LitterBox: A Linter for Scratch Programs

Abstract

Creating programs with block-based programming languages like SCRATCH is easy and fun. Block-based programs can nevertheless contain bugs, in particular when learners have misconceptions about programming. Even when they do not, SCRATCH code is often of low quality and contains code smells, further inhibiting understanding, reuse, and fun. To address this problem, in this paper we introduce LITTERBOX, a linter for SCRATCH programs. Given a program or its public project ID, LITTERBOX checks the program against patterns of known bugs and code smells. For each issue identified, LITTERBOX provides not only the location in the code, but also a helpful explanation of the underlying reason and possible misconceptions. Learners can access LITTERBOX through an easy to use web interface with visual information about the errors in the block-code, while for researchers LITTERBOX provides a general, open source, and extensible framework for static analysis of SCRATCH programs.

Code Perfumes: Reporting Good Code to Encourage Learners

“…Code smells also lower the likelihood of a project being remixed and thus hurt the Scratch community [34]. In order to find code smells in Scratch projects there are automated tools to analyse the programs like Hairball [3], Quality Hound [33], and LitterBox [8].…”

“…As testing as well as verification require some form of specification, a more accessible approach to support learners is through static code analysis tools (i.e., linters), which can recognise common errors based on rules and patterns [16,25]. In particular, misconceptions of learners often manifest in bugs that follow recurring bug patterns which tools like LitterBox [8] can automatically detect: A bug pattern is a code idiom that is likely to be a defect [14], and bug patterns for Scratch have been demonstrated to occur frequently [9].…”

“…However, not all bug patterns imply solution patterns. For example, the Terminated Loop bug pattern [8] checks if there are loops that stop during the first iteration due to a stop block without guarding condition. The absence of a stop block in a loop would already be sufficient to avoid the bug pattern, but it hardly justifies positive feedback, as that would simply commend the absence of bugs and not the presence of a solution pattern.…”

“…♠ Nested Loops: The presence of nested loops, where the inner one is accompanied by other blocks preceding or following it to not have a Nested Loop smell [8], might be indicative for attempts to implement advanced control flow [32].…”

“…To address this problem, program analysis tools such as Litter-Box [8], Hairball [3] and Quality Hound [33] can automatically detect bug patterns and code smells. By identifying and pointing out bugs and code smells in learners' programs such tools can provide support for young programmers during their learning process.…”