Two-Dimensional Parson’s Puzzles: The Concept, Tools, and First Observations

Abstract

Executive SummaryParson's programming puzzles are a family of code construction assignments where lines of code are given, and the task is to form the solution by sorting and possibly selecting the correct code lines. We introduce a novel family of Parson's puzzles where the lines of code need to be sorted in two dimensions. The vertical dimension is used to order the lines, whereas the horizontal dimension is used to change control flow and code blocks based on indentation as in Python. Python blocks have no explicit begin/end statements or curly braces to mark where the block starts or stops. Instead, indentation is used to define starts and stops of blocks and functions.In addition, we introduce tools supporting two-dimensional Parson's puzzles: (1) MIT licensed JavaScript widget to embed our puzzles to any HTML, and (2) server to create, share, and solve puzzles.We have observed how experienced programmers solve our puzzles. Such users often start by dragging the method signature to the beginning and continue by defining majority of the control flow (i.e., loop statements, assignments, conditional statements). Only after these are done, details, including initialization of variables and handling of corner cases, are dragged to correct positions in the middle of the previously structured code. This shows that even experts are not able to solve puzzles linearly, i.e., line by line, starting from the first. Thus, user interfaces (UIs) should minimize the work needed to insert a line between two adjacent lines of existing code. In some of the existing Parson's Puzzle UIs this is not the case.Another observation we made is that too often users don't ask or use automated feedback. Why this happens needs further investigations. Perhaps experienced users are too proud to ask a tool to help them (especially when being observed), or perhaps users don't recognize when they are stuck and should ask for help. Providing constant feedback is one way to tackle this problem. However, the obvious downside of the constant feedback is that solving an exercise can become trial-and-error repetition.

Increasing Adoption of Smart Learning Content for Computer Science Education

“…Js-parsons (Figure 16), is used to give automated feedback from visual code construction exercises [40]. These are a type of scaffolded program construction tasks where the learner is given a set of code fragments, blocks of a single or multiple lines of code, and from these the task is to piece together a program.…”

“…• Problem-solving support tools (e.g., Problets [56,57], js-parsons [40,39], WadeIn [11]) that have the student learn concepts by solving targeted problems. Many of these are adaptive (e.g., Problets [55], ELM-ART [12], JavaGuide [36]) and can be classified as intelligent tutoring systems.…”

How Do Students Solve Parsons Programming Problems? -- Execution-Based vs. Line-Based Feedback

“…There may exist multiple possible arrangements that meet the requirements. Avoiding such situations is one of the challenges in designing non-trivial Parsons problems when using this approach to assess correctness [2].…”

“…Indeed, originally, Parsons problems were implemented on top of a multiplechoice question engine called Hot-Potatoes [1]. A similar assessment approach has also been utilized in js-parsons [2] and its mobile extension MobileParsons [3] -although in both of these tools code fragments are placed on a two dimensional grid and this requires some preprocessing of the data (e.g. normalizing the indentations).…”

“…Typically, a fragment is a single line of code but may also consist of multiple consecutive lines. Previously, we have implemented a drag-and-drop web environment for Parsons problems in Python called js-parsons [2]. In Python, code blocks are defined by indentation and an indented statement falls into the block of a surrounding (control) structure that has lower indentation.…”