Research

badge

As a junior researcher, I’m interested in two primary subjects: automated project management and object-oriented programming. You can find me in Scopus, Google Scholar and DBLP too. I’m also a co-organizer of ICCQ.

If you are a student and looking for a research topic for your diploma, pick one of the following and text me in Telegram: @yegor256. I will help you to do the research and may also play a role of your scientific advisor.

Jumps Elimination in OOP: Recently, we developed DeJump, an automated tool that modifies programs written in EOLANG, our own object-oriented language, such that “jumping objects” (GOTO, BREAK, and CONTINUE) are replaced with recursion or other declarative mechanisms. A research paper about the tool was recently accepted at a computer science conference (CSAI’23). However, even though the tool exists, it has not been tested with a larger amount of source code, and it needs a few important improvements in its algorithm. Thus, we suggest creating the second version of the tool, with the required improvements and testing it with larger blocks of code, in order to demonstrate its effectiveness. A new research paper is expected to be written and published.

SLR of Object Models: There is a famous λ-calculus that underlies functional programming languages, like Haskell or Scala. At the same time, for object-oriented programming languages, more than one calculus has been introduced over the last thirty years, such as Object-Z and VDM. However, most modern object-oriented languages, such as Java 18+ or Swift, may not be fully formalized with the help of said calculi. There are certain limitations in every calculus. In 2021, we made an attempt to conduct a detailed analysis of existing object-oriented formal models in the form of a Systematic Literature Review (SLR). Unfortunately, we didn’t have enough time to study each calculus deeply enough, which is why we stopped the research project. Now, we suggest continuing this work, researching existing models, and publishing a full academic paper. This paper, if published in a reputable journal or conference, could become a significant contribution to the theoretical sector of object-oriented programming.

Algorithmization of Object-Oriented Programs Written in EO: EOLANG, our experimental strictly object-oriented programming language, is very effective for modeling objects and their relationships. However, it falls short when programming algorithms and mathematics, because the code tends to run much slower than its procedural counterparts written in C++ or Rust. We suggest creating a code optimization tool, which will identify blocks of code inside EO programs that can be translated to Rust to enhance performance. Thus, we aim to develop a tool that inserts native code blocks into the original EO program, bridging two programming paradigms—object-oriented in EO and procedural in Rust—by connecting them both ways through the JNI interface. The main challenge lies in pinpointing the most suitable code blocks; their introduction should increase the overall program performance while considering the additional overhead introduced by JNI roundtrips.

ChatGPT in Backlog Prioritization: There is an existing open-source tool, created more than five years ago, that assists GitHub programmers in decomposing their tasks into smaller ones and automatically creating sub-tasks: 0pdd. We aim to enhance this tool with Machine Learning, enabling it to prioritize tasks more effectively. Additionally, we want to integrate ChatGPT into the tool to automatically estimate the value of each task and suggest fast-tracks for those implementing tasks. We have already made a theoretical attempt to identify an effective solution and have published our findings in the IEEE Access journal. Now, implementing these findings in practice and incorporating ChatGPT could represent a significant contribution to computer science and provide invaluable assistance to hundreds of GitHub programmers.

ChatGPT for GitHub Anomalies Detection: The quality of source code and the discipline of repository maintenance play a crucial role in the success of a software product. However, there is no automated tool on the market that appraises the quality of an entire repository and generates recommendations for its maintainers, with a focus on architectural issues and process flaws. In 2021, we attempted to create such a tool and even published some theoretical findings in this direction. However, we didn’t employ ChatGPT, as it didn’t exist at that time. Now, we believe it would be beneficial to combine our project analysis algorithm with the capabilities of LLM, to generate the most effective and useful advice for software teams.

Analysis of Heap Performance in Different OSs: The heap is the primary storage for variable-sized memory blocks in modern operating systems and virtual machines. Allocating a slice of bytes in the heap and then releasing it back is a time-consuming operation, requiring several hundred CPU cycles. However, the exact number of cycles it takes to allocate and free memory chunks in different virtual machines and OSs remains unclear. We suggest studying this subject, performing experiments on a sufficiently large number of testing platforms, summarizing and analyzing the results, and then publishing a research paper. Such an analysis might assist creators of programming languages and compilers in making better design decisions.

Encapsulation Strategy in Design Patterns: In object-oriented programming, many design patterns are recommended for use. It’s commonly believed that if programmers adhere to these patterns in their code, the code quality will improve due to clearer design. We hypothesize that most design patterns emphasize the encapsulation of behaviors rather than data. In other words, the objects participating in design patterns are typically “dataless” objects. We propose studying this subject through a Systematic Literature Review (SLR) of existing literature on design patterns, aiming to either confirm or refute our hypothesis. The results of our research might be useful for compiler and programming language designers, prompting them to treat objects differently if they are dataless.

Study Programmers’ Preferences for Object Extensions: In object-oriented programming, additional functionality can be added to classes using inheritance, decoration, composition, or by simply expanding existing classes with new code or methods. It is commonly believed that most programmers, especially those with over 10 years of practical coding experience, prefer decoration or composition, as these methods typically result in superior design. To validate this assumption, we propose conducting a survey among a sizeable group of programmers. We’ll present them with various code snippets and ask them to choose a method for modifications. The findings from our research might offer insights to designers of new programming languages about programmers’ perceptions of OOP.

Cost of OOP in Different Languages: Earlier studies by various researchers and practitioners have indicated that object-oriented programming features, such as encapsulation and polymorphism, are more performance-intensive compared to their procedural counterparts (like static methods and global variables). However, a direct comparison between programming languages has not been undertaken. We propose such a comparison to determine in which object-oriented programming language programmers incur the highest performance cost for using the object paradigm. The results of this research might assist other researchers in better understanding the practical usability of object-oriented programming.

You may also pick one of our open source projects, and extend it with a new feature, presenting this work as engineering+science:

In each of them, there are over a hundred unresolved issues and pull requests. I can help you figure our how to formulate your coding contribution such that it would look like a diploma work.

sixnines availability badge   GitHub stars