100 likes | 204 Vues
This study examines how the implementation of a meta-model affects the time spent by volunteers solving programming problems across different systems. We found that the average time saved was 43% when using the meta-model, especially among new programmers. The experiment involved 6 volunteers who were tasked with writing simple programs in Java, with a focus on code correctness with minimal debugging. Results highlighted the effectiveness of sample code usage and the importance of naming conventions in reducing problem-solving time.
E N D
Time spent (min) 10 20 0 10 20 0 P1 F F D B B B B D A B B B D D D D E C C D B D E F D B C C A A A E C C B B A D F F C C F D A E B A A E C F F A E A E B C C A F D F A F F A C E P2 P3 P4 P5 E F E P6 E P7 E P8 C P9 E P10 P11 E P12 B D F D C A avg
Time spent (min) 10 20 0 10 20 0 P1 F F D B B B B D A B B B D D D D E C C D B D E F D B C C A A A E C C B B A D F F C C F D A E B A A E C F F A E A E B C C A F D F A F F A C E P2 P3 P4 P5 E F E P6 E P7 E P8 C P9 E P10 P11 E P12 B D F D C A avg
Time spent (min) 10 20 0 P1 E C A D B F P2 C A E D B F P3 A C E D B F P4 E C A B D F P5 D F A B C E P6 C E A B C E P7 C D A F B E P8 B D E F A C P9 D B E C A F P10 D B E A C F P11 F C B A D E P12 F B C D E A avg B D A C F F D F C A E B
Experiment design • 4 systems, 12 problems • Problem example, correct classes • 6 volunteers, friends of the authors • 3 gr, 1 undergr, 2 eng, 4 Java, 2 not (c++/php) • No one know data source, nor participate the work • 2 minutes to train • Write a simple program to solve the problem • Without debugging, we check if it is correct • 10 lines to 30. • All the four sources, two with meta-model, others source • How to calculate time • All the time spent, divided, 2-5 minutes training, divided • 20 minutes, unsolved -> 30 minutes (subsequent operations)
Experiment process • With meta-model • All use the meta-models, browse and choose • Some check, some choose to believe after the first • Extreme sample • Relations, browse nearby, reuse the prefix code • Without meta-model • Most choose sample client and get result • All volunteers use code search, cross-ref, but many complaint for suffering the problem of naming rules, may fail • Do not notice the nearby code • Some volunteer (D) choose other sample XML data, but blocked by dealing with the node list • One volunteer (E) was scared by the big scale of client,
Result (in summary) • Saved 43% time in average • Horizontal • All improved (due to copy/paste) • For related problems, save significantly (2/6) • For the one with big size, significantly • For the one with good names, significantly • Vertical • Stable with meta-model • Especially important for a freshman (PHP/DOM)
23 31 28. 1505, ExistsClientTest • 19 29 23. 1320, CoreMonitoringTest • 132 114 131. 2562, JabRefTest • 9 18 14. 2807, GeoRssTest • 84 308 112. 12516, JonasAdminTest • 25 81 25. 5142, FlickrjTest • 11 26 7. 1780, JBossStatusTest • 36 21 53. 1253, WSDLHelper
http://www.schemacentral.com/sc/ooxml/s-shared-bibliography.xsd.htmlhttp://www.schemacentral.com/sc/ooxml/s-shared-bibliography.xsd.html