View the Project on GitHub TiarkRompf/cs352

Project 2: Branches - Loops - Error handling

Due 11:59pm Monday Jan 22


In this project, we will enrich the simple arithmetic expression language we defined in the first project.

At the end of the project, we will be able to parse full programs; the only things missing will be functions and arrays. Our parser will generate an intermediate representation in the form of an Abstract Syntax Tree (AST).

Step 1: Getting Started

The project has been designed and tested for Linux/Mac OS. If you only have Windows installed on your personal laptop, consider running Linux in a VM or using the lab machines for the project.

If you use remote access to work on your project, please use one of the lab machines pod1-1 to pod1-20 with the suffix cs.purdue.edu (e.g. pod1-1.cs.purdue.edu)

Download the skeleton file here.

unzip proj2.zip
cd proj2

Step 2: Project Structure

Browse the Files

A description of the different files is provided below. The files you need to complete are Parser.scala, SemanticAnalyzer.scala, Interpreter.scala, and Compiler.scala. Following the path of the first project, we are going to implement the parser step by step. Follow the comments in the code and find the TODOs. The other part are also independent and can be develop separately.


This file defines multiple classes that are used to generate the code and run it on your machine. Nothing needs to be modified, but it is recommanded to read it and have an idea of what is happening behind the scenes.


As we are generating assembly code which is OS dependent, we are using GCC to do the heavy lifting for us. The boostrap file is a generic C file that is calling a function entry_point and is printing the result in stdout. Our compiler will generate the file gen/gen.s and will be assembled and bootstrapped by gcc:

gcc bootstrap.c gen.s -o out

out will then print the result of our compiled expression.


The main function is defined in this file. The data flow in this file can be seen as:

Read from File / Read from Command Line -> Parser -> Semantic Analyzer -> Interpreter or Compiler

You will be implementing many different parsers in order to test them through the main function.

The AST generated will then go through the semantic analyzer. If there are no errors, then the code is going to be interpreted or compiled. We provide you with one interpreter and one compiler. You will have to write one of each yourself.

As our language becomes more complex, it will become more useful to read the code from a source file. A folder with sample files examples/ has been provided. Here some examples how to run the code:

> run
Usage: run PROG [OPTION]
   or: run FILE [OPTION]
OPTION: compX86, intStack, intValue (default)
> run examples/valid_arithm.scala              ---> interpreter with the provided interpreter
> run examples/valid_arithm.scala compX86      ---> x86 compiler (your code)
> run examples/valid_arithm.scala intStack     ---> stack interpreter (your code)

However, it is still possible to run code from the command line, as in project 1:

> run "val x = 5; x"
> run "val x = 5; x" compX86
> run "val x = 5; x" intStack


This class contains the definition of our intermediate language.

As we discussed in class, parsing the file one character at a time is not enough when we introduce more complex constructs. Also, our single digit numbers were a little bit limiting. The Scanner class defined in this file tokenizes the code. You will have to implement the getNum method. Check out the other methods implemented there, as they are going to be useful for the next part.

The rest of the file is the definition of each parser we are building, starting from a generic precedence arithmetic parser to the full language we want to target, including variables, if statements, and loops. What you have to do for this project is highlighted with TODOs in the code. We encourage you to read the comments and code carefully before proceeding.


While the parser is in charge of verifying that the input string follows the defined syntax, the semantic analyzer verifies that the program described is meaningful and follows the rules. For example:

val x = 2; x = 5

has a valid syntax. However 'val' is immutable: it is not allowed to be reassigned to.

Using the functions error and warn, enforce the following semantic rules:

Note that we allow a variable to be declared multiple times. The last definition will be used when a reference to it is found. In this case only a warning is raised.


As we have seen in class, now that our programing language accepts more than just mathematical expressions, we need to define the required behavior of our language. An interpreter can be used to define this meaning.

We provide you with an interpreter that is fully functional. Your job is to complete the Stack-based interpreter.


In this file, one compiler emitting Scala-like code has already been implemented. You need to implement a second compiler that can convert our AST into x86_64 code.

In the previous project, we considered two different ways of generating the assembly code, each of which has pros and cons: Stack-based code, which is not very efficient, but can handle arbitrarily complex expressions; and Register-based code, which is efficient, but can not handle arbitrarily complex expressions. For this assignment, we will use a register-based approach.

Tests src/test/scala/project2/*Test.scala

These files contain some unit tests for the first parsers. You will have to write your own tests for the others. Every task should have at least 2 tests otherwise points will be deducted. There are some functions given to you in order to make the implementation easier.


You should turn in the proj2 directory. Please run an 'sbt clean' and './cleanall.sh' before submitting.

To turn in your project create a ZIP file named <purdueemailusername>-proj<N>.zip of the proj2 directory for example axhebraj-proj2.zip and upload it to the corresponding assignment on Brightspace.

No other file formats or naming conventions will be accepted as submissions. Verify your submission by downloading the ZIP file you uploaded on Brightspace and extracting its content. The uncompressed content should be the proj2 folder containing the code.


Your project will be tested against a set of unit tests. The weights of each task will be distributed as follow:

Task Weight
Scanner.getNum 5%
ArithParser 10%
LetParser 5%
BranchParser 10%
VariableParser 15%
LoopParser 5%
SemanticAnalyzer 10%
StackInterpreter 15%
X86Compiler 25%