Language Migration in non-CS Introductory Programming through Mutual Language Translation Environment

Abstract

In the past decade, improvements have been made to the environments used for introductory programming education, including by the introduction of visual programming languages such as Squeak and Scratch. However, migration from these languages to text-based programming languages such as C and Java is still a problem. Hence, using the OpenBlocks framework proposed at the Massachusetts Institute of Technology, we developed a system named BlockEditor, which can translate bidirectionally between Block (the block language used here) and Java. We conducted an empirical study of this system in an introductory programming course taken by approximately 100 university students not majoring in computer science. When students were given opportunities to select the language to solve their programming assignments, we traced their selection by tracking working time with BlockEditor or Java for each individual student. The results illustrate the nature of the seamless migration from Block to Java, and show that there is great diversity in the timing and speed of migration to Java by each individual. Additionally, we found that students with low selfevaluation of their skill chose to use Block at a significantly higher rate than did students with high self-evaluation. This suggests that BlockEditor can act as scaffolding for students by promoting mixed programming between Block and Java in their migration phase.

Lessons in combining block-based and textual programming

“…Experimentally, we did see users deploy both modes, and seem to understand the connection between the two. Similar to DiSalvo [49], and like Weintrop [50] and Matsuzawa [20] as well, however, we found that a number of participants very strongly preferred the textual mode and were even scornful of the tiled mode and its presence, to the point of using text almost exclusively even when the system design made it more difficult to complete tasks in that way. Three participants explicitly noted unprompted that they were predisposed to dislike GUIs.…”

“…In our experiment, which used exactly parallel languages in both textual and visual modes, and in Matsuzawa et al's [20], which did not, understanding the relation between the two did not seem to be an especial problem, but intimidation by text was evident. Students in Matsuzawa et al's study with lower self-rating of their skill avoided the text view, but none of our self-rating questions showed a strong correlation with use of either mode; we did not have a generic "rate your programming skill" question, however.…”

“…Matsuzawa et al's experiment using both Java and Block, which translates to and from Java, in an introductory programming course finds a wide range of different levels of use of textual and block editing [20]. The BlockEditor system supports exporting to Java and importing from Java, but makes no particular explicit mapping between the two in itself; it does not appear that this caused any widespread trouble for learners, which may suggest that Tiled Grace's emphasis on making the mapping manifest through animation is unnecessary, or it may be a reflection of the teaching structure employed.…”

“…Matsuzawa et al [20] built BlockEditor, an editor for a visual language called Block that can save the program to textual Java. The Java code can be edited and automatically reimported into BlockEditor.…”

“…As in BlockEditor [20], permitting both views avoids the transition event altogether, so that moving from blocks to text becomes a process rather than an event. A programmer can start to move to text as early as the first day, and draw out the process as long as necessary until they are truly comfortable working with textual code.…”