattreynolds@chromium.org
Mark suggested I leave this comment here:

It sounds like today, we can give suggestions on an empty omnibox (which is fine), but if you actually hit enter on the empty omnibox in this case, we'll navigate somewhere, at least on Android.  That isn't fine; we always want the omnibox text itself, without reference to the dropdown, to give a clue as to what will happen when you hit enter.  This is particularly important on Android where we don't indicate selection in the dropdown, so it's not obvious that the first match is actually the "selected entry".

There are a couple potential solutions here:
* Don't suggest things on an empty omnibox (probably undesirable)
* Add a placeholder "<enter query>" non-navigable match (we do this on desktop if you e.g. type just "?")
* Add a placeholder match, and also change the dropdown to hide this match, so there is no visible "selected entry"
* Generalize the previous bullet to hide the "what you typed" and placeholder matches in general (we tried this during 1993 and the metrics were bad, but I think we were trying on desktop, and it may make sense on Android given how we style the dropdown)
* Change the code so there actually might _be_ no selected entry (probably harder and riskier than the previous bullets, and would have the same effect)
I believe the third and fifth bullets are effectively equivalent from the user perspective.  Thus, I think Peter offered four meaningfully different user interaction possibilities.
Correct.

Comment 7 by rolfe@chromium.org, Mar 29 2017

Cc: -rolfe@chromium.org maxwalker@chromium.org
Removing myself and adding maxwalker as omnibox design contact. Might also be something bbergher is interested in, for Assist work.
Labels: Hotlist-Polish
maxwalker@, what do you think the right behavior is?
Components: -UI>Browser>Omnibox UI>Browser>Omnibox>ZeroSuggest
Labels: Needs-Feedback
Awaiting UX or PM feedback.
Ping maxwalker and emilyschechter. For what it's worth, this current behavior doesn't strike me as a problem.
omnibox needs-feedback ping. thanks!
Labels: -Needs-Feedback
Owner: maxwalker@chromium.org
[status update]

maxwalker@ mailed me separately to inquire about metrics about this.  It's clearly on his radar.  Officially assigning to him; when a decision is made, assign it back to the omnibox team.

Sign in to add a comment