Code Completion by Modeling Flattened Abstract Syntax Trees as Graphs

Code completion has become an essential component of integrated development environments. Contemporary code methods rely on the abstract syntax tree (AST) to generate syntactically correct code. However, they cannot fully capture sequential and repetitive patterns writing structural information AST. To alleviate these problems, we propose a new approach named CCAG, which models flattened sequence partial AST as graph. CCAG uses our proposed Graph Attention Block different dependencies in graph for representation learning completion. The sub-tasks are optimized via multi-task task balance is automatically achieved using uncertainty without need tune weights. experimental results show that superior performance than state-of-the-art approaches it able provide intelligent