path: root/src/lib.rs

#![warn(missing_docs)]
//! This top level package provides necessary functions for Emacs to
//! call.

extern crate chain;
extern crate grammar;

use chain::{atom::DefaultAtom, default::DefaultChain, Chain};
use grammar::Grammar;

// For printing forests
use chain::item::{default::DefaultForest, ForestLabel, ForestLabelType};
use grammar::{GrammarLabel, TNT};
use graph::LabelGraph;

/// This struct is the representation of a parser.
///
/// When the user constructs a parser, an instance of this struct will
/// be constructed and the user will receive an opaque pointer to this
/// struct.
#[derive(Debug, Clone)]
#[repr(C)]
pub struct Parser {
    chain: DefaultChain,
}

impl Parser {
    /// Construct a parser from the grammar string.
    ///
    /// The grammar is supposed to conform to the Augmented
    /// Backus-Naur Format.  See RFC 5234 for exact details of this
    /// format.
    pub fn new(s: &str) -> Result<Self, String> {
        let grammar: Grammar = s.parse().map_err(|err| format!("{err}"))?;

        println!("grammar: {grammar}");

        let atom: DefaultAtom =
            DefaultAtom::from_grammar(grammar).map_err(|err| format!("{err}"))?;

        DefaultChain::unit(atom)
            .map_err(|err| format!("{err}"))
            .map(|chain| Self { chain })
    }
}

/// Actual function that is called through C ABI.
///
/// The parameter `ERROR_LEN` is supposed to point to an integer,
/// which will be set to the length of the error message, if and only
/// if an error occurs, in which case the `ERROR_STR` will be set to
/// point to the actual error message.
///
/// It is expected that `*ERROR_STR` should hold the value `NULL` .
#[no_mangle]
extern "C" fn new_parser(
    grammar_string: *mut std::os::raw::c_char,
    error_vec: *mut LenVec<std::os::raw::c_char>,
) -> *mut Parser {
    let parsed_str;

    let error_len = unsafe { (*error_vec).len };
    let error_cap = unsafe { (*error_vec).capacity };

    unsafe {
        match std::ffi::CStr::from_ptr(grammar_string).to_str() {
            Ok(ccstr) => {
                parsed_str = ccstr.to_string();
            }
            Err(e) => {
                let mut e_string = format!("error: {e}");

                let e_string_len_slice = e_string.len().to_be_bytes();
                let e_string_cap_slice = e_string.capacity().to_be_bytes();

                for i in 0..8 {
                    *(error_len.add(i)) = e_string_len_slice.get(i).copied().unwrap();
                    *(error_cap.add(i)) = e_string_cap_slice.get(i).copied().unwrap();
                }

                (*error_vec).data = e_string.as_mut_ptr() as *mut std::os::raw::c_char;

                std::mem::forget(e_string);

                return std::ptr::null_mut();
            }
        }
    }

    match Parser::new(&parsed_str) {
        Ok(result) => unsafe {
            for i in 0..8 {
                *(error_len.add(i)) = 0;
            }

            Box::into_raw(Box::new(result))
        },
        Err(e) => unsafe {
            let mut e_string = format!("error: {e}");

            let e_string_len_slice = e_string.len().to_be_bytes();
            let e_string_cap_slice = e_string.capacity().to_be_bytes();

            for i in 0..8 {
                *(error_len.add(i)) = e_string_len_slice.get(i).copied().unwrap();
                *(error_cap.add(i)) = e_string_cap_slice.get(i).copied().unwrap();
            }

            (*error_vec).data = e_string.as_mut_ptr() as *mut std::os::raw::c_char;

            std::mem::forget(e_string);

            std::ptr::null_mut()
        },
    }
}

#[no_mangle]
extern "C" fn clean_parser(parser: *const std::ffi::c_void) {
    unsafe {
        drop(Box::from_raw(parser as *mut Parser));
    }
}

/// To make it easier to pass arrays to and from C.
#[repr(C)]
pub struct LenVec<T> {
    /// This must be an array of unsigned char of length 8.
    ///
    /// A less length leads to access to invalid memory location,
    /// while a longer length will be ignored, possibly leading to
    /// wrong length.
    ///
    /// The length will be interpreted as a 64-bits integer stored in
    /// big endian format.
    pub len: *mut std::os::raw::c_uchar,
    /// This must be an array of unsigned chars of length 8.
    ///
    /// This is only used so that Rust knows how to reconstruct the
    /// data from C, in order for Rust to deallocate the objects
    /// exposed by this library.
    pub capacity: *mut std::os::raw::c_uchar,
    /// The actual pointer to the data.
    ///
    /// In case this should be set by the function on the Rust end,
    /// this field should be `NULL`.
    pub data: *mut T,
}

#[no_mangle]
extern "C" fn clean_signed(vec: *mut LenVec<std::os::raw::c_char>, flag: std::os::raw::c_uchar) {
    let len = usize::from_be_bytes(
        unsafe { std::slice::from_raw_parts((*vec).len, 8) }
            .try_into()
            .unwrap(),
    );
    let capacity = usize::from_be_bytes(
        unsafe { std::slice::from_raw_parts((*vec).capacity, 8) }
            .try_into()
            .unwrap(),
    );

    if (flag & 1) != 0 {
        drop(unsafe { Vec::from_raw_parts((*vec).len, 8, 8) });
    }

    if (flag & 2) != 0 {
        drop(unsafe { Vec::from_raw_parts((*vec).capacity, 8, 8) });
    }

    if (flag & 4) != 0 {
        drop(unsafe { String::from_raw_parts((*vec).data as *mut u8, len, capacity) });
    }

    if (flag & 8) != 0 {
        drop(unsafe { Box::from_raw(vec) });
    }
}

#[no_mangle]
extern "C" fn clean_unsigned(vec: *mut LenVec<std::os::raw::c_uchar>, flag: std::os::raw::c_uchar) {
    let len = usize::from_be_bytes(
        unsafe { std::slice::from_raw_parts((*vec).len, 8) }
            .try_into()
            .unwrap(),
    );
    let capacity = usize::from_be_bytes(
        unsafe { std::slice::from_raw_parts((*vec).capacity, 8) }
            .try_into()
            .unwrap(),
    );

    if (flag & 1) != 0 {
        drop(unsafe { Vec::from_raw_parts((*vec).len, 8, 8) });
    }

    if (flag & 2) != 0 {
        drop(unsafe { Vec::from_raw_parts((*vec).capacity, 8, 8) });
    }

    if (flag & 4) != 0 {
        drop(unsafe { Vec::<u8>::from_raw_parts((*vec).data, len, capacity) });
    }

    if (flag & 8) != 0 {
        drop(unsafe { Box::from_raw(vec) });
    }
}

#[no_mangle]
extern "C" fn parser_recognize(
    parser: *mut Parser,
    input_vec: *mut LenVec<std::os::raw::c_uchar>,
    error_vec: *mut LenVec<std::os::raw::c_char>,
    reset_p: std::os::raw::c_uchar,
) -> std::os::raw::c_int {
    let input_len = usize::from_be_bytes(
        unsafe { std::slice::from_raw_parts((*input_vec).len, 8) }
            .try_into()
            .unwrap(),
    );

    let mut parser_box;
    let input_array_len = input_len;
    let input_array;

    let error_len = unsafe { (*error_vec).len };
    let error_cap = unsafe { (*error_vec).capacity };

    unsafe {
        parser_box = Box::from_raw(parser);

        input_array = std::slice::from_raw_parts((*input_vec).data, input_array_len);
    }

    // If the parser has already been used before, reset it to the
    // initial state.

    if reset_p != 0 && !parser_box.chain.history().is_empty() {
        match DefaultChain::unit(parser_box.chain.atom().clone()) {
            Ok(chain) => {
                parser_box.chain = chain;
            }
            Err(e) => {
                let mut e_string = format!("error: {e}");

                Box::leak(parser_box);

                let e_string_len_slice = e_string.len().to_be_bytes();
                let e_string_cap_slice = e_string.capacity().to_be_bytes();

                unsafe {
                    for i in 0..8 {
                        *(error_len.add(i)) = e_string_len_slice.get(i).copied().unwrap();
                        *(error_cap.add(i)) = e_string_cap_slice.get(i).copied().unwrap();
                    }

                    (*error_vec).data = e_string.as_mut_ptr() as *mut std::os::raw::c_char;
                }

                std::mem::forget(e_string);

                return 0;
            }
        }
    }

    if input_array_len.rem_euclid(8) != 0 {
        let mut e_string =
            format!("error: input length should be divisible by 8, but got {input_array_len}");

        let e_string_len_slice = e_string.len().to_be_bytes();
        let e_string_cap_slice = e_string.capacity().to_be_bytes();

        Box::leak(parser_box);

        unsafe {
            for i in 0..8 {
                *(error_len.add(i)) = e_string_len_slice.get(i).copied().unwrap();
                *(error_cap.add(i)) = e_string_cap_slice.get(i).copied().unwrap();
            }

            (*error_vec).data = e_string.as_mut_ptr() as *mut std::os::raw::c_char;
        }

        std::mem::forget(e_string);

        return 0;
    }

    #[cfg(target_pointer_width = "64")]
    let input_iter = input_array
        .chunks_exact(8)
        .map(|chunk| usize::from_be_bytes(<[u8; 8]>::try_from(chunk).unwrap()));

    #[cfg(not(target_pointer_width = "64"))]
    compile_error!("this program assumes to be run on 64-bits machines");

    for (index, token) in input_iter.enumerate() {
        let chain_result = parser_box.chain.chain(token, index, true);

        if let Err(e) = chain_result {
            let mut e_string = format!("error: {e}");

            let e_string_len_slice = e_string.len().to_be_bytes();
            let e_string_cap_slice = e_string.capacity().to_be_bytes();

            Box::leak(parser_box);

            unsafe {
                for i in 0..8 {
                    *(error_len.add(i)) = e_string_len_slice.get(i).copied().unwrap();
                    *(error_cap.add(i)) = e_string_cap_slice.get(i).copied().unwrap();
                }

                (*error_vec).data = e_string.as_mut_ptr() as *mut std::os::raw::c_char;
            }

            std::mem::forget(e_string);

            return 0;
        }
    }

    match parser_box.chain.epsilon() {
        Ok(result) => {
            Box::leak(parser_box);

            result as std::os::raw::c_int
        }
        Err(e) => {
            let mut e_string = format!("error: {e}");

            let e_string_len_slice = e_string.len().to_be_bytes();
            let e_string_cap_slice = e_string.capacity().to_be_bytes();

            Box::leak(parser_box);

            unsafe {
                for i in 0..8 {
                    *(error_len.add(i)) = e_string_len_slice.get(i).copied().unwrap();
                    *(error_cap.add(i)) = e_string_cap_slice.get(i).copied().unwrap();
                }

                (*error_vec).data = e_string.as_mut_ptr() as *mut std::os::raw::c_char;
            }

            std::mem::forget(e_string);

            0
        }
    }
}

#[no_mangle]
extern "C" fn parser_parse(
    parser: *mut Parser,
    input_vec: *mut LenVec<std::os::raw::c_uchar>,
    error_vec: *mut LenVec<std::os::raw::c_char>,
    reset_p: std::os::raw::c_uchar,
) -> *mut LenVec<std::os::raw::c_uchar> {
    let input_len = usize::from_be_bytes(
        unsafe { std::slice::from_raw_parts((*input_vec).len, 8) }
            .try_into()
            .unwrap(),
    );

    let mut parser_box;
    let input_array;

    let error_len = unsafe { (*error_vec).len };
    let error_cap = unsafe { (*error_vec).capacity };

    unsafe {
        parser_box = Box::from_raw(parser);

        input_array = std::slice::from_raw_parts((*input_vec).data, input_len);
    }

    // If the parser has already been used before, reset it to the
    // initial state.

    if reset_p != 0 && !parser_box.chain.history().is_empty() {
        match DefaultChain::unit(parser_box.chain.atom().clone()) {
            Ok(chain) => {
                parser_box.chain = chain;
            }
            Err(e) => {
                let mut e_string = format!("error: {e}");

                Box::leak(parser_box);

                let e_string_len_slice = e_string.len().to_be_bytes();
                let e_string_cap_slice = e_string.capacity().to_be_bytes();

                unsafe {
                    for i in 0..8 {
                        *(error_len.add(i)) = e_string_len_slice.get(i).copied().unwrap();
                        *(error_cap.add(i)) = e_string_cap_slice.get(i).copied().unwrap();
                    }

                    (*error_vec).data = e_string.as_mut_ptr() as *mut std::os::raw::c_char;
                }

                std::mem::forget(e_string);

                return std::ptr::null_mut();
            }
        }
    }

    if input_len.rem_euclid(8) != 0 {
        let mut e_string =
            format!("error: input length should be divisible by 8, but got {input_len}");

        let e_string_len_slice = e_string.len().to_be_bytes();
        let e_string_cap_slice = e_string.capacity().to_be_bytes();

        Box::leak(parser_box);

        unsafe {
            for i in 0..8 {
                *(error_len.add(i)) = e_string_len_slice.get(i).copied().unwrap();
                *(error_cap.add(i)) = e_string_cap_slice.get(i).copied().unwrap();
            }

            (*error_vec).data = e_string.as_mut_ptr() as *mut std::os::raw::c_char;
        }

        std::mem::forget(e_string);

        return std::ptr::null_mut();
    } else if input_len == 0 {
        Box::leak(parser_box);

        return std::ptr::null_mut();
    }

    #[cfg(target_pointer_width = "64")]
    let input_iter = input_array
        .chunks_exact(8)
        .map(|chunk| usize::from_be_bytes(<[u8; 8]>::try_from(chunk).unwrap()));

    #[cfg(not(target_pointer_width = "64"))]
    compile_error!("this program assumes to be run on 64-bits machines");

    let mut last_pos: usize = 0;
    let mut last_token: usize = 0;

    for (index, token) in input_iter.enumerate() {
        last_pos = index;
        last_token = token;

        let chain_result = parser_box.chain.chain(token, index, false);

        if let Err(e) = chain_result {
            let mut e_string = format!("error: {e}");

            let e_string_len_slice = e_string.len().to_be_bytes();
            let e_string_cap_slice = e_string.capacity().to_be_bytes();

            Box::leak(parser_box);

            unsafe {
                for i in 0..8 {
                    *(error_len.add(i)) = e_string_len_slice.get(i).copied().unwrap();
                    *(error_cap.add(i)) = e_string_cap_slice.get(i).copied().unwrap();
                }

                (*error_vec).data = e_string.as_mut_ptr() as *mut std::os::raw::c_char;
            }

            std::mem::forget(e_string);

            return std::ptr::null_mut();
        }
    }

    match parser_box.chain.epsilon() {
        Ok(result) => {
            if result {
                let forest = parser_box.chain.end_of_input(last_pos + 1, last_token);

                match forest {
                    Ok(forest) => {
                        Box::leak(parser_box);

                        let mut bytes = bytes::forest_to_bytes(&forest);

                        let bytes_len = bytes.len().to_be_bytes().to_vec();

                        let bytes_capacity = bytes.capacity().to_be_bytes().to_vec();

                        let bytes_vec: LenVec<std::os::raw::c_uchar> = LenVec {
                            len: Box::leak(bytes_len.into_boxed_slice()).as_mut_ptr(),
                            capacity: Box::leak(bytes_capacity.into_boxed_slice()).as_mut_ptr(),
                            data: bytes.as_mut_ptr(),
                        };

                        std::mem::forget(bytes);

                        Box::into_raw(Box::new(bytes_vec))
                    }
                    Err(e) => {
                        let mut e_string = format!("error: {e}");

                        let e_string_len_slice = e_string.len().to_be_bytes();
                        let e_string_cap_slice = e_string.capacity().to_be_bytes();

                        Box::leak(parser_box);

                        unsafe {
                            for i in 0..8 {
                                *(error_len.add(i)) = e_string_len_slice.get(i).copied().unwrap();
                                *(error_cap.add(i)) = e_string_cap_slice.get(i).copied().unwrap();
                            }

                            (*error_vec).data = e_string.as_mut_ptr() as *mut std::os::raw::c_char;
                        }

                        std::mem::forget(e_string);

                        std::ptr::null_mut()
                    }
                }
            } else {
                Box::leak(parser_box);

                std::ptr::null_mut()
            }
        }
        Err(e) => {
            let mut e_string = format!("error: {e}");

            let e_string_len_slice = e_string.len().to_be_bytes();
            let e_string_cap_slice = e_string.capacity().to_be_bytes();

            Box::leak(parser_box);

            unsafe {
                for i in 0..8 {
                    *(error_len.add(i)) = e_string_len_slice.get(i).copied().unwrap();
                    *(error_cap.add(i)) = e_string_cap_slice.get(i).copied().unwrap();
                }

                (*error_vec).data = e_string.as_mut_ptr() as *mut std::os::raw::c_char;
            }

            std::mem::forget(e_string);

            std::ptr::null_mut()
        }
    }
}

fn read_label(label: *mut std::os::raw::c_uchar) -> ForestLabel<GrammarLabel> {
    let status: u8 = unsafe { *label };

    let label_status: ForestLabelType = match status {
        0 => ForestLabelType::Plain,
        1 => ForestLabelType::Packed,
        _ => ForestLabelType::Cloned(usize::from_be_bytes(
            unsafe { std::slice::from_raw_parts(label.add(1), 8) }
                .try_into()
                .unwrap(),
        )),
    };

    let start = usize::from_be_bytes(
        unsafe { std::slice::from_raw_parts(label.add(9), 8) }
            .try_into()
            .unwrap(),
    );

    let end = usize::from_be_bytes(
        unsafe { std::slice::from_raw_parts(label.add(17), 8) }
            .try_into()
            .unwrap(),
    );

    let discriminant: u8 = unsafe { *label.add(25) };

    let content = usize::from_be_bytes(
        unsafe { std::slice::from_raw_parts(label.add(26), 8) }
            .try_into()
            .unwrap(),
    );

    let inner_label: GrammarLabel = match discriminant {
        0 => GrammarLabel::new_closed(TNT::Ter(content), start, end),
        1 => GrammarLabel::new_closed(TNT::Non(content), start, end),
        _ => GrammarLabel::new_closed(content, start, end),
    };

    ForestLabel::new(inner_label, label_status)
}

macro_rules! return_error {
    ($err:expr, $elen:ident, $ecap:ident, $evec:ident) => {
        let mut e_string = $err;

        let e_string_len_slice = e_string.len().to_be_bytes();
        let e_string_cap_slice = e_string.capacity().to_be_bytes();

        unsafe {
            for i in 0..8 {
                *($elen.add(i)) = e_string_len_slice.get(i).copied().unwrap();
                *($ecap.add(i)) = e_string_cap_slice.get(i).copied().unwrap();
            }

            (*$evec).data = e_string.as_mut_ptr() as *mut std::os::raw::c_char;
        }

        std::mem::forget(e_string);

        return;
    };
}

#[no_mangle]
extern "C" fn print_forest(
    forest_vec: *mut LenVec<std::os::raw::c_uchar>,
    error_vec: *mut LenVec<std::os::raw::c_char>,
    filename: *mut std::os::raw::c_char,
) {
    let forest_len = usize::from_be_bytes(
        unsafe { std::slice::from_raw_parts((*forest_vec).len, 8) }
            .try_into()
            .unwrap(),
    );

    let error_len = unsafe { (*error_vec).len };
    let error_cap = unsafe { (*error_vec).capacity };

    if forest_len < 27 {
        return_error!(
            format!("forest bytes length {forest_len} < 27"),
            error_len,
            error_cap,
            error_vec
        );
    }

    let nodes_len = usize::from_be_bytes(
        unsafe { std::slice::from_raw_parts((*forest_vec).data.add(11), 8) }
            .try_into()
            .unwrap(),
    );

    println!("the forest has {nodes_len} nodes");

    let special_marks = unsafe { std::slice::from_raw_parts((*forest_vec).data.add(8), 3) };

    if special_marks != [114, 101, 112] {
        return_error!(
            format!(
                "the forest does not begin with the special mark\nThe first bytes are: \
                     {:?}\n",
                special_marks
            ),
            error_len,
            error_cap,
            error_vec
        );
    }

    let labels_offset = usize::from_be_bytes(
        unsafe { std::slice::from_raw_parts((*forest_vec).data.add(19), 8) }
            .try_into()
            .unwrap(),
    );

    println!("labels_offset = {labels_offset}");

    if forest_len < labels_offset + 34 * nodes_len || forest_len < (27 + 16 * nodes_len) {
        return_error!(
            format!(
                "the forest length is too small: {forest_len}\n\
                               labels offset + 34 * nodes_len = {}, all {nodes_len} \
                               nodes take {}\n",
                labels_offset + 34 * nodes_len,
                27 + 16 * nodes_len
            ),
            error_len,
            error_cap,
            error_vec
        );
    }

    let mut total_degree = 0usize;

    let preamble = "digraph forest {
    fontname=\"Helvetica,Arial,sans-serif\"
    node [fontname=\"Helvetica,Arial,sans-serif\", ordering=out]
    edge [fontname=\"Helvetica,Arial,sans-serif\"]
    rankdir=LR;\n";

    let mut post = String::new();

    for node in 0..nodes_len {
        let degree = usize::from_be_bytes(
            unsafe { std::slice::from_raw_parts((*forest_vec).data.add(27 + 16 * node), 8) }
                .try_into()
                .unwrap(),
        );

        total_degree += degree;

        post.push_str(&format!(
            "    {node} [label = \"{node}:{}\"]\n",
            read_label(unsafe { (*forest_vec).data.add(labels_offset + 34 * node) })
        ));
    }

    println!("total degree = {total_degree}");

    let correct_len: usize = 27 + 50 * nodes_len + 8 * total_degree;

    println!("correct length = {correct_len}");

    if forest_len != correct_len {
        return_error!(
            format!("the forest length {forest_len} should be equal to: {correct_len}\n"),
            error_len,
            error_cap,
            error_vec
        );
    }

    for source in 0..nodes_len {
        let degree = usize::from_be_bytes(
            unsafe { std::slice::from_raw_parts((*forest_vec).data.add(27 + 16 * source), 8) }
                .try_into()
                .unwrap(),
        );

        let node_offset = usize::from_be_bytes(
            unsafe { std::slice::from_raw_parts((*forest_vec).data.add(27 + 16 * source + 8), 8) }
                .try_into()
                .unwrap(),
        );

        if forest_len <= node_offset + 8 * degree {
            return_error!(
                format!(
                    "the forest length {forest_len} is <= {node_offset} + 8 * {degree} = {}\n",
                    node_offset + 8 * degree
                ),
                error_len,
                error_cap,
                error_vec
            );
        }

        for i in 0..degree {
            let target = usize::from_be_bytes(
                unsafe {
                    std::slice::from_raw_parts((*forest_vec).data.add(node_offset + 8 * i), 8)
                }
                .try_into()
                .unwrap(),
            );

            post.push_str(&format!("    {source} -> {target}\n"));
        }
    }

    post.push_str("}\n");

    let result = format!("{preamble}{post}");

    let parsed_filename = match unsafe { std::ffi::CStr::from_ptr(filename).to_str() } {
        Ok(ccstr) => ccstr,
        Err(e) => {
            return_error!(format!("error: {e}"), error_len, error_cap, error_vec);
        }
    };

    if std::fs::metadata(parsed_filename).is_ok() {
        let _ = std::fs::remove_file(parsed_filename);
    }

    let file = std::fs::File::options()
        .write(true)
        .create(true)
        .open(parsed_filename);

    use std::io::Write;

    match file {
        Ok(mut file) => {
            if let Err(e) = file.write_all(result.as_bytes()) {
                return_error!(format!("error: {e}"), error_len, error_cap, error_vec);
            }
        }
        Err(e) => {
            return_error!(format!("error: {e}"), error_len, error_cap, error_vec);
        }
    }
}

// TODO: Write a function to print the node label of a forest and
// expose it to C ABI.
//
// This can be used in LLDB.

/// This struct is a wrapper around the forest.
///
/// This is used so that we can call a C function receiving a pointer
/// to a forest struct.
#[derive(Debug, Clone)]
#[repr(C)]
pub struct CForest {
    forest: DefaultForest<ForestLabel<GrammarLabel>>,
}

/// Print the label of the node with id `node` in the forest `forest`.
///
/// The parameter `node` should point to 8 bytes of unsigned
/// characters, which forms a number in 64 bits, in the *big endian*
/// format.
#[no_mangle]
extern "C" fn print_forest_node(forest: *mut CForest, node: *mut std::os::raw::c_uchar) {
    let node = usize::from_be_bytes(
        unsafe { std::slice::from_raw_parts(node, 8) }
            .try_into()
            .unwrap(),
    );

    let forest = unsafe { (*forest).forest.clone() };

    let Ok(Some(label)) = forest.vertex_label(node) else {
        return;
    };

    println!("node {node} has label {label}");
}

pub mod bytes;