Description

In this part of the project, we will add error handling to the parser and create a semantic analyzer. At the end of this phase, we will generate quadruples of three-address code as our intermediate                           representation.

Though it is described last, the IR generation is the most important part of this project as it will allow you to continue with your code base in part 3c. Make sure that you generate at least some IR.

Error Handling in the Parser

JavaCUP supports a special “error” token that a production can go to. For example, we may have an error in a statement and thus we would add a production such as:

The semicolon (and additional tokens afterward) is a point of synchronization that gets the parser back on track by discarding erroneous input off of the stack.

Parser Error Handling Requirements

For your project, handle the following four error situations:

1.    Statements synchronized by a semicolon

2.   Variable declarations synchronized by a semicolon

3.    Class and method bodies synchronized by a }

4.    Formal and actual parameter lists synchronized by a )

For each parser error (regardless of using the error symbol), report the line and character position that the error occurred on:

Name Analysis

With the AST that the first part of the compiler generated, we can now do the name and type analysis for MiniJava.

Name analysis discovers identifier names that are improperly defined.

We will handle two types of errors in this step:

1.    Redefinition of an identifier in a given scope

2.    Use of an undefined identifier

To do name analysis and later parts of the project, you will need to construct a symbol table. You should

use either the tree of hashtables or the stack of hashtables to represent your nested scopes.   As you build the symbol table, or in a second pass, you may encounter the errors listed above.

Name Analysis Requirements

If you discover more than one declaration of an identifier in a scope, output the message:

If you discover that an undefined identifier is used, output the message:

Type Checking

With AST and Symbol Table in hand, we next need to do type checking.

We will identify the following type errors:

1.   Assignment to a class or method name

2.   Assignment from a class or method name

3.   Attempting to call a class or method name

4.    Calling a method with the wrong number and/or types of arguments

5.   Applying an operator to the wrong type

6.    Using a non-boolean expression in an if or while condition

7.   Type mismatch during assignment

Type Checking Requirements

If there is an attempt to assign to a class or method name, or the keyword this, output the message:

If there is an attempt to assign from a class or method name, output the message:

If an operator is applied to a class or method name, output the message:

If an attempt to call something that isn’t a method (class, variable, etc.) is made, output:

If a method is called with the wrong number of arguments:

If a method is called with the wrong type of argument:

If you attempt to use an arithmetic or comparison operator (<, +, -, *) with non-integer operands, output:

If you attempt to use a Boolean operator with non-boolean types (&&), output:

If you attempt to use .length on any type other than int [], output:

If the condition of an if or while statement does not evaluate to a boolean, output:

If the right hand side of an assignment does not match the type of the left hand side, output:

If you try to use this in a static method (main), output:

Note:

•    For each error, the character output should be the index of the first letter of the offending code

•    System.out.println is declared as taking a single int parameter

IR Generation

The last part of this phase will be to generate IR code. We will generate quadruples of three-address code.

These are the IR statements to use:

You will likely want the operands as pointers to entries in the symbol table. The keyword “this” is a local variable implicitly defined as the first parameter in each non-static method. When calling a non-static    method, pass the left hand side of the dot operator as the value of parameter this.

For example:

should be translated into:

Treat true and false as constants of your choosing.