Compare revisions

    "version": "2.0.0",

    "tasks": [

        {

            "type": "cargo",

            "subcommand": "build",

            "type": "shell",

            "label": "cargo check --example main_span_expr",

            "command": "cargo check --example main_span_expr",

            "problemMatcher": [

                "$rustc"

            ]

            ],

            "group": {

                "kind": "build",

                "isDefault": true

            }

        },

        {

            "type": "shell",

            "label": "cargo check --example main_span_expr_custom_err",

            "command": "cargo check --example main_span_expr_custom_err",

            "problemMatcher": [

                "$rustc"

            ],

            "group": {

                "kind": "build",

                "isDefault": true

            }

        },

        {

            "type": "shell",

            "label": "cargo run --example precedence",

            "command": "cargo run --example precedence",

            "problemMatcher": [

                "$rustc"

            ],

            "group": {

                "kind": "build",

                "isDefault": true

            }

        },

        {

            "type": "shell",

            "label": "cargo run --example main2",

            "command": "cargo run --example main2",

            "problemMatcher": [

                "$rustc"

            ],

            "group": {

                "kind": "build",

                "isDefault": true

            }

        },

        {

            "type": "shell",

            "label": "cargo run --example tmp",

            "command": "cargo run --example tmp",

            "problemMatcher": [

                "$rustc"

            ],

            "group": {

                "kind": "build",

                "isDefault": true

            }

        }

    ]

}

 No newline at end of file

# This file is automatically @generated by Cargo.

# It is not intended for manual editing.

[[package]]

name = "bytecount"

version = "0.3.2"

source = "registry+https://github.com/rust-lang/crates.io-index"

[[package]]

name = "cfg-if"

version = "0.1.9"

source = "registry+https://github.com/rust-lang/crates.io-index"

[[package]]

name = "first"

name = "crust"

version = "0.1.0"

dependencies = [

 "nom 5.0.1 (registry+https://github.com/rust-lang/crates.io-index)",

 "nom_locate 1.0.0 (registry+https://github.com/rust-lang/crates.io-index)",

]

[[package]]

 "version_check 0.1.5 (registry+https://github.com/rust-lang/crates.io-index)",

]

[[package]]

name = "nom_locate"

version = "1.0.0"

source = "registry+https://github.com/rust-lang/crates.io-index"

dependencies = [

 "bytecount 0.3.2 (registry+https://github.com/rust-lang/crates.io-index)",

 "memchr 2.2.1 (registry+https://github.com/rust-lang/crates.io-index)",

 "nom 5.0.1 (registry+https://github.com/rust-lang/crates.io-index)",

]

[[package]]

name = "rustc_version"

version = "0.2.3"

source = "registry+https://github.com/rust-lang/crates.io-index"

[metadata]

"checksum bytecount 0.3.2 (registry+https://github.com/rust-lang/crates.io-index)" = "f861d9ce359f56dbcb6e0c2a1cb84e52ad732cadb57b806adeb3c7668caccbd8"

"checksum cfg-if 0.1.9 (registry+https://github.com/rust-lang/crates.io-index)" = "b486ce3ccf7ffd79fdeb678eac06a9e6c09fc88d33836340becb8fffe87c5e33"

"checksum lexical-core 0.4.3 (registry+https://github.com/rust-lang/crates.io-index)" = "b8b0f90c979adde96d19eb10eb6431ba0c441e2f9e9bdff868b2f6f5114ff519"

"checksum memchr 2.2.1 (registry+https://github.com/rust-lang/crates.io-index)" = "88579771288728879b57485cc7d6b07d648c9f0141eb955f8ab7f9d45394468e"

"checksum nom 5.0.1 (registry+https://github.com/rust-lang/crates.io-index)" = "c618b63422da4401283884e6668d39f819a106ef51f5f59b81add00075da35ca"

"checksum nom_locate 1.0.0 (registry+https://github.com/rust-lang/crates.io-index)" = "f932834fd8e391fc7710e2ba17e8f9f8645d846b55aa63207e17e110a1e1ce35"

"checksum rustc_version 0.2.3 (registry+https://github.com/rust-lang/crates.io-index)" = "138e3e0acb6c9fb258b19b67cb8abd63c00679d2851805ea151465464fe9030a"

"checksum ryu 1.0.0 (registry+https://github.com/rust-lang/crates.io-index)" = "c92464b447c0ee8c4fb3824ecc8383b81717b9f1e74ba2e72540aef7b9f82997"

"checksum semver 0.9.0 (registry+https://github.com/rust-lang/crates.io-index)" = "1d7eb9ef2c18661902cc47e535f9bc51b78acd254da71d375c2f6720d9a40403"

[package]

name = "first"

name = "crust"

version = "0.1.0"

authors = ["Per Lindgren <per.lindgren@ltu.se>"]

edition = "2018"

[dependencies]

nom = "5.0.1"

nom_locate = "1.0.0"

# Repo for the D7050E course 2019

The repo will be updated througout the course and includes a draft outline of the course and hints towards reaching the learning goals.

## Course Aim

Fundamental theories about computation and different models of computation. Construction of compilers. Lexical analysis, syntax analysis, and translation into abstract syntax.Regular expressions and grammars, context-free languages and grammars, lexer and parser generators. Identifier handling and symbol table organization. Type-checking, logical inference systems. Intermediate representations and transformations for different languages. Code optimization and register allocation. Machine code generation for common architectures.

In the course you will learn and develop your skills through hands on implementation work building your own complier from scratch. In this way theoretical aspects such as formal grammars, Structural Operational Semantics (SOS), and type rule formalisations becomes tangible. We will even touch upon memory safety and how guarantees can be achieved through static (compilet time) borrow checking. Compiler backend (code optimization etc.) will be discussed in context of LLVM, which you will optionally interface as a library for code generation.

## Draft outline

### W1 The big picture, parsing, semantic analysis, code generation.

Practical assigment:

- Define a minimal subset of Rust, including 

  - Function definitions

  - Commands (let, assignment, if then (else), while)

  - Expressions (includig function calls)

  - Primitive types (boolean, i32) and their literals

  - Explicit types everywhere

  - Explicit return(s)

- Begin writing a parser for expressions in Rust using `nom` (parser combinator library)

### W2 Formal languages and Structural Operational Semantics

Theory:

- Regular expressions and automata

- EBNF

- Structural Operational Semantics

Practical assignment:

- Formulate an EBNF for your language

- Continue on the parser implementation (you may use other tools)

### W3 Context Free Grammars, Push Down Automata and Type Checking

Theory:

- DFA/NFA (regular expressions)

- Push Down Automata (PDA) for Context Free Grammars (CFG)

- Typing Rules and their Derivations

Practical assignment:

- Formulate SOS rules for your language

- Finish parser

- Implement interpreter. Panic! on run-time error.

### W4 Parsing strategies, Mutability and Memory References

Theory:

- Parsing stratigies, pros and cons. L(1), LALR, Parising Expression Grammars (PEG), etc.

- Mutability and memory references

Practical assignment

- Formalize type rules for your language (optional)

- Start to implement type checker

- Extend parser/AST/interpreter to support `&` and `&mut. Panic! on run-time error.

### W5 Borrow checking

Theory:

- Linear types and memory safety

- The Rust borrow model

Practical assigmnent

- Finish type checker. (A program passing type checking should never run into panics in the interpreter due to type errors.)

- Start to implement borrow checker

### W6 LLVM

Theory:

- SSA form

- Concept of `unique`

- Code optimization techniques (performed by LLVM)

- LLVM API (a minimal subset)

Practical assignment

- Borrow checker implementation.

- Optional. Use LLVM as library for code generation.

### W7 Wrapping it up

Practical assignment

- Compiler harness (cli interface)

- Finish work on the compiler

### W8 Home Exam

You will get the home exam to work on the last weeks of the course. This may imply further theoretical exercises and experiments on your compiler.

### Examination

You will each be scheduled 30 minutes to present Your home exam to Jingsen and me, based on which Your grade will be determined. Schudule will be agreed on later using Doodle.

## Files

In this repo you find some examples using `nom` to parse expressions.

- main.rs

Simple recursive decent parsing.

- main*

Shows different approaches to introduce location information and custom error types.

- examples/aron.rs

- examples/climb.rs

Shows two approches to do precedence climbing.

---

## Your parser

- You may implement your parser using any tool of choice.

- You are NOT required to account for operator precedence in expressions, however you MUST support parantesized sub expressions. (+ for precedence towards higher grades)

- You are NOT required to account for location information, but your error messages will be better if you do. (+ for spans, towards higher grades)

- Error recovery is NOT required (+ for recovery towards higher grades)

## Your interpreter

- Your interpreter should be able to correctly execute programs according to your SOS.

- Your interpreter should panic (with an appropriote error message) when encountering an evaluation error (e.g., 1 + false)

## Your type checker

- Your type checker should reject ill-typed programs according to your typing rules.

- (+ for higher grades)

  - span information in type errors

  - multiple error reporting

  - type inference (relaxing explicit typing where possible)

## Your borrow checker

- Your borrow checker should reject borrow errors according to lexical scoping

- (+ for higher grades)

  - Non Lexical Lifetimes (likely hard)

## Your LLVM bindings (Optional)

Implement for higher grades

- Basic code generation.

- Pass `noalias` where possible allowing for better optimization (assuming your borrowchecker prevents aliasing).

- Other attributes, intrinsics, etc. that enables further LLVM optimizations.

 No newline at end of file

use nom::{branch, bytes::complete::tag, character::complete::digit1, error, Err};

use nom_locate::LocatedSpan;

const INPUT: &str = "-2+3**2*3/5-4";

//const INPUT: &str = "2+-3";

//const INPUT: &str = "2+30000000000000000000000";

//const INPUT: &str = "2";

//const INPUT: &str = "30000000000000000000000";

//const INPUT: &str = "3+2a";

const UNARYS: [Funcmap; 1] = [Funcmap {

    keyword: "-",

    prec: 4,

    ass: Ass::Right,

    func: Function::UnSub,

}];

const INFIXS: [Funcmap; 5] = [

    Funcmap {

        keyword: "**",

        prec: 3,

        ass: Ass::Right,

        func: Function::Pow,

    },

    Funcmap {

        keyword: "*",

        prec: 2,

        ass: Ass::Left,

        func: Function::Mult,

    },

    Funcmap {

        keyword: "/",

        prec: 2,

        ass: Ass::Left,

        func: Function::Div,

    },

    Funcmap {

        keyword: "+",

        prec: 1,

        ass: Ass::Left,

        func: Function::Add,

    },

    Funcmap {

        keyword: "-",

        prec: 1,

        ass: Ass::Left,

        func: Function::Sub,

    },

];

struct Expr<'a> {

    span: Span<'a>,

    val: Value<'a>,

}

enum Value<'a> {

    Int(i32),

    UnFunc(Function, Box<Expr<'a>>),

    Func(Function, Box<Expr<'a>>, Box<Expr<'a>>),

}

#[derive(Clone, Copy)]

enum Function {

    UnSub,

    Pow,

    Mult,

    Div,

    Add,

    Sub,

}

struct Funcmap {

    keyword: &'static str,

    prec: u8,

    ass: Ass,

    func: Function,

}

enum Ass {

    Left,

    Right,

}

type Span<'a> = LocatedSpan<&'a str>;

type SpanFuncmap<'a> = (Span<'a>, &'a Funcmap);

type IResult<'a, I, O, E = Error<'a>> = Result<(I, O), Err<E>>;

fn main() {

    match parse(Span::new(INPUT)) {

        Ok(tree) => {

            println!("{:#?}", SimpleExpr::new(&tree));

        }

        Err(Error {

            val: Some(val),

            error,

            ..

        }) => println!(

            "{:#?} at line {}, column {}:\n\t{}",

            error.description(),

            val.line,

            val.get_utf8_column(),

            val.fragment,

        ),

        Err(err) => panic!(err),

    }

}

fn parse<'a>(input: Span) -> Result<Expr, Error> {

    match parse_expr(input) {

        Ok((Span { fragment: "", .. }, tree)) => Ok(tree),

        Ok((input, _)) => Err(Error {

            input: input,

            val: Some(input),

            error: ErrorKind::NotRecognised,

        }),

        Err(Err::Incomplete(_)) => Err(Error {

            input: input,

            val: Some(input),

            error: ErrorKind::Incomplete,

        }),

        Err(Err::Error(err)) => Err(err),

        Err(Err::Failure(err)) => Err(err),

    }

}

fn parse_expr<'a>(input: Span) -> IResult<Span, Expr> {

    branch::alt((parse_infix, parse_expr_nobin))(input)

}

fn parse_expr_nobin<'a>(input: Span) -> IResult<Span, Expr> {

    branch::alt((parse_unary, parse_expr_nobin_noun))(input)

}

fn parse_expr_nobin_noun<'a>(input: Span) -> IResult<Span, Expr> {

    parse_value(input)

}

fn parse_value(input: Span) -> IResult<Span, Expr> {

    // More primitive type categories go here (branch::alt() if > 1):

    parse_number(input)

}

fn parse_number(input: Span) -> IResult<Span, Expr> {

    // More number types go here (branch::alt() if > 1):

    parse_int(input)

}

fn parse_int(input: Span) -> IResult<Span, Expr> {

    let (input, digits) = digit1(input)?;

    let int = match digits.fragment.parse() {

        Ok(int) => int,

        Err(err) => {

            return Err(Err::Failure(Error {

                input,

                val: Some(digits),

                error: ErrorKind::ParseInt(err),

            }))

        }

    };

    Ok((

        input,

        Expr {

            span: digits,

            val: Value::Int(int),

        },

    ))

}

fn parse_unary(input: Span) -> IResult<Span, Expr> {

    let (input, (span, func_map)) = tag_unary(input)?;

    let (input, right) = parse_expr_nobin(input)?;

    Ok((

        input,

        Expr {

            span,

            val: Value::UnFunc(func_map.func, Box::new(right)),

        },

    ))

}

fn parse_infix(input: Span) -> IResult<Span, Expr> {

    // Initialize with minimum precedence 1.

    parse_infix_first(input, 1)

}

fn parse_infix_first<'a>(input: Span, min_prec: u8) -> IResult<Span, Expr> {

    // First, find left hand side expression. Search for everything but BinOps.

    let (input, left) = parse_expr_nobin(input)?;

    parse_infix_left(input, min_prec, left)

}

fn parse_infix_prec<'a>(input: Span, min_prec: u8) -> IResult<Span, Expr> {

    // Almost identical to parse_infix_first(), but we do not accept unary ops.

    let (input, left) = parse_expr_nobin_noun(input)?;

    parse_infix_left(input, min_prec, left)

}

fn parse_infix_left<'a>(

    input_bin: Span<'a>,

    min_prec: u8,

    left: Expr<'a>,

) -> IResult<'a, Span<'a>, Expr<'a>> {

    // See if we can find an infix function. If not, return what we have.

    let (input, (span, func_map)) = match tag_infix(input_bin) {

        Ok(res) => res,

        Err(Err::Error(Error { input, .. })) => return Ok((input, left)),

        Err(err) => return Err(err),

    };

    // Does the new infix fulfill our precedence criteria?

    if func_map.prec < min_prec {

        return Ok((input_bin, left));

    }

    // Parse the right-hand-side.

    let new_min_prec = match func_map.ass {

        Ass::Left => func_map.prec + 1,

        Ass::Right => func_map.prec,

    };

    let (input, right) = parse_infix_prec(input, new_min_prec)?;

    // Put together the infix function with arguments. Continue forward.

    let expr = Expr {

        span,

        val: Value::Func(func_map.func, Box::new(left), Box::new(right)),

    };

    parse_infix_left(input, min_prec, expr)

}

fn tag_unary(input: Span) -> IResult<Span, SpanFuncmap> {

    tag_func(input, &UNARYS)

}

fn tag_infix(input: Span) -> IResult<Span, SpanFuncmap> {

    tag_func(input, &INFIXS)

}

fn tag_func<'a>(input: Span<'a>, funcs: &'a [Funcmap]) -> IResult<'a, Span<'a>, SpanFuncmap<'a>> {

    for func_map in funcs.iter() {

        match tag(func_map.keyword)(input) {

            Ok((input, span)) => return Ok((input, (span, &func_map))),

            Err(Err::Error(_)) => (),

            Err(err) => return Err(err),

        }

    }

    Err(Err::Error(Error {

        input,

        val: None,

        error: ErrorKind::Nom(error::ErrorKind::Tag),

    }))

}

#[derive(Debug)]

enum SimpleExpr<'a> {

    Int(&'a i32),

    UnSub(Box<SimpleExpr<'a>>),

    Pow(Box<SimpleExpr<'a>>, Box<SimpleExpr<'a>>),

    Mult(Box<SimpleExpr<'a>>, Box<SimpleExpr<'a>>),

    Div(Box<SimpleExpr<'a>>, Box<SimpleExpr<'a>>),

    Add(Box<SimpleExpr<'a>>, Box<SimpleExpr<'a>>),

    Sub(Box<SimpleExpr<'a>>, Box<SimpleExpr<'a>>),

}

impl<'a> SimpleExpr<'a> {

    fn new(tree: &'a Expr) -> Self {

        match tree {

            Expr {

                val: Value::Int(int),

                ..

            } => SimpleExpr::Int(&int),

            Expr {

                val: Value::UnFunc(Function::UnSub, x),

                ..

            } => SimpleExpr::UnSub(Box::new(SimpleExpr::new(x))),

            Expr {

                val: Value::Func(Function::Pow, x, y),

                ..

            } => SimpleExpr::Pow(Box::new(SimpleExpr::new(x)), Box::new(SimpleExpr::new(y))),

            Expr {

                val: Value::Func(Function::Mult, x, y),

                ..

            } => SimpleExpr::Mult(Box::new(SimpleExpr::new(x)), Box::new(SimpleExpr::new(y))),

            Expr {

                val: Value::Func(Function::Div, x, y),

                ..

            } => SimpleExpr::Div(Box::new(SimpleExpr::new(x)), Box::new(SimpleExpr::new(y))),

            Expr {

                val: Value::Func(Function::Add, x, y),

                ..

            } => SimpleExpr::Add(Box::new(SimpleExpr::new(x)), Box::new(SimpleExpr::new(y))),

            Expr {

                val: Value::Func(Function::Sub, x, y),

                ..

            } => SimpleExpr::Sub(Box::new(SimpleExpr::new(x)), Box::new(SimpleExpr::new(y))),

            _ => panic!(),

        }

    }

}

impl<'a> error::ParseError<Span<'a>> for Error<'a> {

    fn from_error_kind(input: Span<'a>, kind: error::ErrorKind) -> Self {

        Error {

            input,

            val: None,

            error: ErrorKind::Nom(kind),

        }

    }

    fn append(_: Span<'a>, _: error::ErrorKind, other: Self) -> Self {

        other

    }

}

struct Error<'a> {

    input: Span<'a>,

    val: Option<Span<'a>>,

    error: ErrorKind,

}

enum ErrorKind {

    NotRecognised,

    Incomplete,

    ParseInt(std::num::ParseIntError),

    Nom(error::ErrorKind),

}

impl ErrorKind {

    fn description(&self) -> String {

        match self {

            ErrorKind::NotRecognised => String::from("Failed to parse input."),

            ErrorKind::Incomplete => String::from("There was not enough data."),

            ErrorKind::ParseInt(err) => format!("Parse Int Error ({})", &err),

            ErrorKind::Nom(err) => String::from(err.description()),

        }

    }

}

extern crate nom;

use std::iter::Peekable;

use std::slice::Iter;

use nom::{

    branch::alt,