This commits adds stats about clicking on stripe marker with the following payload: severity, total number of markers in a file, file type
GitOrigin-RevId: fbf54fa363471dc3c8227c4f90af6152696bbead
Commands via Driver are executed in a different thread (JMX one) so ThreadLocal variable with current context is missing and all spans from the commands appear in the root.
GitOrigin-RevId: 646c4a5b1fa6610ce5c97bc109e74824bc32d7ad
Try to avoid `expr as Type?` when `expr as Type` is possible.
Usually, this is further cleaned up to just `expr`
by the "useless cast" post-processing.
KTIJ-29147
GitOrigin-RevId: 91ee0d265b5ee0f123581df7b269181c91e96061
`arrayOfNulls` call returns an `Array<T?>`, but it seems that the type argument
to the call should be just `T`.
KTIJ-29147
GitOrigin-RevId: 52d73f8f35fe94fd5a4803f15731cc39bb9d0c79
Migrate to AsyncTreeModel, just like with the Package Dependencies panel.
One difference is that this model needs to sort children, so introduce
a boolean parameter for that to distinguish between this and package dependencies.
Another important note is that instead of a regular NBRA, we use the model's
Invoker to retrieve package sets from the selected nodes. This is because the tree
model is single-threaded by its nature. The read lock would prevent
concurrent access, but it's still hard to reason about.
GitOrigin-RevId: 2095337ecceed015980212341e0075243d9c257e
There are other slow ops inside it already, so why keep this
one outside and get a lot of useless reports?
GitOrigin-RevId: 4444a1c871b6e2e87b651922d4cf37ae3970f490
This move directory with classes helper should only be responsible for moving Java files. #KTIJ-30171 Fixed
GitOrigin-RevId: 7d4b11ae7b0185920c60f51f7a3a4f5e1f7e7006
(MP-6452) Fix failing project structure and packaging tests
(MP-6452) Make sure that machine learning score is calculated correctly for each plugin
Previously, there was a cache with mutable list of features. There were two issues with it. First, because of mutability, it could store mlScore for a different query sometimes. Second, there were no guarantees on consequent call of ranking and features logging before the next ranking event happens, so the feature cache could be inconsistent in certain scenarios.
(MP-6452) Introduce search index in plugin manager events that corresponds to actual query order
I also change the definition of plugin manager session. The session is restarted only in two cases: when a user clicks on the Plugins section in settings or when the search is restarted after a plugin installation.
(MP-6452) Add additional features: is result ordered by ML, is user internal, experiment group and version
(MP-6452) Fix missing date of latest plugin update, add days since latest updates and textual features
(MP-6452) Apply suggestions from the code review
(MP-6452) Integrate the ranking plugin with the search and the logging group
In feature extractors, during the first run, I calculate the features for a model to predict plugin relevance and on the second run the features to report to the MP collector (with predicted score, for example). I also implement features cache to calculate the most of the features only once
(MP-6452) Implement the baseline plugin manager session id definition and log it to FUS and MP recorders
I attach the start of a session to enableSearch method of PluginManagerConfigurable that gets triggered on "Plugins" setting group selection in the menu
(MP-6452) Add plugin for plugins ranking in the Plugin Manager
ML in SE: add embedding search integration test subsystem tag
Merge-request: IJ-MR-130364
Merged-by: Evgeny Abramov <Evgeny.Abramov@jetbrains.com>
GitOrigin-RevId: 9136d316aec2ede74bec07798dd8db16e7849f54
It works in JPS, and enables `IdeParcelizeResolveExtension`
when there is a corresponding compiler plugin jar (namely "parcelize-compiler.jar")
specified with the `-Xplugin=` compiler argument
Currently this feature is under registry flag and disabled by default;
it should be enabled by default when it becomes clear that it
works as intended and does not break anything, and then the registry
flag should be completely removed.
Side note: `getInstance` function from `KotlinFacet.kt` returns
`compilerArguments`, but they do not include the arguments from the
"additionalArguments" sections in the Kotlin Facets.
This most probably means that currently the use-sites of this function
are not functioning properly, and need to be checked later - for example,
`KotlinSerializationEnabledChecker`
GitOrigin-RevId: 7fc0ff4dc07cf12770e004ccde93314a6b365ee2
platform.posix.DIR is a type alias to a forward declaration
that hadn't been supported until KT-62936 was fixed.
GitOrigin-RevId: 1d726fbaa731007eb18a8c8329b7c8e358cce2f7
