While the output in the UI was fine, certain job configurations containing non-aggregatable fields could trigger errors on the Elasticsearch side.
This PR fixes it by adding an additional query for fieldCaps to check first which fields are actually aggregatable.
* Refactor vis components to TypeScript
* Fix issue with ResizeChecker
* Fix calling onInit for no data
* Explicit named export
* Add title to vistype
* Fix error in test file
* Move onInit to no VisualizationNoResults
* Make listenOnChange changeable
* Add memoize util
* Use memoize for no results check
* Address issue with uiState
* Optimize memoize function
For categorization jobs, job validation would report that mlcategory isn't an aggregatable field. This fix checks the job configuration and only reports the error if the job config isn't using categorization_field_name and the detector field isn't set to mlcategory.
- Migrates the ml-form-label directive to use EUI/React.
- Exposes both FormLabel and JsonTooltip as React components from individual files so they can be used in a React context when the wrapping element also has been already ported to React.
- Adds jests based tests for the FormLabel and JsonTooltip components. They try where possible to make the same assertions like the mocha based tests. The mocha based tests are kept in the code for now so the code gets still tested in a angular based context and as a reference to have the same mocha/jest based tests side by side as a reference for migration.
- The FormLabel component is done in a way so it supports transclusion in both cases when used with React alone (using the children prop) and angularjs (using a ref callback and angular's transclude()).
* avoid day long gaps in sample data
* avoid using toISOString to avoid an timezone problems
* unskip sample test now that problem is fixed
* use much better cj algorithm for translating time
* cjcenizal review updates
* update funtion name in install.js
* push source reference date back a week
* Replace notify.warning with toastNotifications in region map, vega, index_pattern, redirect_when_missing, graph, monitoring, and ML
* Link to index patterns from Graph toast.
* Delete RouteBasedNotifier.
* Remove courierNotifier and SearchTimeout and ShardFailure errors.
* Remove warning and custom notifier types.
/api/_xpack/usage was added as a target for 6.4.0 but it will not be
used. Instead, the /api/stats response will include usage info on
everything that gets registered with the usage service in
/src/server/usage
/api/_kibana/v1/stats is a GET API that was added in 6.2, during a point
where we thought providing usage stats through a public API would be OK
for capturing internally, with the benefit of having it be visible.
However, we've pivoted away from that idea because it doesn't line up
too well with the existing flow of data, where usage stats are combined
with the "Kibana stats" such as process uptime and number of requests.
We want to shift how we collect stats from Kibana, but it will be
gradual. It might be a while before we have an architecture that makes
sense for a standalone public API for the usage stats
This endpoint was never documented, and isn't used anywhere in the code.
It does incur a maintenance cost though.
Therefore, instead of waiting for a next major version to remove this
API, I'm removing it for 6.4. It will be marked in the release notes as
a breaking change. Since it was never documented, it should not provide
a problem.
* partial progress on reactifying users
* progress on EUIfication of users screen
* removing Angular stuff
* adding data-test-subj="passwordConfirmationInput"
* removing data-test-subj="userFormEmailInput" refs from tests
* fixing selector for role assignment
* some functional test fixes
* fixing some functional tests
* fixing last functional test
* removing stray console log
* fixing warnings
* attempting to fix flaky test
* trying again to fix flaky test
* PR feedback
* PR feedback
* fixing issue where form tried to submit
* adding sleep to allow user to load
* Design edits
Mainly adding wrapper EUI page elements, but also shifted around form elements.
* Fixed console error and added responsive prop to table
* addressing PR feedback
* A few more PR feedback
- Fixed alignment of table
- Removed the tooltip from the lock icon and placed the description inline.
- Changed delete button to an empty button
* addressing more PR feedback
* adding email field back in
* adding back username validation
* restoring original error message
* fixing dumb null error
Many modules are still in beta which made the beta label not very useful in the overview page. This disables the beta label for now.
The label is still shown when clicking on module.
In order to make the awesome new kfetch api easier to consume in purely TypeScript projects, and since it's a pretty small module with very few dependencies, I converted it to TypeScript.
Along with kfetch I also started a type definition file for `ui/chrome` that we can extend as we go, but will likely be unnecessary after #19992
Just like our ESLint rules, I think we should enforce named exports over default exports via TSLint. No changes to the code were necessary because of this linting rule because everyone is already following the pattern we established with ESLint, but TSLint wasn't actually verifying the pattern.
* [Stats API] Set API field names per spec
* fix jest tests
* fix api integration test
* trash the original metrics collector
- constantly accumulating stats over time does not align with the existing behavior, which is to reset the stats to 0 whenever they are pulled
* move some logic out of the collector types combiner into inline
- change the signature of sourceKibana
* Make a new stats collector for the API
- to not clear the data when pulling via the api
- fetching is a read-only thing
* isolate data transforms for api data and upload data
* no static methods
* remove external in bytes
* remove the _stats prefix for kibana and reporting
* update jest test snapshot
* fix collector_types_combiner test
* fix usage api
* add test suite todo comment
* reduce some loc change
* roll back mysterious change
* reduce some more loc change
* comment correction
* reduce more loc change
* whitespace
* comment question
* fix cluster_uuid
* fix stats integration test
* fix bulk uploader test, combineTypes is no longer external
* very important comments about the current nature of stats represented and long-term goals
* add stats api tests with/without authentication
* fix more fields to match data model
* fix more tests
* fix jest test
* remove TODO
* remove sockets
* use snake_case for api field names
* restore accidental removal + copy/paste error
* sourceKibana -> getKibanaInfoForStats
* skip usage test on legacy endpoint
* fix api tests
* more comment
* stop putting a field in that used to be omitted
* fix the internal type to ID the usage data for bulk uploader
* correct the kibana usage type value, which is shown as-is in the API
* more fixes for the constants identifying collector types + test against duplicates
* add a comment on a hack, and a whitespace fix
* add preview for creating scripted fields
* Move previewed results into flyout. (#25)
* Move previewed results into flyout. Execute script as soon as the tab is selected.
* Format error in a danger callout.
* add functional test for preview
* move executeScript functionallity into lib so it can be used to validate on save
* validate script on save
* add functional test for script validation during save
* fix jest tests
* fix jest tests
* only validate script when field is scripted
* only add script callouts and help flyout to DOM for scripted fields
* move scripting stuff to its own render method
* right justify delete button to have same style as delete button for user management
* use data-test-subj key to find invalid script error msg in functional tests
* move help flyout link out of formRow help text and into its own row
* add savedObjects to tutorial schema, add savedObjects to APM, add bulk create endpoint
* SavedObjectInstaller component
* bulkCreate fixes
* fix tutorial jest test
* update from sqren review
* updated copy
* move saved object json into seperate json files
* minor commit clean up
* ensure isMounted before setting state after async call, allow manifest to customize saved object install message
* remove duplicated logic for getting config xpack.apm.indexPattern
* refactor get index pattern title
* add functional test that loads APM saved objects
* remove extra await
* display overwrite message
* use angular free savedObjectClient
* functional test cleanup
* handle bulkRequest exception and add jest tests for SavedObjectsInstaller
* use Promise.reject instead of throw
* update copy
* [ML] Adding close jobs menu item
* changing icon
* updates based on review
* adding extra check for close failure
* adding extra guard against missing response
* Improve tests for histogram, tag cloud, input controls and enable lab mode.
* enable skipped tests
* Update shakespeare test navigation to visualize
* Add sleep to shakespeare test
* Skip shakespeare test for now, it is not clear why it is failing and does not fail locally, need the other tests to run in CI
* Revert selector for selectAggregation, re-enable shakespeare test
* skip vega spec text test, was not fixed with changes
* Enable only visualize tests to run, need to see results and it's being blocked by other test failures
* Enable all tests again, visualization tests all passed
* Skipping sample data test due to bug #20807, let's see if rest of the tests pass
Fixes#20694
Implements super basic new platform `core` system, which includes two services: `core.injectedMetadata` and `core.legacyPlatform`. The `core` currently has two responsibilities:
1. read the metadata from the DOM and initialize the `ui/metadata` module with legacy metadata, proving out how we plan to expose data from the new platform through the existing APIs/modules to the legacy platform.
2. bootstrap the legacy platform by loading either `ui/chrome` or `ui/test_harness`
Because `core` mutates the `ui/metadata` module before bootstrapping the legacy platform all existing consumers of `ui/metadata` won't be impacted by the fact that metadata loading was moved into the new platform. We plan to do this for many other services that will need to exist in both the legacy and new platforms, like `ui/chrome` (see #20696).
