drawing-lang-interpreter

Practical assignments for the XDU compiler course: The toy interpreter of function drawing language, written by XDU Zhang Yi(19030500131).

Introduction

This is an interpreter of a function drawing language. The language uses a simple syntax to specify various details of the function to be drawn.

I plan to use C++(C++17) to complete the interpreter and use some third-party libraries, including

• OpenCV: Used to generate a drawn image of the function.
• googletest: Used for unit test.

This is my first attempt to write a toy interpreter, so obviously it will have many shortcomings. I try to make myself able to deal with various practical situations, for which I refer to the implementation of LLVM/clang a lot.

Lex

There are some keywords in the language, which are defined in include/Lex/KeywordDef.h.

• origin
• scale
• rot
• is
• to
• step
• draw
• for
• from
• t

Keywords are not case-sensitive.

There are also some operators:

• ;
• ( and )
• ,
• + - * / and **

The rules for identifiers in the languages are similar to those in most languages. As required, the identifier should consist of letters followed by any number of letters or numbers. I additionally allow underscores(_) in the identifier.

All numbers in this language are floating point numbers and support (optional) decimal point. However, there are some special cases:

1. How to deal with 1.1.1? If the lexer combines 1.1.1 into a complete token to represent a literal value, and then reports that the literal value does not conform to the specification, this will make the error message clearer. However, strictly speaking, 1.1.1 is not a legal token, so the lexer will generate a token with a value of 1.1 and complain about the invalid character ".".
2. How to deal with 0001? The lexer will treat it as a constant.
3. Should the lexer convert the sequence of characters that represent numbers into real floating-point numbers? Here I chose the same approach as clang, that is, no conversion.

This is a hand-written lexer. The lexer will cache the obtained token to implement look_ahead.

Grammar

It will be completed when I write the parser.

Development progress

1. Write basic facilities.

   • string_ref: This is a similar implementation of LLVM StringRef class.
   • file_manager: This class represents a single source file and manages the lifetime of the buffer.

2. Lexer

   • token_kind: Defines the kind of token.
   • KeywordDef.h: Defines all of the keywords.
   • token: Represents a token.
   • lexer: The lexer.

3. Diagnosis

   The diagnostics subsystem is used to send the error or warning information found by the interpreter to the user.

   • diag_data saves the relevant attributes of the diagnosis result.
   • The DiagTypes.h file defines all diagnostic information that can be reported.
   • diag_engine is used to generate a diagnostic object.
   • diag_builder is used to replace the placeholders in the diagnostic message with the actual parameters provided by the interpreter.
   • diag_consumer defines the way to show diagnostic information to the user.