Release notes
What changed in each Atlas release, as you notice it in the app. Atlas runs on openFHIR Enterprise; the engine’s own changes are in its release notes.
Unreleased
Added
Slotting in a model that maps another FHIR resource writes the reference + slot mapping the engine needs to step into that resource: from a Bundle through its entries of that type, from any other resource through a Reference element you name. Every way of slotting a model in (Slot it in…, + Slot, reusing a model, the create wizard) asks under FHIR resource whether the model maps the same resource or enters a new one, with a default from the two models’ types.
A new problem,
slot-resource-mismatch, flags a slot the engine would skip because its model maps another resource. The Slot archetype field offers Wrap in a reference tofor it, as one edit you can undo. The ConceptMap editor and the inline terminology rows suggest code systems and codes from the tenant’s FHIR packages and core FHIR: system fields offer the systems (those already used, and in inline rows those of the element’s binding, first), code and display fields the codes of the group’s or row’s system, and picking one fills the other when it is empty. A line under each system field says which package it comes from and how many of its codes are known, or that no loaded package knows it. Anything can still be typed.
The unmapped rule of a ConceptMap group has an Other map field for mode other-map.
A user guide for Atlas: guides for every task, a reference page for each screen, every problem and message explained, keyboard shortcuts, deep links and what Atlas remembers. Each release of Atlas has its own version of the guide.
Help links into the guide of the version you run: Documentation in the sidebar, a help icon next to every page title and in the Workspace top bar, a link on each row of the Problems dock to that problem’s explanation, How coverage is counted in the coverage legends, and When is versioning available? where versioning is unavailable.
API access links the Atlas guide Connect your system next to the openFHIR API documentation.
A template uploaded without a context mapper opens in the Workspace on New context mapper: pick the FHIR version and the profile (or base resource) the template maps to, and Atlas creates the context mapper, named after the template, ready for its start and models. Until then the Template coverage offers no + model.
An empty Engine state page says when the git remote has commits the engine hasn’t pulled, most likely your mappings, and offers Import from the remote, which imports them the way the Git remote settings’ The repository — import it into the engine does. Load demo mappings asks first in that case.
FHIR packages take several files at once, or a
.zipof them: npm packages, and loose StructureDefinitions, ValueSets and CodeSystems (as.jsonor Bundles), which are stored together as one package you can name. ValueSets and CodeSystems no longer need a profile beside them. ConceptMaps among the files go to Terminology. The files are listed before Upload, and afterwards the page says what was stored and which files were skipped. The Upload profiles / packages button of the create-model and context wizards does the same.
Changed
+ Slot on an archetype occurrence works like NOT YET MAPPED, with the occurrence chosen: it lists the archetype’s models (this template’s first, then other templates’, which are added to the context when picked) with New model…, or names a new model when none exists. It no longer picks a model by name on its own (the occurrence’s own, or the one nothing slots in), and adds nothing until you choose. A failure shows in that dialog.
Use an existing mapping? says New model… instead of Create new instead.
The Slot archetype suggestions offer New model for
… after the archetype’s existing models, not only when it has none.+ Slot (and + model) no longer show on an archetype occurrence whose enclosing model already maps elements inside it, and Slot into shows that occurrence greyed with the reason: a model slotted in there would map the same elements twice. A mapping that only leads into a slot there (a reference + slot) doesn’t count.
The ConceptMap editor’s fields and buttons look like the Workspace’s (grey instead of orange).
An upload on the Engine state page stores StructureDefinition, ValueSet and CodeSystem
.jsonfiles (and Bundles of them) as one package in the FHIR packages when the upload is applied, instead of skipping them; a Bundle’s ConceptMaps go to the terminology store.One file that can’t be stored stops a FHIR packages upload before anything is stored, naming the file; the files stay listed to fix and upload again.
An upload whose plan has errors can be imported anyway: tick Import despite these errors in the preview. The engine imports what it can store and leaves out the rest, marked in the preview and the result. Unticked, nothing is imported, as before.
The repository — import it into the engine (first connection to a branch with content) imports right away, without a preview and a second Import into the engine click. Errors in the repository don’t stop the import (with an engine that allows it).
Plans that haven’t been applied (previews, and imports stopped by errors) count what they would do (2 to create · 1 to delete) instead of 2 created · 1 deleted. A pull or restore stopped by errors says Nothing imported.
openEHR attribute suggestions follow the engine, which reads every spelling of an attribute the same way in every field: manual entries and condition fields suggest the RM name (
magnitude,units,value/valuefor a coded text’s display text) instead of the flat pipe (|magnitude), a text element also offers a coded text’s attributes, and a node that is a data value itself (null_flavour) offers its attributes. Condition fields suggest the multimedia attributes anddefining_code/terminology_idthe way the engine resolves them.openehr-not-in-template is also reported for an attribute pinned through the element’s
valuethat none of its data value types has (…/value/magnitudeon a coded text), and for a manual entry’s openEHR path; a flat pipe written onto the node (items[at0004]|code) is no longer reported as a missing node. The unsaved flat path shown for an attribute path includes its flat attribute.The API access page labels the engine’s REST documentation openFHIR API documentation.
Installs without internet access can turn the help links off: an empty
atlas.docs.base-urlhides them.When the openFHIR engine has versioning turned off, the Versioning and Git remote pages say so and how to turn it on, instead of saying the engine doesn’t support versioning.
Resubmit on the Reviews tab is offered for pull requests closed without merging (a new pull request with the same changes), and no longer on open ones.
Push asks you to pull first, with Pull… at hand, when the branch has changes the engine hasn’t taken (or its history was rewritten), and to connect, with a link to the Git remote settings, when the branch is empty or was deleted. On a branch the engine never synced with it offers both, connecting first. The settings then offer Connect again for a deleted branch.
Git remote settings warn when a save would remove the stored token: it stays with the hosts it was given for, so a new repository host, an API base on a host of its own, or the first repository for a token stored before any removes it unless you enter a new one. Atlas says so after saving. A new repository URL gets its provider and API base detected again unless you change them yourself.
Fixed
After a few minutes in Atlas, every page failed with errors from the engine while you still looked signed in. Atlas now renews your sign-in with the engine as you work; once your openFHIR sign-in has really ended, it sends you to the sign-in page.
Pull, Restore and importing the repository from the Git remote settings show the plan with its errors (an artefact referencing one that is neither imported nor in the engine, say) instead of Could not preview the pull (or the restore, or the import), and apply it anyway: errors in what is imported don’t block any of them.