For this project, you need to install Java, Gradle, and Git (and optionally, a Git GUI client, such as TortoiseGit or GitHub Desktop). Please check the compatibility matrix for Java and Gradle versions.
This project's main goal was to acquire and apply the theoretical principals of the Compilers' course unit. This was achieved by building a compiler for programs written in the Java-- language. The main parts of the project are Syntactic error controller, Semantic analysis, OLLIR and Jasmin Code generation.
There are three important subfolders inside the main folder. First, inside the subfolder named javacc you will find the initial grammar definition. Then, inside the subfolder named src you will find the entry point of the application. Finally, the subfolder named tutorial contains code solutions for each step of the tutorial. JavaCC21 will generate code inside the subfolder generated.
To compile and install the program, run gradle installDist. This will compile your classes and create a launcher script in the folder ./build/install/comp2022-00/bin. For convenience, there are two script files, one for Windows (comp2022-00.bat) and another for Linux (comp2022-00), in the root folder, that call tihs launcher script.
After compilation, a series of tests will be automatically executed. The build will stop if any test fails. Whenever you want to ignore the tests and build the program anyway, you can call Gradle with the flag -x test.
To test the program, run gradle test. This will execute the build, and run the JUnit tests in the test folder. If you want to see output printed during the tests, use the flag -i (i.e., gradle test -i).
You can also see a test report by opening ./build/reports/tests/test/index.html.
For the first checkpoint the following is required:
- Convert the provided e-BNF grammar into JavaCC grammar format in a .jj file
- Resolve grammar conflicts, preferably with lookaheads no greater than 2
- Include missing information in nodes (i.e. tree annotation). E.g. include the operation type in the operation node.
- Generate a JSON from the AST
To help converting the JavaCC nodes into a JSON format, we included in this project the JmmNode interface, which can be seen in src-lib/pt/up/fe/comp/jmm/ast/JmmNode.java. The idea is for you to use this interface along with the Node class that is automatically generated by JavaCC (which can be seen in generated). Then, one can easily convert the JmmNode into a JSON string by invoking the method JmmNode.toJson().
Please check the JavaCC tutorial to see an example of how the interface can be implemented.
We also included in this project the class src-lib/pt/up/fe/comp/jmm/report/Report.java. This class is used to generate important reports, including error and warning messages, but also can be used to include debugging and logging information. E.g. When you want to generate an error, create a new Report with the Error type and provide the stage in which the error occurred.
We have included the interface src-lib/pt/up/fe/comp/jmm/parser/JmmParser.java, which you should implement in a class that has a constructor with no parameters (please check src/pt/up/fe/comp/CalculatorParser.java for an example). This class will be used to test your parser. The interface has a single method, parse, which receives a String with the code to parse, and returns a JmmParserResult instance. This instance contains the root node of your AST, as well as a List of Report instances that you collected during parsing.
To configure the name of the class that implements the JmmParser interface, use the file config.properties.
The project is divided in four compilation stages, that you will be developing during the semester. The stages are Parser, Analysis, Optimization and Backend, and for each of these stages there is a corresponding Java interface that you will have to implement (e.g. for the Parser stage, you have to implement the interface JmmParser).
The testing framework, which uses the class TestUtils located in src-lib/pt/up/fe/comp, has methods to test each of the four compilation stages (e.g., TestUtils.parse() for testing the Parser stage).
In order for the test class to find your implementations for the stages, it uses the file config.properties that is in root of your repository. It has four fields, one for each stage (i.e. ParserClass, AnalysisClass, OptimizationClass, BackendClass), and initially it only has one value, pt.up.fe.comp.SimpleParser, associated with the first stage.
During the development of your compiler you will update this file in order to setup the classes that implement each of the compilation stages.
Finished developing missing parts of the compiler.
The following verifications are implemented in the typeverfification folder : (_src/pt/up/fe/comp/analysis/analyser/typeverfification)
-
ArrayIndexNotIntCheck - In a array access it checks if the index is of type int or not, if it isn't an error is created AssignmentsCheck - Checks in every assignment if the types match in order for it too be correct
-
AssignmentsCheck - Checks in every assignment if the types match in order for it too be correct
-
IncompatibleArguments - Checks in a call to a function if the parameters of the function being called match with the arguments of the call
-
IncompatibleReturn - Checks if the return of a function has the same type as the declaration of the same function. OperatorsCheck - Checks if the types are correct in the "Times", "And", "Plus", "Minus", "Less" , "Divide", "Not" operands
-
WhileIfConditionCheck - Checks if the conditions presented in If and While statements are of type boolean
The following verifications are implemented in the methodverification folder : (src/pt/up/fe/comp/analysis/analyser/methodverification)
-
CallToUndeclared method - checks whether or not a method is declared, if it is imported from another class or if the Super class extends another class which contains the method
-
ThisCallCheck - Checks if there is any instance of this on the main method, because the main method is static
This step consists in generating OLLIR code derived from the AST and Symbol table and finaly generating the set of JVM instructions to be accepted by jasmin.
While & If's - These nodes are implemented using the same structure. A condition is generated with its opposite operand. That way, when the new condition is met, the code skips the normal instructions , breaking the loop or parsing the else statements.
Assignments, Array Access & Operations - These statements are structured in such a way, so that when a temporary variable is created, a prefix statement(with the auxiliary variable set up) and in another line, the new temporary variable(to be parsed by the parent afterwards ) are added to the ollir string
Variable generation - The temporary variables are generated incrementaly with a "t" tag and it's respective number.
Types - All the types generated by ollir, returns, identifiers, functions... are evaluated by searching on the symbol table for the type in question. Checks were created to identify variables with the same name.
Fields - The field generation and access was also developed considering many cases. For example, when storing a value in an index of a field array all the procedures are taken to generate valide code.
Disclaimer : Every feature was implemented accordingly to the pretended standarts, which means any valid input should work.
While & If's - Condition loops were implemented accordingly, each if clause if covered and the Backend is able to generate if clauses and loops.
Instructions - Every java-- instruction was covered, and the backend can generated a .j file correspondent with the code given.
Optimizations - Our compiler generates JVM code with lower cost instructions.
Disclaimer : Every feature was implemented accordingly to the pretended standarts, which means any valid input should work.
The highlights of our project are:
- Organized code
- Complete and detailed AST
- Robust semantic analysis
- Jasmin Optimizations
Since we hadn't enough time, we didn't implement all the optimizations, which would have made this an even better project.
| Name | UP | Contribution |
|---|---|---|
| André Santos | up201907879 | 25% |
| João Andrade | up201905589 | 25% |
| Sérgio Estêvão | up201905680 | 25% |
| Sofia Germer | up201907461 | 25% |