rust/src/comp/syntax/codemap.rs

import std::vec;
import std::uint;
import std::str;
import std::istr;
import std::term;
import std::io;
import std::option;
import std::option::some;
import std::option::none;

type filename = istr;

type file_pos = {ch: uint, byte: uint};

/* A codemap is a thing that maps uints to file/line/column positions
 * in a crate. This to make it possible to represent the positions
 * with single-word things, rather than passing records all over the
 * compiler.
 */
type filemap =
    @{name: filename, start_pos: file_pos, mutable lines: [file_pos]};

type codemap = @{mutable files: [filemap]};

type loc = {filename: filename, line: uint, col: uint};

fn new_codemap() -> codemap { ret @{mutable files: []}; }

fn new_filemap(filename: filename, start_pos_ch: uint, start_pos_byte: uint)
   -> filemap {
    ret @{name: filename,
          start_pos: {ch: start_pos_ch, byte: start_pos_byte},
          mutable lines: [{ch: start_pos_ch, byte: start_pos_byte}]};
}

fn next_line(file: filemap, chpos: uint, byte_pos: uint) {
    file.lines += [{ch: chpos, byte: byte_pos}];
}

type lookup_fn = fn(file_pos) -> uint;

fn lookup_pos(map: codemap, pos: uint, lookup: lookup_fn) -> loc {
    let a = 0u;
    let b = vec::len(map.files);
    while b - a > 1u {
        let m = (a + b) / 2u;
        if lookup(map.files[m].start_pos) > pos { b = m; } else { a = m; }
    }
    let f = map.files[a];
    a = 0u;
    b = vec::len(f.lines);
    while b - a > 1u {
        let m = (a + b) / 2u;
        if lookup(f.lines[m]) > pos { b = m; } else { a = m; }
    }
    ret {filename: f.name, line: a + 1u, col: pos - lookup(f.lines[a])};
}

fn lookup_char_pos(map: codemap, pos: uint) -> loc {
    fn lookup(pos: file_pos) -> uint { ret pos.ch; }
    ret lookup_pos(map, pos, lookup);
}

fn lookup_byte_pos(map: codemap, pos: uint) -> loc {
    fn lookup(pos: file_pos) -> uint { ret pos.byte; }
    ret lookup_pos(map, pos, lookup);
}

tag opt_span {
     //hack (as opposed to option::t), to make `span` compile
    os_none;
    os_some(@span);
}
type span = {lo: uint, hi: uint, expanded_from: opt_span};

fn span_to_str(sp: &span, cm: &codemap) -> istr {
    let cur = sp;
    let res = ~"";
    let prev_file = none;
    while true {
        let lo = lookup_char_pos(cm, cur.lo);
        let hi = lookup_char_pos(cm, cur.hi);
        res += istr::from_estr(
            #fmt["%s:%u:%u: %u:%u",
                 if some(lo.filename) == prev_file {
                     "-"
                 } else {
                     istr::to_estr(lo.filename)
                 }, lo.line, lo.col, hi.line, hi.col]);
        alt cur.expanded_from {
          os_none. { break; }
          os_some(new_sp) {
            cur = *new_sp;
            prev_file = some(lo.filename);
            res += ~"<<";
          }
        }
    }

    ret res;
}

fn emit_diagnostic(sp: &option::t<span>, msg: &istr, kind: &istr, color: u8,
                   cm: &codemap) {
    let ss = ~"";
    let maybe_lines: option::t<@file_lines> = none;
    alt sp {
      some(ssp) {
        ss = span_to_str(ssp, cm) + ~" ";
        maybe_lines = some(span_to_lines(ssp, cm));
      }
      none. { }
    }
    io::stdout().write_str(ss);
    if term::color_supported() {
        term::fg(io::stdout().get_buf_writer(), color);
    }
    io::stdout().write_str(istr::from_estr(#fmt["%s:", istr::to_estr(kind)]));
    if term::color_supported() { term::reset(io::stdout().get_buf_writer()); }
    io::stdout().write_str(istr::from_estr(#fmt[" %s\n",
                                                istr::to_estr(msg)]));

    maybe_highlight_lines(sp, cm, maybe_lines);
}

fn maybe_highlight_lines(sp: &option::t<span>, cm: &codemap,
                         maybe_lines: option::t<@file_lines>) {

    alt maybe_lines {
      some(lines) {
        // If we're not looking at a real file then we can't re-open it to
        // pull out the lines
        if lines.name == ~"-" { ret; }

        // FIXME: reading in the entire file is the worst possible way to
        //        get access to the necessary lines.
        let file = io::read_whole_file_str(lines.name);
        let fm = get_filemap(cm, lines.name);

        // arbitrarily only print up to six lines of the error
        let max_lines = 6u;
        let elided = false;
        let display_lines = lines.lines;
        if vec::len(display_lines) > max_lines {
            display_lines = vec::slice(display_lines, 0u, max_lines);
            elided = true;
        }
        // Print the offending lines
        for line: uint in display_lines {
            io::stdout().write_str(
                istr::from_estr(#fmt["%s:%u ",
                                     istr::to_estr(fm.name), line + 1u]));
            let s = get_line(fm, line as int, file);
            if !istr::ends_with(s, ~"\n") { s += ~"\n"; }
            io::stdout().write_str(s);
        }
        if elided {
            let last_line = display_lines[vec::len(display_lines) - 1u];
            let s = #fmt["%s:%u ",
                         istr::to_estr(fm.name), last_line + 1u];
            let indent = str::char_len(s);
            let out = ~"";
            while indent > 0u { out += ~" "; indent -= 1u; }
            out += ~"...\n";
            io::stdout().write_str(out);
        }


        // If there's one line at fault we can easily point to the problem
        if vec::len(lines.lines) == 1u {
            let lo = lookup_char_pos(cm, option::get(sp).lo);
            let digits = 0u;
            let num = lines.lines[0] / 10u;

            // how many digits must be indent past?
            while num > 0u { num /= 10u; digits += 1u; }

            // indent past |name:## | and the 0-offset column location
            let left = istr::char_len(fm.name) + digits + lo.col + 3u;
            let s = "";
            while left > 0u { str::push_char(s, ' '); left -= 1u; }

            s += "^";
            let hi = lookup_char_pos(cm, option::get(sp).hi);
            if hi.col != lo.col {
                // the ^ already takes up one space
                let width = hi.col - lo.col - 1u;
                while width > 0u { str::push_char(s, '~'); width -= 1u; }
            }
            io::stdout().write_str(istr::from_estr(s + "\n"));
        }
      }
      _ { }
    }
}

fn emit_warning(sp: &option::t<span>, msg: &istr, cm: &codemap) {
    emit_diagnostic(sp, msg, ~"warning", 11u8, cm);
}
fn emit_error(sp: &option::t<span>, msg: &istr, cm: &codemap) {
    emit_diagnostic(sp, msg, ~"error", 9u8, cm);
}
fn emit_note(sp: &option::t<span>, msg: &istr, cm: &codemap) {
    emit_diagnostic(sp, msg, ~"note", 10u8, cm);
}

type file_lines = {name: istr, lines: [uint]};

fn span_to_lines(sp: span, cm: codemap::codemap) -> @file_lines {
    let lo = lookup_char_pos(cm, sp.lo);
    let hi = lookup_char_pos(cm, sp.hi);
    let lines = [];
    for each i: uint in uint::range(lo.line - 1u, hi.line as uint) {
        lines += [i];
    }
    ret @{name: lo.filename, lines: lines};
}

fn get_line(fm: filemap, line: int, file: &istr) -> istr {
    let begin: uint = fm.lines[line].byte - fm.start_pos.byte;
    let end: uint;
    if line as uint < vec::len(fm.lines) - 1u {
        end = fm.lines[line + 1].byte - fm.start_pos.byte;
    } else {
        // If we're not done parsing the file, we're at the limit of what's
        // parsed. If we just slice the rest of the string, we'll print out
        // the remainder of the file, which is undesirable.
        end = istr::byte_len(file);
        let rest = istr::slice(file, begin, end);
        let newline = istr::index(rest, '\n' as u8);
        if newline != -1 { end = begin + (newline as uint); }
    }
    ret istr::slice(file, begin, end);
}

fn get_filemap(cm: codemap, filename: istr) -> filemap {
    for fm: filemap in cm.files {
        if fm.name == filename { ret fm; }
    }
    //XXjdm the following triggers a mismatched type bug
    //      (or expected function, found _|_)
    fail; // ("asking for " + filename + " which we don't know about");
}
//
// Local Variables:
// mode: rust
// fill-column: 78;
// indent-tabs-mode: nil
// c-basic-offset: 4
// buffer-file-coding-system: utf-8-unix
// compile-command: "make -k -C $RBUILD 2>&1 | sed -e 's/\\/x\\//x:\\//g'";
// End:
//