Java的語法樹直譯器

A Syntax-Tree-Based Interpreter for Java

10.29715/JIMCSA.200806.0006

周敬斐(Ching-Fei Chou)；郭肇安(Chao-An Kuo)；廖賀田(Heh-Tyan Liaw)

Java ； 語法樹 ； 直譯器 ； Java ； Syntax Tree ； Interpreter

資訊管理展望

10卷1期（2008 / 06 / 01）

111 - 135

繁體中文

Java程式語言具有物件導向(object-oriented)、可攜(portable)與多執行緒(multi-threading)等優點，是當今最風行的程式語言之一。目前所有的Java 原始檔都必須被編譯成稱為位元組碼(bytecode)的中間語言(intermediate language)，再交由Java虛擬機(java virtual machine)進行直譯。但位元組碼與Java的原始碼之間並沒有本質上的關聯，位元組碼在直譯上也沒有特別便利之處。 本論文主張以Java語法樹(syntax tree)當成中間語言，並提出對它進行直譯(interpreting)的技術。語法樹是製作剖析器(parser)時就必須用到的資料結構，它在先天上就和原始碼有最密切的關連。我們直接在語法樹的各種節點上配置適當的方法(method)，運用遞迴呼叫(recursive call)與動態繫結(dynamic binding)來製作直譯器。我們也利用例外機制(exception mechanism)來脫離進入多層呼叫後的控制流程(control flow)。為配合目前被廣泛使用的位元組碼，我們還提出語法樹與位元組碼之間的銜接機制，讓這兩種不同的直譯機制可以互相呼叫。 本論文所提出的直譯機制非常容易製作，是開發直譯器的優良途徑。我們所實作的直譯器在大型程式上的真實運作已有良好的表現。由於語法樹與原始程式之間的對應既直接又自然，未來在製作除錯(debug)機制時也會比較容易。此外，還可以進一步在語法樹上配置圖形展示的能力，讓直譯器在程式語言與資料結構的教學上對學生提供直覺性的幫助。

The Java programming language, with advantages such as object-orientation, portability and multi-threading, is one of the most popular programming languages. In current time, the Java source code has to be compiled into an intermediate language, the bytecode, before it can be interpreted by the Java virtual machine. However, there is no intrinsic relationship between bytecode and its source code, and there is no special benefit in interpreting the bytecode either. In this paper, we recommend adoption of the syntax tree to be the intermediate language for Java, and we propose techniques for interpreting it. The syntax tree is an essential data structure when parsing the source code. Furthermore, it bears strongest relation with the source code. By appending suitable methods into various types of nodes in the syntax tree, the interpreter can be implemented through recursive call and dynamic binding. We also utilize the exception mechanism to quit from the control flow of deep method calls. In order to cooperate with the popular bytecode, we propose a bridge mechanism to interconnect the interpretations of both bytecode and the syntax tree. It is very easy to build an interpreter based on the syntax tree. Our syntax-tree-based interpreter for Java has been successfully implemented, and it works well on interpreting practical Java software. Since the correspondence between the syntax tree and the source code is direct and natural, it will be easier to implement a debugging mechanism in the future. Furthermore, we can equip graphical rendering functions into the syntax tree, so that the interpreter can provide instinctive assistance in teaching the Java programming language and data structures.