Miscellaneous TODOs

[ekuiter]

Some TODOs and ideas for future extensions.

• cross-tree constraint editor (=> moved to Constraint editor #16)
• provide other techniques for drawing links (not only curved, but also straight lines etc.) as settings in the settings panel
• BDD updating (i.e., do not always recompute SAT, but only update (not reconstruct) a BDD) for analysis instead of SAT (is it feasible in spite of space complexity?)
• abstract away things like a Concurrency- or SynchronizationStrategy, if we implement different ones (=> use different kernels)
• let users decide which undo/redo model to use (at least: global undo +
  some sort of local undo) (=> only chronological undo)
• currently, we identify features by name. DarwinSPL identifies features by
  reference (i.e., UUID), which simplifies renaming operations and allows
  duplicate feature names. is this a good idea? => yes, simplifies concurrency.
• handle missing elements of FeatureIDE's feature model format: struct, constraints, properties, calculations, comments, feature-order, rule (=> replaced by our own feature model format tailored to collaborative editing)
• publish project, add logo (favicon, readme, github deploy button)
• when project is public, add Travis CI (and maybe coverage) badge to README.md (possibly also deploy batches for Netlify, Bluemix and Azure)
• might combine with a configuration recommender system (which features are chosen often together, see Pereira et al.)
• EFM (extended/attributed feature model) support
• auto-calculate metrics for configurations (from extended attributes)
• recommend good/often-used configurations
• minimap for big feature models (like Sublime Text)
• analysis results webrtc? how to guarantee certain analysis results? i.e., the editor should have a mode in which it's guaranteed that the FM will never be inconsistent (no products) and otherwise edits will be rejected
• "share feature model" via link so that others can see it (i.e., routing for projects and artifacts - e.g., http://.../project/artifact)
• a nice way to visualize kernel state in our editor? visually integrated into the editor (debug mode)?
• logging and statistics for operation types, conflicts, kernel performance (garbage collection), undo etc.
• SPLOT seems to be used in a lot of literature, especially for its feature model repository. maybe it would be good to add the SPLOT feature model format as input type to variED/FeatureIDE
• PDF export seems to work, but PDF is then broken in LaTEX export
• for validity and sanity checks in the kernel, use :pre and :post of defn (https://github.com/bbatsov/clojure-style-guide#pre-post-conditions)
• in the kernel, use alter-var-root instead of def (https://github.com/bbatsov/clojure-style-guide#alter-var)
• instead of implementing feature models once in Clojure, implement them both in TypeScript and Java. This way, no more conversions to/from Clojure are necessary (better performance for bigger feature models, can avoid FM-encodecall) (see Optimize kernel performance #32)
• display a * in the window title whenever unflushed messages exist (consistent with the onBeforeUnload handler)
• make the command palette touch friendly (scrollable)
• use code spliting to load the kernel only when requested (dynamic import). This is useful to shorten the initial load time. When a user requests to join, the dynamic import can be issued simultaneously with the JOIN_REQUEST (Promise.all), which needs some time anyway.
• add a "feedback box" where users can suggest improvements etc.
• make a general notification scheme, where the server notifies the client (error or info level) and the client displays the message as a toast (e.g., "booting up kernel" etc.)
• import SPLOT models (has a rich feature model repository)
• move subtree only shows non-collapsed features, is this intentional?
• the vertically split constraint view vanishes in Firefox when the command palette is opened
• collapsing, then expanding a feature does not remove the feature number indicator below