BREAKING

If you want to use the dependent java apps (formerly solverd, bbmri and the current palga) you need to upgrade the MolgenisWebAppSecurity config class.

Features

• Added debug logging for user mapping when OIDC is configured
• Update Guava dependency to 30.1
• #9169 Add some new default i18n labels for new data explorer
• Audit logging
  • Can log data modification events for system and non-system entity types.
  • Logs settings changes
  • Logs authentication events
• M7874: Add roles from OIDC merged ID-token and userInfo claims
  Check: https://molgenis.gitbook.io/molgenis/access-control/guide-groups-roles
  • Auto-configuration enabled via issuerURL
  • Added audit events to see which roles you get from the VO groups and roles claims
• Configurable maxLength of attribute with string type (string, text, email, etc.)
• Maximum upload file size is configurable as a runtime property max.file.mb

Fair API

• Prioritize IRI attribute over iri created with BASE_URI + id for DCAT Resources
• Name the blank nodes with the valueof the Entity ID

Bugfixes

• Style attribute tree popovers in data explorer
• #9084: Dash not allowed in data api filters
• #9166: Error when downloading a file with a comma in the name
• #9159: Use bootstrap classes in tables
• #9026: Old version of JWKS URI used
• #9237 & #9240 Unnecessary synchronizations in decorator registries

Frontend features & fixes

See the release notes of the accompanying frontend release.

Bugfixes

• #9166 Error when downloading a file with a comma in the name

Bugfixes

• #9141: Python scripts do not run in docker
• #9142: Expression evaluation in old forms is broken

Features

• M6290: Use Graal JS as JavaScript engine
• M7513: Make OIDC email attribute configurable
• #9116: Show number of active user sessions in the user manager plugin
• #9120: Show if user has active session
• #9099: Add per entity data explorer settings table during bootstrap phase.
• #9129: Add active theme endpoint at '/theme/style.css'
• #9146: Made the sorta back button better readable

New theme workflow

Extracted theming from MOLGENIS platform and integrating it in the http://github.com/molgenis/molgenis-theme repository.
The default are configured in AppSettings and these are now served from these locations:

• BLUE theme bootstrap 4: /@molgenis-ui/molgenis-theme/dist/themes/mg-molgenis-blue-4.css
• BLUE theme bootstrap 3: /@molgenis-ui/molgenis-theme/dist/themes/mg-molgenis-blue-3.css

Default all themes will be served from https://unpkg.com.

If you use the theme-manager the possible themes will be served from https://unpkg.com/molgenis-ui/molgenis-theme. If you choose one that url will be served.

Use Graal JS as JavaScript engine

How?

There are two APIs to choose from, the ScriptEngine and the GraalVM Polyglot Context.
The latter gives more fine-grained control which we need to tweak performance.

Differences with Nashorn

There are some serious differences between the two.

Compiled expressions cache

We explicitly cached compiled expressions for Nashorn. The Graal Engine caches all scripts it has ever compiled. So the cache can go.

Graal JS does not allow you to share bindings

We used to bind once and then copy them around between contexts. That is not allowed in Graal JS. The MagmaScript context is rather expensive to create and our scripts are tiny ($('height').value()) so this is problematic.

No cleanup allowed

You cannot remove keys from a Context so if anyone does a var a = 5 in their MagmaScript, there's no getting rid of it.

Explicit enter

Threads need to enter the context before they can use it. If you explicitly enter the context before issuing a series of evaluations, that saves the cost of leaving and re-entering between calls.

Proxying is back

In Rhino we could send an Entity instance into JavaScript. This broke in Nashorn so we had to copy the entire Entity into a Map<String, Object>. In Graal we can proxy again!

Bugfixes

• #9053: OIDC User mapper does not respect the userNameAttribute configured in OidcClient
• #9078: ACEs not removed when deleting roles (Including a migration step that cleans up leftover ACEs)
• #6360: Separate sign-up form and OIDC settings
• #9139: SameSite warnings for Google Analytics cookies
• Use glyphicons in data explorer rather than gifs, to allow styling in all colors

Features

• Upgrade Fair Data Point to model version DCAT2
• Add search-query string operator to use in searching entities
  Available in /api/v2 and /api
  Documentation: RSQL usage
• Updated frontend apps
  • app-manager | 1.0.0
  • core-ui | 0.3.2
  • data-row-edit | 5.0.0
  • navigator | 0.2.1
  • searchall | 0.1.5
  • security | 0.4.1
  • settings | 4.0.1
  • scripts | 3.0.0

Bugfixes

• #9015: When you make ROLE_ANONYMOUS a viewer on your group, they cannot see metadata
• #9018: Google Analytics gets incorrect URLs
• #9073: Fix typo's in data row edit l10n
• #9086: Added new security-ui l10n strings
• #9082: You are now able to remove a group that has role extensions
• #9011: Updating one value via REST API v2 fails for reference attributes
• #9103: Intro page for Questionnaire is shown in application default language
• Added message when browser invalidates input, but doesn't return a value to Vue Forms to validate.
• Export DateTime in UTC

Other

• Switch to Hikari ConnectionPool
  • Scale down connections when idle
  • Add connection pool metrics
    Access it at https://my-molgenis.org/api/metrics/prometheus. The user needs to have the METRICS role)
• Replace ui-form:form_not_a_valid_url default message with ui-form:form_not_a_valid_hyperlink message

Bugfixes

• #8985: OIDC personal permissions lost
• #9044: Elasticsearch break after cleanup script

Features

• Allow user to select multiple matching ontology terms for a given input term in the SORTA / ontology
  module story 7385.
• Allow user flag a matched term for review in the SORTA / ontology
  module

Bugfixes

• #9024: Activation link is not idempotent (MS ATP issue)
• #9023: Google analytics does not work if privacy-friendly settings are not set

Search changes (M7343)

Standard tokenizer

For text field analysis in ElasticSearch, we switch from whitespace tokenizer to the standard tokenizer
and remove the word_delimiter filter which was causing too many false positive search results.

The word_delimiter filter was designed to remove punctuation from complex identifiers, such as product IDs or part numbers.

When upgrading to 8.4, all existing repositories will be reindexed with these new settings.

Case-insensitive like

The Like operator becomes case-insensitive.

token attributes

You can tag an attribute of type STRING, HYPERLINK, URL or ENUM with tag token.
It will then be analyzed using the keyword analyzer and searching on that field will only match case-insensitive equal values.

case-sensitive attributes

• If an attribute with tag token is also tagged case-sensitive, Search queries on that field are case-sensitive.
• If an attribute is tagged case-sensitive, Like queries on that field are case-sensitive.

The case-sensitive and token tags are created for you.

See also the RSQL guide and API test cases.

Features

• Expose selected theme

Bugfixes

• #8926 Duplicate basePath /api in swagger documentation
• #4080 Update active session authentication roles
• #8951 Redirect plugin fails in case of non top level item
• #6902 Deleting a manual match in SORTA throws an error
• #8976 Default apiV3 page size is incorrect
• #8548 Redirect plugin can redirect to URLs that are not present in the menu
• #8965 Indexing Ontologies takes quite a long time
• #8969: SORTA - ngram broken/removed (in favor of faster indexing?)

Reintroduced

• #7862 Ontology deletion non-practical

Other

• #8949 Upgrade dependencies

Bugfixes

• Can not redirect to arbitrary urls anymore. Redirects need to be in the menu administrated.

Bugfixes

• #8978 Personal permissions lost when logged in using OIDC

Features

• Filter biobanks in negotiator

Bugfixes:

#6902 Deleting a manual match in SORTA throws an error