[rustc.git] / src / libsyntax / codemap.rs

// Copyright 2012 The Rust Project Developers. See the COPYRIGHT
// file at the top-level directory of this distribution and at
// http://rust-lang.org/COPYRIGHT.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.
//
// ignore-lexer-test FIXME #15679

//! The CodeMap tracks all the source code used within a single crate, mapping
//! from integer byte positions to the original source code location. Each bit
//! of source parsed during crate parsing (typically files, in-memory strings,
//! or various bits of macro expansion) cover a continuous range of bytes in the
//! CodeMap and are represented by FileMaps. Byte positions are stored in
//! `spans` and used pervasively in the compiler. They are absolute positions
//! within the CodeMap, which upon request can be converted to line and column
//! information, source code snippets, etc.

pub use self::MacroFormat::*;

use std::cell::RefCell;
use std::num::ToPrimitive;
use std::ops::{Add, Sub};
use std::rc::Rc;

use libc::c_uint;
use serialize::{Encodable, Decodable, Encoder, Decoder};

pub trait Pos {
    fn from_usize(n: usize) -> Self;
    fn to_usize(&self) -> usize;
}

/// A byte offset. Keep this small (currently 32-bits), as AST contains
/// a lot of them.
#[derive(Clone, Copy, PartialEq, Eq, Hash, PartialOrd, Debug)]
pub struct BytePos(pub u32);

/// A character offset. Because of multibyte utf8 characters, a byte offset
/// is not equivalent to a character offset. The CodeMap will convert BytePos
/// values to CharPos values as necessary.
#[derive(Copy, PartialEq, Hash, PartialOrd, Debug)]
pub struct CharPos(pub usize);

// FIXME: Lots of boilerplate in these impls, but so far my attempts to fix
// have been unsuccessful

impl Pos for BytePos {
    fn from_usize(n: usize) -> BytePos { BytePos(n as u32) }
    fn to_usize(&self) -> usize { let BytePos(n) = *self; n as usize }
}

impl Add for BytePos {
    type Output = BytePos;

    fn add(self, rhs: BytePos) -> BytePos {
        BytePos((self.to_usize() + rhs.to_usize()) as u32)
    }
}

impl Sub for BytePos {
    type Output = BytePos;

    fn sub(self, rhs: BytePos) -> BytePos {
        BytePos((self.to_usize() - rhs.to_usize()) as u32)
    }
}

impl Pos for CharPos {
    fn from_usize(n: usize) -> CharPos { CharPos(n) }
    fn to_usize(&self) -> usize { let CharPos(n) = *self; n }
}

impl Add for CharPos {
    type Output = CharPos;

    fn add(self, rhs: CharPos) -> CharPos {
        CharPos(self.to_usize() + rhs.to_usize())
    }
}

impl Sub for CharPos {
    type Output = CharPos;

    fn sub(self, rhs: CharPos) -> CharPos {
        CharPos(self.to_usize() - rhs.to_usize())
    }
}

/// Spans represent a region of code, used for error reporting. Positions in spans
/// are *absolute* positions from the beginning of the codemap, not positions
/// relative to FileMaps. Methods on the CodeMap can be used to relate spans back
/// to the original source.
#[derive(Clone, Copy, Debug, Hash)]
pub struct Span {
    pub lo: BytePos,
    pub hi: BytePos,
    /// Information about where the macro came from, if this piece of
    /// code was created by a macro expansion.
    pub expn_id: ExpnId
}

pub const DUMMY_SP: Span = Span { lo: BytePos(0), hi: BytePos(0), expn_id: NO_EXPANSION };

// Generic span to be used for code originating from the command line
pub const COMMAND_LINE_SP: Span = Span { lo: BytePos(0),
                                         hi: BytePos(0),
                                         expn_id: COMMAND_LINE_EXPN };

#[derive(Clone, PartialEq, Eq, RustcEncodable, RustcDecodable, Hash, Debug, Copy)]
pub struct Spanned<T> {
    pub node: T,
    pub span: Span,
}

impl PartialEq for Span {
    fn eq(&self, other: &Span) -> bool {
        return (*self).lo == (*other).lo && (*self).hi == (*other).hi;
    }
    fn ne(&self, other: &Span) -> bool { !(*self).eq(other) }
}

impl Eq for Span {}

impl Encodable for Span {
    /* Note #1972 -- spans are encoded but not decoded */
    fn encode<S: Encoder>(&self, s: &mut S) -> Result<(), S::Error> {
        s.emit_nil()
    }
}

impl Decodable for Span {
    fn decode<D: Decoder>(_d: &mut D) -> Result<Span, D::Error> {
        Ok(DUMMY_SP)
    }
}

pub fn spanned<T>(lo: BytePos, hi: BytePos, t: T) -> Spanned<T> {
    respan(mk_sp(lo, hi), t)
}

pub fn respan<T>(sp: Span, t: T) -> Spanned<T> {
    Spanned {node: t, span: sp}
}

pub fn dummy_spanned<T>(t: T) -> Spanned<T> {
    respan(DUMMY_SP, t)
}

/* assuming that we're not in macro expansion */
pub fn mk_sp(lo: BytePos, hi: BytePos) -> Span {
    Span {lo: lo, hi: hi, expn_id: NO_EXPANSION}
}

/// Return the span itself if it doesn't come from a macro expansion,
/// otherwise return the call site span up to the `enclosing_sp` by
/// following the `expn_info` chain.
pub fn original_sp(cm: &CodeMap, sp: Span, enclosing_sp: Span) -> Span {
    let call_site1 = cm.with_expn_info(sp.expn_id, |ei| ei.map(|ei| ei.call_site));
    let call_site2 = cm.with_expn_info(enclosing_sp.expn_id, |ei| ei.map(|ei| ei.call_site));
    match (call_site1, call_site2) {
        (None, _) => sp,
        (Some(call_site1), Some(call_site2)) if call_site1 == call_site2 => sp,
        (Some(call_site1), _) => original_sp(cm, call_site1, enclosing_sp),
    }
}

/// A source code location used for error reporting
pub struct Loc {
    /// Information about the original source
    pub file: Rc<FileMap>,
    /// The (1-based) line number
    pub line: usize,
    /// The (0-based) column offset
    pub col: CharPos
}

/// A source code location used as the result of lookup_char_pos_adj
// Actually, *none* of the clients use the filename *or* file field;
// perhaps they should just be removed.
pub struct LocWithOpt {
    pub filename: FileName,
    pub line: usize,
    pub col: CharPos,
    pub file: Option<Rc<FileMap>>,
}

// used to be structural records. Better names, anyone?
pub struct FileMapAndLine { pub fm: Rc<FileMap>, pub line: usize }
pub struct FileMapAndBytePos { pub fm: Rc<FileMap>, pub pos: BytePos }

/// The syntax with which a macro was invoked.
#[derive(Clone, Copy, Hash, Debug)]
pub enum MacroFormat {
    /// e.g. #[derive(...)] <item>
    MacroAttribute,
    /// e.g. `format!()`
    MacroBang
}

#[derive(Clone, Hash, Debug)]
pub struct NameAndSpan {
    /// The name of the macro that was invoked to create the thing
    /// with this Span.
    pub name: String,
    /// The format with which the macro was invoked.
    pub format: MacroFormat,
    /// The span of the macro definition itself. The macro may not
    /// have a sensible definition span (e.g. something defined
    /// completely inside libsyntax) in which case this is None.
    pub span: Option<Span>
}

/// Extra information for tracking macro expansion of spans
#[derive(Hash, Debug)]
pub struct ExpnInfo {
    /// The location of the actual macro invocation, e.g. `let x =
    /// foo!();`
    ///
    /// This may recursively refer to other macro invocations, e.g. if
    /// `foo!()` invoked `bar!()` internally, and there was an
    /// expression inside `bar!`; the call_site of the expression in
    /// the expansion would point to the `bar!` invocation; that
    /// call_site span would have its own ExpnInfo, with the call_site
    /// pointing to the `foo!` invocation.
    pub call_site: Span,
    /// Information about the macro and its definition.
    ///
    /// The `callee` of the inner expression in the `call_site`
    /// example would point to the `macro_rules! bar { ... }` and that
    /// of the `bar!()` invocation would point to the `macro_rules!
    /// foo { ... }`.
    pub callee: NameAndSpan
}

#[derive(PartialEq, Eq, Clone, Debug, Hash, RustcEncodable, RustcDecodable, Copy)]
pub struct ExpnId(u32);

pub const NO_EXPANSION: ExpnId = ExpnId(-1);
// For code appearing from the command line
pub const COMMAND_LINE_EXPN: ExpnId = ExpnId(-2);

impl ExpnId {
    pub fn from_llvm_cookie(cookie: c_uint) -> ExpnId {
        ExpnId(cookie as u32)
    }

    pub fn to_llvm_cookie(self) -> i32 {
        let ExpnId(cookie) = self;
        cookie as i32
    }
}

pub type FileName = String;

pub struct FileLines {
    pub file: Rc<FileMap>,
    pub lines: Vec<usize>
}

/// Identifies an offset of a multi-byte character in a FileMap
#[derive(Copy)]
pub struct MultiByteChar {
    /// The absolute offset of the character in the CodeMap
    pub pos: BytePos,
    /// The number of bytes, >=2
    pub bytes: usize,
}

/// A single source in the CodeMap
pub struct FileMap {
    /// The name of the file that the source came from, source that doesn't
    /// originate from files has names between angle brackets by convention,
    /// e.g. `<anon>`
    pub name: FileName,
    /// The complete source code
    pub src: String,
    /// The start position of this source in the CodeMap
    pub start_pos: BytePos,
    /// Locations of lines beginnings in the source code
    pub lines: RefCell<Vec<BytePos> >,
    /// Locations of multi-byte characters in the source code
    pub multibyte_chars: RefCell<Vec<MultiByteChar> >,
}

impl FileMap {
    /// EFFECT: register a start-of-line offset in the
    /// table of line-beginnings.
    /// UNCHECKED INVARIANT: these offsets must be added in the right
    /// order and must be in the right places; there is shared knowledge
    /// about what ends a line between this file and parse.rs
    /// WARNING: pos param here is the offset relative to start of CodeMap,
    /// and CodeMap will append a newline when adding a filemap without a newline at the end,
    /// so the safe way to call this is with value calculated as
    /// filemap.start_pos + newline_offset_relative_to_the_start_of_filemap.
    pub fn next_line(&self, pos: BytePos) {
        // the new charpos must be > the last one (or it's the first one).
        let mut lines = self.lines.borrow_mut();
        let line_len = lines.len();
        assert!(line_len == 0 || ((*lines)[line_len - 1] < pos));
        lines.push(pos);
    }

    /// get a line from the list of pre-computed line-beginnings
    ///
    pub fn get_line(&self, line_number: usize) -> Option<String> {
        let lines = self.lines.borrow();
        lines.get(line_number).map(|&line| {
            let begin: BytePos = line - self.start_pos;
            let begin = begin.to_usize();
            let slice = &self.src[begin..];
            match slice.find('\n') {
                Some(e) => &slice[..e],
                None => slice
            }.to_string()
        })
    }

    pub fn record_multibyte_char(&self, pos: BytePos, bytes: usize) {
        assert!(bytes >=2 && bytes <= 4);
        let mbc = MultiByteChar {
            pos: pos,
            bytes: bytes,
        };
        self.multibyte_chars.borrow_mut().push(mbc);
    }

    pub fn is_real_file(&self) -> bool {
        !(self.name.starts_with("<") &&
          self.name.ends_with(">"))
    }
}

pub struct CodeMap {
    pub files: RefCell<Vec<Rc<FileMap>>>,
    expansions: RefCell<Vec<ExpnInfo>>
}

impl CodeMap {
    pub fn new() -> CodeMap {
        CodeMap {
            files: RefCell::new(Vec::new()),
            expansions: RefCell::new(Vec::new()),
        }
    }

    pub fn new_filemap(&self, filename: FileName, src: String) -> Rc<FileMap> {
        let mut files = self.files.borrow_mut();
        let start_pos = match files.last() {
            None => 0,
            Some(last) => last.start_pos.to_usize() + last.src.len(),
        };

        // Remove utf-8 BOM if any.
        // FIXME #12884: no efficient/safe way to remove from the start of a string
        // and reuse the allocation.
        let mut src = if src.starts_with("\u{feff}") {
            String::from_str(&src[3..])
        } else {
            String::from_str(&src[..])
        };

        // Append '\n' in case it's not already there.
        // This is a workaround to prevent CodeMap.lookup_filemap_idx from
        // accidentally overflowing into the next filemap in case the last byte
        // of span is also the last byte of filemap, which leads to incorrect
        // results from CodeMap.span_to_*.
        if src.len() > 0 && !src.ends_with("\n") {
            src.push('\n');
        }

        let filemap = Rc::new(FileMap {
            name: filename,
            src: src.to_string(),
            start_pos: Pos::from_usize(start_pos),
            lines: RefCell::new(Vec::new()),
            multibyte_chars: RefCell::new(Vec::new()),
        });

        files.push(filemap.clone());

        filemap
    }

    pub fn mk_substr_filename(&self, sp: Span) -> String {
        let pos = self.lookup_char_pos(sp.lo);
        (format!("<{}:{}:{}>",
                 pos.file.name,
                 pos.line,
                 pos.col.to_usize() + 1)).to_string()
    }

    /// Lookup source information about a BytePos
    pub fn lookup_char_pos(&self, pos: BytePos) -> Loc {
        self.lookup_pos(pos)
    }

    pub fn lookup_char_pos_adj(&self, pos: BytePos) -> LocWithOpt {
        let loc = self.lookup_char_pos(pos);
        LocWithOpt {
            filename: loc.file.name.to_string(),
            line: loc.line,
            col: loc.col,
            file: Some(loc.file)
        }
    }

    pub fn span_to_string(&self, sp: Span) -> String {
        if self.files.borrow().len() == 0 && sp == DUMMY_SP {
            return "no-location".to_string();
        }

        let lo = self.lookup_char_pos_adj(sp.lo);
        let hi = self.lookup_char_pos_adj(sp.hi);
        return (format!("{}:{}:{}: {}:{}",
                        lo.filename,
                        lo.line,
                        lo.col.to_usize() + 1,
                        hi.line,
                        hi.col.to_usize() + 1)).to_string()
    }

    pub fn span_to_filename(&self, sp: Span) -> FileName {
        self.lookup_char_pos(sp.lo).file.name.to_string()
    }

    pub fn span_to_lines(&self, sp: Span) -> FileLines {
        let lo = self.lookup_char_pos(sp.lo);
        let hi = self.lookup_char_pos(sp.hi);
        let mut lines = Vec::new();
        for i in lo.line - 1..hi.line as usize {
            lines.push(i);
        };
        FileLines {file: lo.file, lines: lines}
    }

    pub fn span_to_snippet(&self, sp: Span) -> Result<String, SpanSnippetError> {
        if sp.lo > sp.hi {
            return Err(SpanSnippetError::IllFormedSpan(sp));
        }

        let begin = self.lookup_byte_offset(sp.lo);
        let end = self.lookup_byte_offset(sp.hi);

        if begin.fm.start_pos != end.fm.start_pos {
            return Err(SpanSnippetError::DistinctSources(DistinctSources {
                begin: (begin.fm.name.clone(),
                        begin.fm.start_pos),
                end: (end.fm.name.clone(),
                      end.fm.start_pos)
            }));
        } else {
            let start = begin.pos.to_usize();
            let limit = end.pos.to_usize();
            if start > limit || limit > begin.fm.src.len() {
                return Err(SpanSnippetError::MalformedForCodemap(
                    MalformedCodemapPositions {
                        name: begin.fm.name.clone(),
                        source_len: begin.fm.src.len(),
                        begin_pos: begin.pos,
                        end_pos: end.pos,
                    }));
            }

            return Ok((&begin.fm.src[start..limit]).to_string())
        }
    }

    pub fn get_filemap(&self, filename: &str) -> Rc<FileMap> {
        for fm in &*self.files.borrow() {
            if filename == fm.name {
                return fm.clone();
            }
        }
        panic!("asking for {} which we don't know about", filename);
    }

    pub fn lookup_byte_offset(&self, bpos: BytePos) -> FileMapAndBytePos {
        let idx = self.lookup_filemap_idx(bpos);
        let fm = (*self.files.borrow())[idx].clone();
        let offset = bpos - fm.start_pos;
        FileMapAndBytePos {fm: fm, pos: offset}
    }

    /// Converts an absolute BytePos to a CharPos relative to the filemap and above.
    pub fn bytepos_to_file_charpos(&self, bpos: BytePos) -> CharPos {
        let idx = self.lookup_filemap_idx(bpos);
        let files = self.files.borrow();
        let map = &(*files)[idx];

        // The number of extra bytes due to multibyte chars in the FileMap
        let mut total_extra_bytes = 0;

        for mbc in &*map.multibyte_chars.borrow() {
            debug!("{}-byte char at {:?}", mbc.bytes, mbc.pos);
            if mbc.pos < bpos {
                // every character is at least one byte, so we only
                // count the actual extra bytes.
                total_extra_bytes += mbc.bytes - 1;
                // We should never see a byte position in the middle of a
                // character
                assert!(bpos.to_usize() >= mbc.pos.to_usize() + mbc.bytes);
            } else {
                break;
            }
        }

        assert!(map.start_pos.to_usize() + total_extra_bytes <= bpos.to_usize());
        CharPos(bpos.to_usize() - map.start_pos.to_usize() - total_extra_bytes)
    }

    fn lookup_filemap_idx(&self, pos: BytePos) -> usize {
        let files = self.files.borrow();
        let files = &*files;
        let len = files.len();
        let mut a = 0;
        let mut b = len;
        while b - a > 1 {
            let m = (a + b) / 2;
            if files[m].start_pos > pos {
                b = m;
            } else {
                a = m;
            }
        }
        // There can be filemaps with length 0. These have the same start_pos as
        // the previous filemap, but are not the filemaps we want (because they
        // are length 0, they cannot contain what we are looking for). So,
        // rewind until we find a useful filemap.
        loop {
            let lines = files[a].lines.borrow();
            let lines = lines;
            if lines.len() > 0 {
                break;
            }
            if a == 0 {
                panic!("position {} does not resolve to a source location",
                      pos.to_usize());
            }
            a -= 1;
        }
        if a >= len {
            panic!("position {} does not resolve to a source location",
                  pos.to_usize())
        }

        return a;
    }

    fn lookup_line(&self, pos: BytePos) -> FileMapAndLine {
        let idx = self.lookup_filemap_idx(pos);

        let files = self.files.borrow();
        let f = (*files)[idx].clone();
        let mut a = 0;
        {
            let lines = f.lines.borrow();
            let mut b = lines.len();
            while b - a > 1 {
                let m = (a + b) / 2;
                if (*lines)[m] > pos { b = m; } else { a = m; }
            }
        }
        FileMapAndLine {fm: f, line: a}
    }

    fn lookup_pos(&self, pos: BytePos) -> Loc {
        let FileMapAndLine {fm: f, line: a} = self.lookup_line(pos);
        let line = a + 1; // Line numbers start at 1
        let chpos = self.bytepos_to_file_charpos(pos);
        let linebpos = (*f.lines.borrow())[a];
        let linechpos = self.bytepos_to_file_charpos(linebpos);
        debug!("byte pos {:?} is on the line at byte pos {:?}",
               pos, linebpos);
        debug!("char pos {:?} is on the line at char pos {:?}",
               chpos, linechpos);
        debug!("byte is on line: {}", line);
        assert!(chpos >= linechpos);
        Loc {
            file: f,
            line: line,
            col: chpos - linechpos
        }
    }

    pub fn record_expansion(&self, expn_info: ExpnInfo) -> ExpnId {
        let mut expansions = self.expansions.borrow_mut();
        expansions.push(expn_info);
        ExpnId(expansions.len().to_u32().expect("too many ExpnInfo's!") - 1)
    }

    pub fn with_expn_info<T, F>(&self, id: ExpnId, f: F) -> T where
        F: FnOnce(Option<&ExpnInfo>) -> T,
    {
        match id {
            NO_EXPANSION => f(None),
            ExpnId(i) => f(Some(&(*self.expansions.borrow())[i as usize]))
        }
    }

    /// Check if a span is "internal" to a macro. This means that it is entirely generated by a
    /// macro expansion and contains no code that was passed in as an argument.
    pub fn span_is_internal(&self, span: Span) -> bool {
        // first, check if the given expression was generated by a macro or not
        // we need to go back the expn_info tree to check only the arguments
        // of the initial macro call, not the nested ones.
        let mut is_internal = false;
        let mut expnid = span.expn_id;
        while self.with_expn_info(expnid, |expninfo| {
            match expninfo {
                Some(ref info) => {
                    // save the parent expn_id for next loop iteration
                    expnid = info.call_site.expn_id;
                    if info.callee.name == "format_args" {
                        // This is a hack because the format_args builtin calls unstable APIs.
                        // I spent like 6 hours trying to solve this more generally but am stupid.
                        is_internal = true;
                        false
                    } else if info.callee.span.is_none() {
                        // it's a compiler built-in, we *really* don't want to mess with it
                        // so we skip it, unless it was called by a regular macro, in which case
                        // we will handle the caller macro next turn
                        is_internal = true;
                        true // continue looping
                    } else {
                        // was this expression from the current macro arguments ?
                        is_internal = !( span.lo > info.call_site.lo &&
                                         span.hi < info.call_site.hi );
                        true // continue looping
                    }
                },
                _ => false // stop looping
            }
        }) { /* empty while loop body */ }
        return is_internal;
    }
}

#[derive(Clone, PartialEq, Eq, Debug)]
pub enum SpanSnippetError {
    IllFormedSpan(Span),
    DistinctSources(DistinctSources),
    MalformedForCodemap(MalformedCodemapPositions),
}

#[derive(Clone, PartialEq, Eq, Debug)]
pub struct DistinctSources {
    begin: (String, BytePos),
    end: (String, BytePos)
}

#[derive(Clone, PartialEq, Eq, Debug)]
pub struct MalformedCodemapPositions {
    name: String,
    source_len: usize,
    begin_pos: BytePos,
    end_pos: BytePos
}

#[cfg(test)]
mod test {
    use super::*;

    #[test]
    fn t1 () {
        let cm = CodeMap::new();
        let fm = cm.new_filemap("blork.rs".to_string(),
                                "first line.\nsecond line".to_string());
        fm.next_line(BytePos(0));
        assert_eq!(fm.get_line(0), Some("first line.".to_string()));
        // TESTING BROKEN BEHAVIOR:
        fm.next_line(BytePos(10));
        assert_eq!(fm.get_line(1), Some(".".to_string()));
    }

    #[test]
    #[should_fail]
    fn t2 () {
        let cm = CodeMap::new();
        let fm = cm.new_filemap("blork.rs".to_string(),
                                "first line.\nsecond line".to_string());
        // TESTING *REALLY* BROKEN BEHAVIOR:
        fm.next_line(BytePos(0));
        fm.next_line(BytePos(10));
        fm.next_line(BytePos(2));
    }

    fn init_code_map() -> CodeMap {
        let cm = CodeMap::new();
        let fm1 = cm.new_filemap("blork.rs".to_string(),
                                 "first line.\nsecond line".to_string());
        let fm2 = cm.new_filemap("empty.rs".to_string(),
                                 "".to_string());
        let fm3 = cm.new_filemap("blork2.rs".to_string(),
                                 "first line.\nsecond line".to_string());

        fm1.next_line(BytePos(0));
        fm1.next_line(BytePos(12));
        fm2.next_line(BytePos(24));
        fm3.next_line(BytePos(24));
        fm3.next_line(BytePos(34));

        cm
    }

    #[test]
    fn t3() {
        // Test lookup_byte_offset
        let cm = init_code_map();

        let fmabp1 = cm.lookup_byte_offset(BytePos(22));
        assert_eq!(fmabp1.fm.name, "blork.rs");
        assert_eq!(fmabp1.pos, BytePos(22));

        let fmabp2 = cm.lookup_byte_offset(BytePos(24));
        assert_eq!(fmabp2.fm.name, "blork2.rs");
        assert_eq!(fmabp2.pos, BytePos(0));
    }

    #[test]
    fn t4() {
        // Test bytepos_to_file_charpos
        let cm = init_code_map();

        let cp1 = cm.bytepos_to_file_charpos(BytePos(22));
        assert_eq!(cp1, CharPos(22));

        let cp2 = cm.bytepos_to_file_charpos(BytePos(24));
        assert_eq!(cp2, CharPos(0));
    }

    #[test]
    fn t5() {
        // Test zero-length filemaps.
        let cm = init_code_map();

        let loc1 = cm.lookup_char_pos(BytePos(22));
        assert_eq!(loc1.file.name, "blork.rs");
        assert_eq!(loc1.line, 2);
        assert_eq!(loc1.col, CharPos(10));

        let loc2 = cm.lookup_char_pos(BytePos(24));
        assert_eq!(loc2.file.name, "blork2.rs");
        assert_eq!(loc2.line, 1);
        assert_eq!(loc2.col, CharPos(0));
    }

    fn init_code_map_mbc() -> CodeMap {
        let cm = CodeMap::new();
        // € is a three byte utf8 char.
        let fm1 =
            cm.new_filemap("blork.rs".to_string(),
                           "fir€st €€€€ line.\nsecond line".to_string());
        let fm2 = cm.new_filemap("blork2.rs".to_string(),
                                 "first line€€.\n€ second line".to_string());

        fm1.next_line(BytePos(0));
        fm1.next_line(BytePos(22));
        fm2.next_line(BytePos(40));
        fm2.next_line(BytePos(58));

        fm1.record_multibyte_char(BytePos(3), 3);
        fm1.record_multibyte_char(BytePos(9), 3);
        fm1.record_multibyte_char(BytePos(12), 3);
        fm1.record_multibyte_char(BytePos(15), 3);
        fm1.record_multibyte_char(BytePos(18), 3);
        fm2.record_multibyte_char(BytePos(50), 3);
        fm2.record_multibyte_char(BytePos(53), 3);
        fm2.record_multibyte_char(BytePos(58), 3);

        cm
    }

    #[test]
    fn t6() {
        // Test bytepos_to_file_charpos in the presence of multi-byte chars
        let cm = init_code_map_mbc();

        let cp1 = cm.bytepos_to_file_charpos(BytePos(3));
        assert_eq!(cp1, CharPos(3));

        let cp2 = cm.bytepos_to_file_charpos(BytePos(6));
        assert_eq!(cp2, CharPos(4));

        let cp3 = cm.bytepos_to_file_charpos(BytePos(56));
        assert_eq!(cp3, CharPos(12));

        let cp4 = cm.bytepos_to_file_charpos(BytePos(61));
        assert_eq!(cp4, CharPos(15));
    }

    #[test]
    fn t7() {
        // Test span_to_lines for a span ending at the end of filemap
        let cm = init_code_map();
        let span = Span {lo: BytePos(12), hi: BytePos(23), expn_id: NO_EXPANSION};
        let file_lines = cm.span_to_lines(span);

        assert_eq!(file_lines.file.name, "blork.rs");
        assert_eq!(file_lines.lines.len(), 1);
        assert_eq!(file_lines.lines[0], 1);
    }

    #[test]
    fn t8() {
        // Test span_to_snippet for a span ending at the end of filemap
        let cm = init_code_map();
        let span = Span {lo: BytePos(12), hi: BytePos(23), expn_id: NO_EXPANSION};
        let snippet = cm.span_to_snippet(span);

        assert_eq!(snippet, Ok("second line".to_string()));
    }

    #[test]
    fn t9() {
        // Test span_to_str for a span ending at the end of filemap
        let cm = init_code_map();
        let span = Span {lo: BytePos(12), hi: BytePos(23), expn_id: NO_EXPANSION};
        let sstr =  cm.span_to_string(span);

        assert_eq!(sstr, "blork.rs:2:1: 2:12");
    }
}