Mercem: Method Name Recommendation Based on Call Graph Embedding

Abstract

Comprehensibility of source code is strongly affected by identifier names, therefore software developers need to give good (e.g. meaningful but short) names to identifiers. On the other hand, giving a good name is sometimes a difficult and timeconsuming task even for experienced developers. To support naming identifiers, several techniques for recommending identifier name candidates have been proposed. These techniques, however, still have challenges on the goodness of suggested candidates and limitations on applicable situations. This paper proposes a new approach to recommending method names by applying graph embedding techniques to the method call graph. The evaluation experiment confirms that the proposed technique can suggest more appropriate method name candidates in difficult situations than the state of the art approach.

Learning Based Methods for Code Runtime Complexity Prediction

“…Variable/Method name prediction is a widely attempted problem, wherein Allamanis et al [4] used a convolutional neural network with attention technique to predict method names, Alon et al [5] suggested the use of AST paths to be used as context for generating code embeddings and training classifiers on top of them. Yonai et al [19] used call graphs to compute method embeddings and recommend names of existing methods with function similar to target function.…”

Impact of Evaluation Methodologies on Code Summarization

“…Around 2016, the first work for comment generation [1] and method naming [15] were developed based on encoderdecoder neural networks and attention mechanism. Other prior work extended this basic framework in many directions: by incorporating tree-like code context such as AST [2], [4], [5], [8], [12]; by incorporating graph-like code context such as call graphs and data flow graphs [3], [11], [21], [22]; by incorporating path-like code context such as paths in AST [16], [17]; by incorporating environment context, e.g., class name when generating method names [18]; by incorporating type information [10]; or by using more advanced neural architecture such as transformer [9].…”

“…Over the last several years, there has been a growing interest in applying machine learning (ML) models to tasks that process source code and the natural language elements, such as comment generation [1]- [12], code generation [13], [14], method naming [11], [15]- [18], and code completion [19]. This growing body of work has introduced sophisticated models based on advanced ML, such as deep neural networks [1], [2], [8], [15]- [17], graph neural networks [3], [11], [20]- [22], and transformers [9], [23]. Substantial progress has been reported over years, usually measured in terms of automatic metrics, including BLEU [24], precision, recall, and F1 scores.…”

DeepTC-enhancer

“…Allamanis et al [2] introduce a neural probabilistic language model for source code that can suggest method names. In addition, Yonai et al [45] propose an approach Mercem to recommend method names in source code by applying graph embedding techniques to the call graph. Both approaches target methods whereas DeepTC-Enhancer focuses on suggesting names for automatically generated test cases.…”