Calculator suggestions should lead to a search results page with a calculator widget |
|||||
Issue description
Steps to reproduce:
1. Focus the omnibox and type: 2 * 2.
2. Click on the calculator suggestion ("= 4").
Expected result:
Land on a search results page for the query [2 * 2], with a calculator widget shown at the top of the page.
Actual result:
Land on a search results page for the query "4". There is no calculator widget and the results for [4] are not useful.
Note that calculator suggestions have always behaved this way, it's not a regression. But with the new answer layout making these more prominent, it would be more valuable to improve this behavior.
,
Aug 16
,
Aug 21
,
Aug 21
krb: assigning this to you to see if you can make it work since you were just addressing an issue in the dedupe code. Grab me some time and I can provide a bit more info on how I attempted to implement this earlier.
,
Aug 24
If you look at the URL they give us: https://www.google.com/search?q=4&oq=2*2&... it appears that there is little we can do, unless we want to re-write the URL. If you agree, I'll open a bug against Search.
,
Aug 27
Re comment #5: The suggest server isn't giving Chrome a URL. It's giving us a query, a type (CALCULATOR), and that kind of stuff. We write it into a URL. What additional information do you want them to provide?
,
Aug 29
I tracked down where we construct the search URL, and tried swapping the query and original query. The new URL indeed takes us to the Google calculator page, but unfortunately, this result doesn't show up any more. It is deduped away because the URL matches the WYT result, despite the addition of "oq=4". It's score is only 1250 vs 1300. Some of this, Justin mentioned before. Carving out an exception for calculator types while deduping is possible, but not elegant. (I verified it works.) I suspect we don't want to fiddle with scores. The last thing I can think of is tweaking which terms that this template URL considers relevant. It seems cleanest but I don't know what's involved there. (The idea doesn't sound quite right to me though. The issue is in Chrome, not Search, so it seems the fix should be in Chrome, not Search's template URL. It seems like the problem could repeat if other providers give us calculator results.) Thoughts?
,
Aug 29
I am okay putting into the place the policy that calculator matches are never marked as a duplicate of anything else. I think it makes sense. // A calculator match is an alternate presentation of the user's input. Its value is in its visual display. Hence, it should not be duplicate-eliminated against that match or any others. I agree that the alternatives sound worse.
,
Sep 6
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/cbe8e92f21c2ac2cea90badf3362f94e84d60d02 commit cbe8e92f21c2ac2cea90badf3362f94e84d60d02 Author: Kevin Bailey <krb@chromium.org> Date: Thu Sep 06 19:00:59 2018 [omnibox] Swap Calculator terms in URL to search correctly The current URL generated from a Calculator suggestion searches for the result e.g. "4", not the calculation e.g. "2+2". This CL swaps the terms (actually, it clears the original query) and, further, treats the suggestions as distinct from non-calculator suggestions, so that they don't dupe against other WYT suggestions. Bug: 873666 Change-Id: I006f9754ee7e48d0a21a1f149c4ef76b3ca509fd Reviewed-on: https://chromium-review.googlesource.com/1197342 Commit-Queue: Kevin Bailey <krb@chromium.org> Reviewed-by: Mark Pearson <mpearson@chromium.org> Cr-Commit-Position: refs/heads/master@{#589228} [modify] https://crrev.com/cbe8e92f21c2ac2cea90badf3362f94e84d60d02/components/omnibox/browser/autocomplete_result.cc [modify] https://crrev.com/cbe8e92f21c2ac2cea90badf3362f94e84d60d02/components/omnibox/browser/autocomplete_result.h [modify] https://crrev.com/cbe8e92f21c2ac2cea90badf3362f94e84d60d02/components/omnibox/browser/base_search_provider.cc [modify] https://crrev.com/cbe8e92f21c2ac2cea90badf3362f94e84d60d02/components/omnibox/browser/match_compare.h
,
Sep 7
|
|||||
►
Sign in to add a comment |
|||||
Comment 1 by jdonnelly@chromium.org
, Aug 13