# Roadmap The `afform` extension is a proof-of-concept. It aims to demonstrate the core model/concept in which AngularJS provides the standard data-format and component-model shared by developers (working in code files) and administrators (working in a programmatic GUI). As a proof-of-concept, it is necessarily incomplete. In particular: (a) some functionality is envisioned for additional extensions and (b) within this extension, there are known issues. ## Suite of Extensions This extension is expected to be the base for a suite of related extensions: * `afform`: Base framework and runtime. Provides APIs for developers. * `afform_html`: Present web-based editor for customizing forms in HTML notation. Use Monaco or ACE. (In the near term, it's a simple demonstration that forms can be edited by users in a browser; in the long term, it occupies a smaller niche as a developer/power-admin tool.) * `afform_gui`: Present a web-based editor for customizing forms with drag-drop/multi-pane UX. (In the long term, this is the main UI for admins.) * `afform_auditor`: Report on upgrade compatibility. Test changesets. Highlight dangerous/unknown/unsupported form elements. Score maintainability of the system. ## Documentation * Development and Refactoring: A guide for making changes to supported markup in a maintainable fashion. (*What if I need to rename a tag? What if I need to deprecate a tag? Ad nauseum*) ## Known Issues Within this extension, there are things which need updating/addressing: * Test coverage for key Angular directives (e.g. `af-api4-ctrl`, `af-api4-action`) * There are several `FIXME`/`TODO` declarations in the code for checking pre-conditions, reporting errors, handling edge-cases, etc. * Although afforms can be used in AngularJS, they don't fully support tooling like `cv ang:html:list` and `hook_civicrm_alterAngular` changesets. We'll need a core patch to allow that. (Ex: Define partials via callback.) * We generally need to provide more services for managing/accessing data (e.g. `crm-api3`). * We need a formal way to enumerate the library of available tags/directives/attributes. This, in turn, will drive the drag-drop UI and any validation/auditing. * Haven't decided if we should support a `client_route` property (i.e. defining a skeletal controller and route for any form). On the plus side, make it easier to add items to the `civicrm/a` base-page. On the flipside, we don't currently have a strong use-case, and developers can get the same effect with `civix generate:angular-page` and embedding `
`. * Injecting an afform onto an existing Civi page is currently as difficult as injecting any other AngularJS widget -- which is to say that (a) it's fine for a Civi-Angular page and (b) it's lousy on a non-Angular page. * The data-storage of user-edited forms supports primitive branching and no merging or rebasing. In an ideal world (OHUPA-4), we'd incorporate a merge or rebase mechanism (and provide the diff/export on web+cli). To reduce unnecessary merge-conflicts and allow structured UI for bona-fide merge-conflicts, the diff/merge should be based on HTML elements and IDs (rather than lines-of-text). * API Request Batching -- If a page makes multiple API calls at the same time, they fire as separate HTTP requests. This concern is somewhat mitigated by HTTP/2, but not really -- because each subrequest requires a separate CMS+CRM bootstrap. Instead, the JS API adapter should support batching (i.e. all API calls issued within a 5ms window are sent as a batch). * Default CSS: There's no mechanism for defining adhoc CSS. This is arguably a feature, though, because the CSS classes should be validated (to ensure theme interoperability). * `Civi/Angular/ChangeSet.php` previously had an integrity check that activated in developer mode (`\CRM_Core_Config::singleton()->debug && $coder->checkConsistentHtml($html)`). This has been removed because it was a bit brittle about self-closing HTML tags. However, the general concept of HTML validation should be reinstated as part of the `afform_auditor`. * `hook_alterAngular` is used to inject APIv4 metadata for certain tags. This behavior needs a unit-test.