There's a potential race condition in the search code: if we have an
outstanding search call to a provider for search "A", and if before it comes
back we do a subsearch for "AB", we won't have any results to pass along.
Previously, we used an empty list when storing the provider results, so we
effectively told the remote search app to filter through this empty list for
any search results that meet the new query, meaning we showed the user 0
results for the provider in this case.
Now that we don't store an empty list, but instead store `undefined`, this race
raises a warning. Solve it by doing an initial search query in this case
instead.
The search code isn't too smart about chained subsearches: now, if we hit this
race while already on a subsearch, we'll do an initial search for the subsearch
query instead, but that is much better than showing the user nothing. This
could be fixed in the future for a performance improvement.
Reviewed-by: Florian Müllner <fmuellner@gnome.org>
Long ago, the search system worked in a synchronous manner: providers
were given a query, and results were collected in a single array of
[provider, results] pairs, and then the search display was updated
from that.
We introduced an asynchronous search system when we wanted to potentially
add a Zeitgeist search provider to the Shell in 3.2. For a while, search
providers were either async or sync, which worked by storing a dummy array
in the results, and adding a method for search providers to add results
later.
Later, we removed the search system entirely and ported the remaining
search providers to simply use the API to modify the empty array, but the
remains of the synchronous search system with its silly array still
lingered.
Finally, it's time to modernize. Promises^WCallbacks are the future.
Port the one remaining in-shell search engine (app search) to the new
callback based system, and simplify the remote search system in the
process.
When we create a result actor, cache it, so it can be used for
subsearches of the same initial. For now, to keep memory usage
and the stage graph relatively clean, don't persist the actors
across searches, but maybe we should do this in the future.
This also means that we don't query getResultMetas for items
that we've seen in the same initial search.
https://bugzilla.gnome.org/show_bug.cgi?id=704912
The existing provider system is split between a confusing mess of
RemoteSearch, SearchSystem, SearchDisplay, and ViewSelector, partly
because of the vestigal in-shell search system. Move most of the
logic to search.js so it's easier to read.
We fetch and store the list of providers from the search system when we
construct SearchResults, but we never update this list when providers are
changed at runtime, causing various bugs making the search not seem as
snappy as it should be. Make sure to always fetch the list of providers
from the search system.
search.js used to do a lot more, but now that most of the
functionality has been moved to the remote search system,
it doesn't do a lot. Merge searchDisplay.js into it.
When we reload the remote search providers, we currently try to remove
all remote providers, and then re-scan. It turns out that we sometimes
remove the wrong providers from the remote provider list, causing us to
have some providers not correctly unloaded.
https://bugzilla.gnome.org/show_bug.cgi?id=700283
pushResults, and the original async search API, were originally intended
so search results that weren't immediate could be added as they come in.
Since then, we've decided that the design of search results is that they
should finish at once with all results. Thus, the code was modified so
that pushResults always overwrote the current result set. As such, it makes
sense to rename the method so that the name matches the behavior.
https://bugzilla.gnome.org/show_bug.cgi?id=693836
Early on, search was based on a list of terms, which was like a set
of tags, in that terms were OR'd, and that order didn't matter. As
such, modifying any one of the terms wouldn't produce new results.
Nowadays, providers take the order into account, so a substring
should only be the case if new terms are added to the end.
https://bugzilla.gnome.org/show_bug.cgi?id=693935
This is causing more confusion than anything else these days; the DBus
API is properly documented now and that's what people are expected to
use, the rest are implementation details we're not interested in
exposing.
https://bugzilla.gnome.org/show_bug.cgi?id=681797
These are for all search results except apps (and Wanda).
We also simplify a bit the packing of search results, which removes some
ugly code in navigateFocus() where we needed to call
st_widget_navigate_focus() twice, since the grid icon was composed by
two nested boxes, both focusable.
https://bugzilla.gnome.org/show_bug.cgi?id=681797
Display a '+' icon on the provider icon if there are more results that are
hidden. If the provider icon is clicked, ask the provider to launch itself and
perform a search with the current terms.
https://bugzilla.gnome.org/show_bug.cgi?id=681797
The original design for the overview had buttons for searching for
Wikipedia and Google, but in practice this is a bad idea. The buttons
are the default activations, meaning that using the overview as a
fluent motion of launching something - "firefxo<Enter>", will launch
Google/Wikipedia.
https://bugzilla.gnome.org/show_bug.cgi?id=670168
clutter_actor_get_children requires making a temporary GSList from
a linked list structure, and then creating a JS Array from that GSList.
For simple cases like the number of children, use clutter_actor_get_n_children.
https://bugzilla.gnome.org/show_bug.cgi?id=677426
Now that all searches are async we can remove the code path for the
SearchSystem::search-completed signal which is no longer useful.
This patch ends up fixing the status text not being updated for when
there are no results.
https://bugzilla.gnome.org/show_bug.cgi?id=675328
As shown in the previous commits, synchronous search is easily implemented
by the asynchronous search API. The only reason we still have a
synchronous search API is of historical reasons. Well, we're not a museum,
and git log can keep our fossils safe if need be....
https://bugzilla.gnome.org/show_bug.cgi?id=675328
To allow this to happen, we need to make sure that we don't overwrite the
previousResults when calling the async method. Note that this is a bug of
some sort, we were already using this synchronous style when a remote
search failed.
https://bugzilla.gnome.org/show_bug.cgi?id=675328
We will allow applications to hook into shell's search by registering
a service which implements a well-known DBus interface.
"search-providers" is a reasonable directory name for applications to
drop their registration files, but it conflicts with "search_providers"
used by open search providers - rename the latter to avoid confusion.
https://bugzilla.gnome.org/show_bug.cgi?id=663125
Currently, asynchronous search providers are expected to call
startAsync() in getInitialResultSet()/getSubsearchResultSet(),
which will trigger async mode until the search is canceled or
updated. Switching between synchronous and asynchronous mode like
this makes asynchronous search an implementation detail, but being
transparent to the searchDisplay means that certain optimizations
don't work as expected. Namely, updating asynchronous search results
causes flickering, and the automatic selection never focuses
asynchronous results.
So change the API to require providers being either synchronous (with
the current getInitialResultSet()/getSubsearchResultSet() methods)
or asynchronous (with asynchronous variants), and handle asynchronous
providers explicitly in searchDisplay.
https://bugzilla.gnome.org/show_bug.cgi?id=663125
The last patch in the sequence. Every place that was previously
setting prototype has been ported to Lang.Class, to make code more
concise and allow for better toString().
https://bugzilla.gnome.org/show_bug.cgi?id=664436
All classes that have at least one other derived class (and thus
benefit from the framework) have been now ported. These includes
NMDevice, SearchProvider, AltTab.SwitcherList, and some other
stuff around.
https://bugzilla.gnome.org/show_bug.cgi?id=664436
js2-mode is no longer developed and we recommend js-mode these days,
so switch the modelines to specify that, and make them consistent
across all files.
https://bugzilla.gnome.org/show_bug.cgi?id=660358
As extensions are now expected to provide a "disable" function,
they need to remove search providers they added. Implement the
removal functionality and add a public removeSearchProvider()
method.
https://bugzilla.gnome.org/show_bug.cgi?id=657548
Some search providers may want to change their results, or may not
want to block on an external service to get their results (DBus, etc.)
Set up an infrastructure to allow search providers to add their search
results at a later time.
Based on a patch by Jasper St. Pierre and Seif Lotfy.
https://bugzilla.gnome.org/show_bug.cgi?id=655220
Adding correct annotations to Gio.File.load_contents revealed that gjs
doesn't actually support array+length combinations. For 3.0 this would
be invasive to fix, so add a method to ShellGlobal which does what
we need.
https://bugzilla.gnome.org/show_bug.cgi?id=646333
Search results' meta info currently is expected to have an 'icon'
property holding a Clutter.Texture of a fixed icon size. This
property is used to implement the createIcon() function of BaseIcon,
which is used to actually display the result, ignoring the size
parameter due to the fixed icon size.
Given that all available search providers create this property for
the desired icon size using a function with the same signature, it
appears logical to replace the fixed-sized 'icon' property with
such a function, so that the icon size will be defined by the display
actor rather than the search system.
https://bugzilla.gnome.org/show_bug.cgi?id=643632
Currently section headers in the search view are reactive and run
the corresponding provider's expandSearch() function when clicked,
which should launch an external program displaying all search
results for the section. Unfortunately it is only implemented for
the "Settings" provider, and does not work properly (as it ignores
the search and just launches System Settings).
Also current mockups deemphasize the section header, making the
feature pretty much indiscoverable (except by accident when
interfering with swipe-scrolling).
In conclusion, the feature does not make much sense in its current
form, so remove it.
https://bugzilla.gnome.org/show_bug.cgi?id=643632
With workspace thumbnails, we don't switch workspaces when dragging windows
between workspaces or adding new workspaces, so we also shouldn't switch
on launch.
* Add workspace parameters to shell_doc_system_open(),
shell_app_activate, shell_app_open_new_window()
* Pass a 'params' object when activating items in the overview with
two currently defined parameters: workspace and timestamp. (timestamp
is only implemented where it is easy and doesn't require interface
changes - using the global current timestamp for the shell is almost
always right or at least good enough.)
https://bugzilla.gnome.org/show_bug.cgi?id=640996
Current mockups display all search results as icons as used by
application results, so change the default result display to use
iconGrid/BaseIcon. Remove the custom application results display,
as it is no longer needed.
https://bugzilla.gnome.org/show_bug.cgi?id=634948
With the new layout, search results will be displayed in an independent
view like window previews, applications and possible future additions;
it does not make much sense keeping it with the switching logic, so move
the code to its own file.
Also remove the dash-prefix from the relevant style classes.
https://bugzilla.gnome.org/show_bug.cgi?id=634948
The current search system uses the OR operator to concatenate search
terms. While results which are matched multiple times sort before
other matches, it is almost guaranteed that adding an additional term
to the search increments the number of results, which is rather
surprising.
https://bugzilla.gnome.org/show_bug.cgi?id=610955
This is our convention.
The only exceptions are double quotes for words in comments that give
them a special meaning (though beware that these quotes are not truly
necessary most of the time) and double quotes that need to be a part
of the output string.
The previous commit broke activation when selecting an actor
from a custom renderer. Fix this by explicitly defining a method
activateSelected on the search renderer.
The high level goal is to separate the concern of searching for
things with display of those things; for example in newer mockups,
applications are displayed exactly the same as they look in the
AppWell.
Another goal was optimizing for speed; for example,
application search was pushed mostly down into C, and we avoid
lowercasing and normalizing every item over and over.
https://bugzilla.gnome.org/show_bug.cgi?id=603523
