Varv: Reprogrammable Interactive Software as a Declarative Data Structure

Abstract

Figure 1: Varv Examples: (a) A todo list web application that is inherently extensible. Here, a basic todo list is extended with the ability to complete and delete todos by adding two new concept defnitions and new modifed template defnitions. (b) A board game toolkit that defnes abstractions for board game logic. The games "Checkers" and "Othello" were implemented with the toolkit and then merged into a new "Checkers-O-Thello" game with the addition of a short concept defnition. As Varv applications are represented as data structures, higher-level tooling can be developed including a block-based editor (right), an inspector to go from an element in the view to the corresponding template or data (context menu to the left), and a data inspector for live editing application state (middle).

No Grammar to Rule Them All: A Survey of JSON-style DSLs for Visualization

“…As JSON-style DSLs continue to be developed, it may be useful to explore language workbenches [24], which are a form of tool for designing, composing, and using (often domain-specific) languages. Some work has been done in this direction by JSON Schema structure editors [5]. However, they focus on data validation and not language design.…”

“…None of the external DSLs had SQL-style enduser definable functions. Varv [5] takes extensibility to an extreme via a fully end-user editable application creation external DSL (that includes simple macros); demonstrating that this level of malleability is achievable in external DSLs.…”

“…Beyond familiar uses such as configuration or NoSQL languages (e.g. MongoDB), they are used in domains as varied as statistical analysis [45], web development [5], narrative [15] and game generation [20], chatbots [52], dance [83], and fabrication [106]. We are interested in building a better understanding of JSON-style DSLs precisely because they are so prevalent-although their sudden prominence (Fig.…”

“…This criterion excludes some system types. SVG, HTML, and other high-level markup languages (as well as general-purpose JSON-based DSLs, such as Varv [5]) were excluded because they are capable of producing far more than visualizations. Also excluded were those merely subsetting another language-for instance, GraphScape [50] uses a non-interactive subset of Vega-Lite to explore sequence recommendation.…”

“…However, doing so requires that such hatches be pre-placed in a manner relevant to the new feature-which is difficult to predict. Exploration of DSL malleability in a manner that allows for undesigned features without becoming unapproachably complex for end-users (a concern for the deeply malleable Varv [5]) is a valuable opportunity.…”