Due to the async nature (??) of the trimming code, this is causing severe performance issues
during search.
Looking back through commits, doesn't seem like there's a particularly good reason we were trimming here. All I could find is #9986 (comment) which is lacking explanation of why this is actually useful.
And currently, we're dealing with an origin (not a full url when this was initially written, I think), i.e. https://accounts.firefox.com vs https://accounts.firefox.com/signin. So, the suffix stripping isn't even doing much beyond removing com in vast majority of cases.
So, seems like all of this trimming stuff can be cleaned up.
This introduces a separate HistoryDB type at the PagedHistoryProvider
layer, that doesn't need to deal with positions. Positioning logic in
HistoryDataSource becomes a type conversion between the new type and an
existing History type that UI and ItemKeyedDataSource API is written against.
With this refactor, we entirely eliminate nullability from these types.
We were converting Long timestamps into Ints (and getting negative
numbers back), and treating that as, basically, a position for the
paging API; the paging API would pass us back the obscure negative
number back as an offset, and we'll mishandle it resulting in an
infinite loop.
This patch removes all of the Long -> Int conversions, and introduces an
explicit 'position' that is calculated once we have a full page of
results completed.
* For #22534 - Update homescreen section name to "Recently visited"
* For #22534 - Show both history highlights and groups in Recently visited
For now the metadata groups don't support scoring so as an interim solution we
will show up to 9 items, evenly distributes, first favoring groups sorted by
date then history highlights pre-sorted by default.
Tapping a history highlight will switch to it's already open tab if available
or create a new one in which to load it if needed.
A "Remove" option will also be available for history highlights to remove it
from the screen and also from history.
Currently removing a group / highlight will not query new ones to again show up
to 9 items, this will be implemented separately.
* For #22534 - Rename and refactor historymetadata to recentvisits
The updated feature supports more than history metadata so updating the overall
naming scheme seems needed.
To signal that this is a homescreen feature the entire package is moved to home
* For #22534 - Update UI tests to account for the new items space on the screen
Saw failures about not finding the collection section on screen.
This is probably happening because w are now adding the recent visits to
homescreen above the collections section pushing it off screen.
Since the collections might be obstructed by the toolbar shown on top as a
quick solution we'll scroll to the next homescreen section so that the
collections will be shown above in their entirety.
* Update app/src/main/java/org/mozilla/fenix/home/recentvisits/RecentVisitsFeature.kt
Co-authored-by: Christian Sadilek <christian.sadilek@gmail.com>
* Update app/src/main/java/org/mozilla/fenix/home/recentvisits/RecentVisitsFeature.kt
Co-authored-by: Christian Sadilek <christian.sadilek@gmail.com>
Co-authored-by: Gabriel Luong <gabriel.luong@gmail.com>
Co-authored-by: Christian Sadilek <christian.sadilek@gmail.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
* For #22410 - Refactored tab sorter metrics into a middleware
* For #22410 - Created distribution metric for tab group sizes
* For #22410 - Created tests for tabs tray middleware
* For #22410 - Merge fixes
* For #22410 - Added PR number to metric
* For #22410 - Fixed unit tests post merge. Added waitUntilIdle to new tests.
* For #22410 - Added missing line to middleware to have the Store process actions
* For #22410 - Updated metric expiration to December
* For #22410 - PR Feedback
* For #22410 - Removed else from middleware when
This patch fixes two problems:
1) We were treating "direct tab load" as an event which applies
uniformally to all tabs, even though it's actually an event which
happens for a specific tab. This lead to background tabs (pages opened as new tab)
setting the direct load flag, and then a simultaneously loading
parent tab would incorrectly interpret that flag for itself.
The patch switches this tracking from a simple boolean (are we direct
loading?) to a set of tab IDs that are currently direct loading.
2) In a case when a background tab was loading with a parent who's
search terms were cleared by a direct load, we were not trying to
lookup search terms on the background tab's historyMetadata key,
which exists to capture search terms for this exact scenario.
The patch adds an additional fallback lookup for that path.
Setting this value in FenixApplication.onCreate was buggy because of a race
with restoring BrowserState.
Setting it here would ensure a better granularity of the events and so to more
accurate reporting.
This allows querying from all throughout the app which of the current tabs are
inactive while taking into consideration whether the feature is enabled or not
such that when the feature is disabled it will always return an empty result.
Roger Yang
Gabriel Luong
mcarare
Matt Tighe
Alexandru2909
Grigory Kruglov
Brais Gabín
mavduevskiy
Grisha Kruglov
Mugurell
aime Soriano Pastor
MickeyMoz
Christian Sadilek
Noah Bond
Arturo Mejia
Jonathan Almeida