Rum Compiler

Yo-ho-ho and the Compilation of Rum!

ITMO compilers course project: Implementation of Compiler for Rum language.

• Compiler
  • Getting Started
    • Build
    • Run
    • Build Dependencies
  • Rum Language
    • Syntax
  • Features
    • Types
    • Variables
    • Operators
  • Expressions
    • Operator Precedence Table
  • Statements
  • Standard library
  • Parser
  • AST
  • Environment
  • Interpreter
  • Stack Machine
    • Translation
    • Interpretation
  • Compiler
  • Testing
    • Running the tests
  • License
  • Acknowledgments

Getting Started

To run on your local machine follow the next steps:

Build

To build the Compiler using stack try next command:

stack build

Run

Then run the .rum file with the option you need

stack exec Compiler -- -flag file_name.rum

where next flags are available

• -i — to interpret the program from file_name.rum
• -s — to compile the program from file_name.rum into the presentation of the stack machine with the further interpretation of the stack machine
• -c — to compile the program from file_name.rum into executable file_name

Build Dependencies

Consider next requirements

• stack
• LLVM 4.0

Rum Language

Rum is an artificial dynamically typed programming language, which was developed in learning purposes.

Syntax

Simple program example

fun fact(n : int) : int
begin
  if n < 2 
     then return 1
     else return n * fact(n-1)
  fi
end

n := read();
write(fact(n))

For the further examples let's assume

S := "Yo-ho-ho, and the Bottle of Rum!"

Features

Next features were implemented for the Rum language.

Types

• int — integer
• char — symbol
• str — string (reference type)
• arr — array (reference type)
• void — void type
• bool — boolean (false and true are represented as 0 and 1 respectively, the integer not equal to 0 is considered as true)

Variables

All variable names must begin with a letter of the alphabet or an underscore _ or dollar sign $. After the first initial letter, variable names can also contain letters, numbers and one of these symbols: _, $. Keywords can not be used as variable names. In Rum language these words are reserved: skip, if, then, else, fi, repeat, until, do, od, while, for, fun, begin, end, return, true, false.

Note: All references are considered to start with capital letter when all other variables from with lowercase.

Operators

• Arithmetic : +, -, *, /, %, ^
• Logical : ==, !=, <=, <, >=, >
• Comparison : &&, ||
• Assignment : :=
• Unary : -
• Void : skip

Expressions

Rum language supports different kind of expressions with consideration of precedence and associativity

• Variables — local and global variables support.
• Operations
  • Binary — operations of any nesting level using arithmetic operators
  • Logical operations — using logical operators
  • Comparison operations — using comparison operators
  • Unary operations — using unary operators
• Functions call can be expression and require to evaluate the value to use it in statements

  S := strmake(5, 'm')
  

Operator Precedence Table

Statements

Each statement is separated by ; except the last one in the program, where the program is finished list of statements. Rum implements next types of statements:

• Assignment

  x := 911;
  y := read();
  z := x && (y * 2 - 5 ^ 1);
  ar := {1, 2, 3}
  

• Function call — can be separated statement also, e.g.

  write(42)
  

• Void statement skip
• Conditional statement — several kinds of if statements are supported

  if strcmp(Bottle, "Rum") == 0 
      then yell("Yo-ho-ho!")
  fi
  

  also alternative ending supported

  if strcmp(Bottle, "Rum") == 0 
      then yell("Yo-ho-ho!")
      else Mood := "sad" 
  fi
  

  and unlimited nesting level of alternative with elif

  if strcmp(Bottle, "Rum") == 0 
      then yell("Yo-ho-ho!")
      elif strcmp(Language, "Rum") == 0 
          then 
              yell("Yo-ho-wait..");
              Mood := "exited"
          else Heart := "broken"
  fi
  

• Loops
  • Repeat Until

  repeat
      yell("Yo-ho-ho!");
      drink();
      rumQuantity := rumQuantity - tumbler
  until rumQuantity == 0;
  

  • While

  while (n != 0)
  do
      sing(strcat(n, " men on the dead man's chest"));
      n := n - 1;
      sing("Yo-ho-ho, and the Bottle of Rum")
  od
  

  • For

  sum := 0;
  for i := 1, i <= 100, i := i + 1
  do
      sum := sum + 1
  od
  

  • Function — all function declarations should be done in the beginning of the file. Important not to separate functions with any special symbols.

  fun lang(L : str) : str
  begin
      if (strcmp(L, "Rum") == 0)
      then return "☠️"
      else return "😞"
  end
  

  • Return

Standard library

Rumlude is the library of standard functions of the language. RumludeFuns is the special data type for representing them during parsing to AST.

In the Compiler.Rum.Internal.Rumlude module there is a map from string to RumludeFuns type, so it can be used at parsing stage. The main function of the module is runRumlude which takes as arguments RumludeFuns representative and the list of the parameters, that are going to be arguments of the given function, and returns the result of the work of the standard function.

Here is the list of functions that are based in module with usage examples:

• read()

  — reads integer from standard input

• write(n : int)

  — prints integer n into standard output.

• strlen(S : str) : int

  — returns length of the given string s

  write(strlen(S))
  > 32
  

• strget(S : str, i : int) : int — returns the symbol of the string on the given position i converted into int

  write(strget(S, 1))
  > 111  -- letter 'o'
  

• strsub(S : str, start : int, len: int) : str — returns string, which is build by taking len symbols starting from start position

  D := strsub(S, 15, 18)
  
  > the Bottle of Rum!
  

• strdup(S : str) : str — returns the duplicate of the string

  D := strdup(S)
  
  > Yo-ho-ho, and the Bottle of Rum!
  

• strset(S: str, i : int, ch : char) : str — replaces i-th symbol in the string S with symbol ch

  strset(S, 1, 'u')
  > Yu-ho-ho, and the Bottle of Rum!
  

• strcat(S1 : str, S2 : str) : str — returns the result of concatenation of two strings

  strcat("Yo-ho-ho, ", "and the Bottle of Rum!")
  > Yo-ho-ho, and the Bottle of Rum!
  

• strcmp(S : str, D : str) : int — returns the result of lexical comparison of two strings. The result is equals to 0 if strings S and D are equal, -1 if S is less than D and 1 in other case

  write(strcmp(S, D));
  write(strcmp(S, "Yo-ho-ho"))
  write(strcmp(S, "Yu-ho-ho"))
  write(strcmp(S, "Ya-ho-ho"))
  > 0
  > -1
  > -1
  > 1
  

• strmake(n : int, ch : char) : str — builds the string of length n fulfilled with ch symbol

  strmake(3, 'm')
  > mmm
  

• arrlen(: arr) : int — returns length of the given array

• arrmake(n : int, e : int) : arr — builds unboxed array of the length n fulfilled with the integer e

• Arrmake(n : int, e : int) : arr — builds boxed array of the length n fulfilled with the integer e

Parser

For parsing part of the project monadic parser combinator library MegaParsec. Source files are in the Compiler.Rum.Internal.Parser module of the project. The input is the file and the result of the work of the parser is AST. In case of parsing error the message with additional information (error row and column coordinates) will be shown. (TODO: make beautiful output of errors)

AST

This module gives an idea about the model of the language. Data types and structure of the program can be found in Compiler.Rum.Internal.AST module. Almost all of the types were described before, so here will be the brief enumeration

• Expression
• Statement
• Program — the list of statements

Environment

Different types of environment are used in interpreter and stack machine.

Interpret a newtype was described in Compiler.Rum.Internal.ASTwhich is monad transformer StateT.

newtype Interpret a = 
    Interpret { runInterpret :: StateT Environment (MaybeT IO) a}
    
data Environment = 
    Env { varEnv    :: VarEnv    -- to hold vars
        , refVarEnv :: RefVarEnv -- to hold reference vars
        , funEnv    :: FunEnv    -- to hold custom functions
        , isReturn  :: Bool      -- necessary for funs declarations
        }

For Stack Machine's Environment there is special type in Compiler.Rum.StackMachine.Structure

type InterpretStack = 
    ReaderT ([Instruction], Labels) (StateT StackEnvironment IO) ()
    
    data StackEnvironment = 
        SEnv { vars  :: RefSVarEnv -- store by reference
             , stack :: Stack
             , pos   :: Int        -- current position at Stack
             }

Interpreter

Interpreter receives Program data type and returns evaluated result. All source files are in the Compiler.Rum.Interpreter directory. In Rummer module all evaluation stuff for both expressions and statements take place. interpret function works under Interpret environment and this is main working function for interpreter.

Stack Machine

Work of this part can be separated into two steps.

Translation

Translation of Program into readable by Stack Machine list of Instruction.

Here are permitted instructions

• Nop — No operation
• SBinOp — Binary operation sign
• SLogicOp — Binary operation sign
• SCompOp — Binary operation sign
• Push — Push a value onto the stack
• Pop — Pop the most recent value from the stack
• Jump — Jump unconditionally to a location
• JumpIfTrue — Jump if conditions are met
• JumpIfFalse — Jump if conditions are not met
• Load — Load ordinary variable
• LoadRef — Load reference variable
• Store — Put Variable
• Label — Location coordinates
• SFunCall — Call a function
• SRumludeCall — Call function from standard
• SReturn — Return from a function
• PushNArr — Put array on stack
• LoadArr — Load array
• StoreArr — Store array

Interpretation

Compiler.Rum.StackMachine.Stacker module receives set of Instruction from work of the Translator and transforms data into result. Main function of the module is stacker. It finds the label of starting position in the instructions and start execute them one by one.

Compiler

And finally compilation. Implementation of the Rum language was made with LLVM. Special thanks for Stephen Diehl's work which explains the basics of how to build a compiler in Haskell. For the implementation these two packages are required:

• llvm-hs-pure — pure Haskell representation of the LLVM Intermediate Representation (IR).
• llvm-hs — FFI bindings to LLVM required for constructing the C representation of the LLVM IR and performing optimization and compilation.

Few consistent steps should be done, and the first one is code generation of LLVM IR from the AST. All necessary functions for code generation are located in Compiler.Rum.Compiler.CodeGenmodule and the transformation process is realized in the Compiler.Rum.Compiler.Emmiter module of the project. Compiler.Rum.Compiler.JIT module contains of optimizations for the language.

Testing

There is a bunch of different kind of tests in compiler-tests submodule. Each folder consist of different type of tests:

• core — number of tests to cover all basic features of the language
• expressions & deep-expressions — huge amount of test basically for arithmetic expressions and IO
• performance — tests to compare Compiler efficiency with gcc.

Running the tests

cd to the directory of the test which have to be run and use the next command:

make

This will run tests in the -i, -s and -c mode.

For example,

cd compiler-tests/core && make

License

This project is licensed under the MIT License - see the LICENSE.md file for details

Acknowledgments

Stephen Diehl article "Implementing a JIT Compiled Language with Haskell and LLVM"