rust/src/libcore/io.rs

1433 lines
39 KiB
Rust
Raw Normal View History

// 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.
/*!
2011-12-06 15:58:45 -06:00
Basic input/output
2011-12-06 15:58:45 -06:00
*/
#[forbid(deprecated_mode)];
#[forbid(deprecated_pattern)];
2012-09-04 13:12:17 -05:00
use result::Result;
2012-09-04 13:12:17 -05:00
use cmp::Eq;
use dvec::DVec;
use int;
use libc;
2012-09-04 13:12:17 -05:00
use libc::{c_int, c_long, c_uint, c_void, size_t, ssize_t};
use libc::consts::os::posix88::*;
use libc::consts::os::extra::*;
use option;
use os;
use prelude::*;
use ptr;
use result;
use str;
use uint;
use vec;
2012-08-14 15:38:35 -05:00
#[allow(non_camel_case_types)] // not sure what to do about this
pub type fd_t = c_int;
#[abi = "cdecl"]
extern mod rustrt {
unsafe fn rust_get_stdin() -> *libc::FILE;
unsafe fn rust_get_stdout() -> *libc::FILE;
unsafe fn rust_get_stderr() -> *libc::FILE;
}
// Reading
// FIXME (#2004): This is all buffered. We might need an unbuffered variant
// as well
2012-10-01 16:02:10 -05:00
pub enum SeekStyle { SeekSet, SeekEnd, SeekCur, }
/// The raw underlying reader trait. All readers must implement this.
2012-10-01 16:02:10 -05:00
pub trait Reader {
// FIXME (#2004): Seekable really should be orthogonal.
/// Read up to len bytes (or EOF) and put them into bytes (which
/// must be at least len bytes long). Return number of bytes read.
// FIXME (#2982): This should probably return an error.
fn read(&self, bytes: &[mut u8], len: uint) -> uint;
/// Read a single byte, returning a negative value for EOF or read error.
fn read_byte(&self) -> int;
/// Return whether the stream is currently at EOF position.
fn eof(&self) -> bool;
/// Move the current position within the stream. The second parameter
/// determines the position that the first parameter is relative to.
fn seek(&self, position: int, style: SeekStyle);
/// Return the current position within the stream.
fn tell(&self) -> uint;
}
/// Generic utility functions defined on readers.
2012-10-01 16:02:10 -05:00
pub trait ReaderUtil {
/// Read len bytes into a new vec.
fn read_bytes(&self, len: uint) -> ~[u8];
/// Read up until the first '\n' char (which is not returned), or EOF.
fn read_line(&self) -> ~str;
/// Read n utf-8 encoded chars.
fn read_chars(&self, n: uint) -> ~[char];
/// Read a single utf-8 encoded char.
fn read_char(&self) -> char;
/// Read up until the first null byte (which is not returned), or EOF.
fn read_c_str(&self) -> ~str;
/// Read all the data remaining in the stream in one go.
fn read_whole_stream(&self) -> ~[u8];
/// Iterate over every byte until the iterator breaks or EOF.
fn each_byte(&self, it: fn(int) -> bool);
/// Iterate over every char until the iterator breaks or EOF.
fn each_char(&self, it: fn(char) -> bool);
/// Iterate over every line until the iterator breaks or EOF.
fn each_line(&self, it: fn(&str) -> bool);
/// Read n (between 1 and 8) little-endian unsigned integer bytes.
fn read_le_uint_n(&self, nbytes: uint) -> u64;
/// Read n (between 1 and 8) little-endian signed integer bytes.
fn read_le_int_n(&self, nbytes: uint) -> i64;
/// Read n (between 1 and 8) big-endian unsigned integer bytes.
fn read_be_uint_n(&self, nbytes: uint) -> u64;
/// Read n (between 1 and 8) big-endian signed integer bytes.
fn read_be_int_n(&self, nbytes: uint) -> i64;
/// Read a little-endian uint (number of bytes depends on system).
fn read_le_uint(&self) -> uint;
/// Read a little-endian int (number of bytes depends on system).
fn read_le_int(&self) -> int;
/// Read a big-endian uint (number of bytes depends on system).
fn read_be_uint(&self) -> uint;
/// Read a big-endian int (number of bytes depends on system).
fn read_be_int(&self) -> int;
/// Read a big-endian u64 (8 bytes).
fn read_be_u64(&self) -> u64;
/// Read a big-endian u32 (4 bytes).
fn read_be_u32(&self) -> u32;
/// Read a big-endian u16 (2 bytes).
fn read_be_u16(&self) -> u16;
/// Read a big-endian i64 (8 bytes).
fn read_be_i64(&self) -> i64;
/// Read a big-endian i32 (4 bytes).
fn read_be_i32(&self) -> i32;
/// Read a big-endian i16 (2 bytes).
fn read_be_i16(&self) -> i16;
/// Read a little-endian u64 (8 bytes).
fn read_le_u64(&self) -> u64;
/// Read a little-endian u32 (4 bytes).
fn read_le_u32(&self) -> u32;
/// Read a little-endian u16 (2 bytes).
fn read_le_u16(&self) -> u16;
/// Read a litle-endian i64 (8 bytes).
fn read_le_i64(&self) -> i64;
/// Read a litle-endian i32 (4 bytes).
fn read_le_i32(&self) -> i32;
/// Read a litle-endian i16 (2 bytes).
fn read_le_i16(&self) -> i16;
/// Read a u8 (1 byte).
fn read_u8(&self) -> u8;
/// Read a i8 (1 byte).
fn read_i8(&self) -> i8;
}
impl<T: Reader> T : ReaderUtil {
fn read_bytes(&self,len: uint) -> ~[u8] {
let mut bytes = vec::with_capacity(len);
unsafe { vec::raw::set_len(&mut bytes, len); }
let count = self.read(bytes, len);
unsafe { vec::raw::set_len(&mut bytes, count); }
move bytes
}
fn read_line(&self) -> ~str {
let mut bytes = ~[];
loop {
let ch = self.read_byte();
if ch == -1 || ch == 10 { break; }
bytes.push(ch as u8);
}
str::from_bytes(bytes)
}
fn read_chars(&self, n: uint) -> ~[char] {
// returns the (consumed offset, n_req), appends characters to &chars
fn chars_from_bytes<T: Reader>(bytes: &~[u8], chars: &mut ~[char])
2012-09-26 12:13:43 -05:00
-> (uint, uint) {
let mut i = 0;
let bytes_len = bytes.len();
while i < bytes_len {
let b0 = bytes[i];
let w = str::utf8_char_width(b0);
let end = i + w;
2012-09-26 12:13:43 -05:00
i += 1;
assert (w > 0);
if w == 1 {
chars.push(b0 as char);
loop;
}
// can't satisfy this char with the existing data
if end > bytes_len {
return (i - 1, end - bytes_len);
}
2012-09-26 12:13:43 -05:00
let mut val = 0;
while i < end {
let next = bytes[i] as int;
2012-09-26 12:13:43 -05:00
i += 1;
assert (next > -1);
assert (next & 192 == 128);
2012-09-26 12:13:43 -05:00
val <<= 6;
val += (next & 63) as uint;
}
// See str::char_at
2012-09-26 12:13:43 -05:00
val += ((b0 << ((w + 1) as u8)) as uint)
<< (w - 1) * 6 - w - 1u;
chars.push(val as char);
}
2012-09-26 12:13:43 -05:00
return (i, 0);
}
let mut bytes = ~[];
let mut chars = ~[];
// might need more bytes, but reading n will never over-read
let mut nbread = n;
2012-09-26 12:13:43 -05:00
while nbread > 0 {
let data = self.read_bytes(nbread);
2012-09-26 12:13:43 -05:00
if data.is_empty() {
// eof - FIXME (#2004): should we do something if
// we're split in a unicode char?
break;
}
bytes.push_all(data);
let (offset, nbreq) = chars_from_bytes::<T>(&bytes, &mut chars);
2012-09-26 12:13:43 -05:00
let ncreq = n - chars.len();
// again we either know we need a certain number of bytes
// to complete a character, or we make sure we don't
// over-read by reading 1-byte per char needed
nbread = if ncreq > nbreq { ncreq } else { nbreq };
2012-09-26 12:13:43 -05:00
if nbread > 0 {
bytes = vec::slice(bytes, offset, bytes.len());
}
}
move chars
}
fn read_char(&self) -> char {
2012-09-26 12:13:43 -05:00
let c = self.read_chars(1);
if vec::len(c) == 0 {
2012-08-01 19:30:05 -05:00
return -1 as char; // FIXME will this stay valid? // #2004
}
2012-09-26 12:13:43 -05:00
assert(vec::len(c) == 1);
2012-08-01 19:30:05 -05:00
return c[0];
}
fn read_c_str(&self) -> ~str {
let mut bytes: ~[u8] = ~[];
loop {
let ch = self.read_byte();
if ch < 1 { break; } else { bytes.push(ch as u8); }
}
str::from_bytes(bytes)
}
fn read_whole_stream(&self) -> ~[u8] {
let mut bytes: ~[u8] = ~[];
while !self.eof() { bytes.push_all(self.read_bytes(2048u)); }
move bytes
}
fn each_byte(&self, it: fn(int) -> bool) {
while !self.eof() {
if !it(self.read_byte()) { break; }
}
}
fn each_char(&self, it: fn(char) -> bool) {
while !self.eof() {
if !it(self.read_char()) { break; }
}
}
fn each_line(&self, it: fn(s: &str) -> bool) {
while !self.eof() {
if !it(self.read_line()) { break; }
}
}
// FIXME int reading methods need to deal with eof - issue #2004
fn read_le_uint_n(&self, nbytes: uint) -> u64 {
assert nbytes > 0 && nbytes <= 8;
let mut val = 0u64, pos = 0, i = nbytes;
while i > 0 {
val += (self.read_u8() as u64) << pos;
pos += 8;
i -= 1;
}
val
}
fn read_le_int_n(&self, nbytes: uint) -> i64 {
extend_sign(self.read_le_uint_n(nbytes), nbytes)
}
fn read_be_uint_n(&self, nbytes: uint) -> u64 {
assert nbytes > 0 && nbytes <= 8;
let mut val = 0u64, i = nbytes;
while i > 0 {
i -= 1;
val += (self.read_u8() as u64) << i * 8;
}
val
}
fn read_be_int_n(&self, nbytes: uint) -> i64 {
extend_sign(self.read_be_uint_n(nbytes), nbytes)
}
fn read_le_uint(&self) -> uint {
self.read_le_uint_n(uint::bytes) as uint
}
fn read_le_int(&self) -> int {
self.read_le_int_n(int::bytes) as int
}
fn read_be_uint(&self) -> uint {
self.read_be_uint_n(uint::bytes) as uint
}
fn read_be_int(&self) -> int {
self.read_be_int_n(int::bytes) as int
}
fn read_be_u64(&self) -> u64 {
self.read_be_uint_n(8) as u64
}
fn read_be_u32(&self) -> u32 {
self.read_be_uint_n(4) as u32
}
fn read_be_u16(&self) -> u16 {
self.read_be_uint_n(2) as u16
}
fn read_be_i64(&self) -> i64 {
self.read_be_int_n(8) as i64
}
fn read_be_i32(&self) -> i32 {
self.read_be_int_n(4) as i32
}
fn read_be_i16(&self) -> i16 {
self.read_be_int_n(2) as i16
}
fn read_le_u64(&self) -> u64 {
self.read_le_uint_n(8) as u64
}
fn read_le_u32(&self) -> u32 {
self.read_le_uint_n(4) as u32
}
fn read_le_u16(&self) -> u16 {
self.read_le_uint_n(2) as u16
}
fn read_le_i64(&self) -> i64 {
self.read_le_int_n(8) as i64
}
fn read_le_i32(&self) -> i32 {
self.read_le_int_n(4) as i32
}
fn read_le_i16(&self) -> i16 {
self.read_le_int_n(2) as i16
}
fn read_u8(&self) -> u8 {
self.read_byte() as u8
}
fn read_i8(&self) -> i8 {
self.read_byte() as i8
}
}
fn extend_sign(val: u64, nbytes: uint) -> i64 {
let shift = (8 - nbytes) * 8;
(val << shift) as i64 >> shift
}
// Reader implementations
2012-08-14 15:38:35 -05:00
fn convert_whence(whence: SeekStyle) -> i32 {
2012-08-06 14:34:08 -05:00
return match whence {
2012-08-14 15:38:35 -05:00
SeekSet => 0i32,
SeekCur => 1i32,
SeekEnd => 2i32
};
}
2012-08-14 15:38:35 -05:00
impl *libc::FILE: Reader {
fn read(&self, bytes: &[mut u8], len: uint) -> uint {
unsafe {
do vec::as_mut_buf(bytes) |buf_p, buf_len| {
assert buf_len >= len;
let count = libc::fread(buf_p as *mut c_void, 1u as size_t,
len as size_t, *self);
count as uint
}
}
}
fn read_byte(&self) -> int {
unsafe {
libc::fgetc(*self) as int
}
}
fn eof(&self) -> bool {
unsafe {
return libc::feof(*self) != 0 as c_int;
}
}
fn seek(&self, offset: int, whence: SeekStyle) {
unsafe {
assert libc::fseek(*self,
offset as c_long,
convert_whence(whence)) == 0 as c_int;
}
}
fn tell(&self) -> uint {
unsafe {
return libc::ftell(*self) as uint;
}
}
}
struct Wrapper<T, C> {
base: T,
cleanup: C,
}
// A forwarding impl of reader that also holds on to a resource for the
// duration of its lifetime.
// FIXME there really should be a better way to do this // #2004
impl<R: Reader, C> Wrapper<R, C>: Reader {
fn read(&self, bytes: &[mut u8], len: uint) -> uint {
self.base.read(bytes, len)
}
fn read_byte(&self) -> int { self.base.read_byte() }
fn eof(&self) -> bool { self.base.eof() }
fn seek(&self, off: int, whence: SeekStyle) {
self.base.seek(off, whence)
}
fn tell(&self) -> uint { self.base.tell() }
}
pub struct FILERes {
2012-09-06 21:40:15 -05:00
f: *libc::FILE,
drop {
unsafe {
libc::fclose(self.f);
}
}
2012-06-21 23:30:16 -05:00
}
pub fn FILERes(f: *libc::FILE) -> FILERes {
2012-09-04 17:23:28 -05:00
FILERes {
f: f
}
}
2012-10-01 16:02:10 -05:00
pub fn FILE_reader(f: *libc::FILE, cleanup: bool) -> Reader {
if cleanup {
Wrapper { base: f, cleanup: FILERes(f) } as Reader
} else {
2012-08-14 15:38:35 -05:00
f as Reader
}
}
// FIXME (#2004): this should either be an trait-less impl, a set of
// top-level functions that take a reader, or a set of default methods on
// reader (which can then be called reader)
pub fn stdin() -> Reader {
unsafe {
rustrt::rust_get_stdin() as Reader
}
}
2012-10-01 16:02:10 -05:00
pub fn file_reader(path: &Path) -> Result<Reader, ~str> {
unsafe {
let f = os::as_c_charp(path.to_str(), |pathbuf| {
os::as_c_charp("r", |modebuf|
libc::fopen(pathbuf, modebuf)
)
});
return if f as uint == 0u { result::Err(~"error opening "
+ path.to_str()) }
else {
result::Ok(FILE_reader(f, true))
}
}
}
2012-10-18 11:06:53 -05:00
// Byte readers
pub struct BytesReader {
bytes: &[u8],
mut pos: uint
}
2011-07-27 07:19:39 -05:00
2012-10-18 11:06:53 -05:00
impl BytesReader: Reader {
fn read(&self, bytes: &[mut u8], len: uint) -> uint {
2012-10-18 11:06:53 -05:00
let count = uint::min(len, self.bytes.len() - self.pos);
2012-10-18 11:06:53 -05:00
let view = vec::view(self.bytes, self.pos, self.bytes.len());
vec::bytes::copy_memory(bytes, view, count);
self.pos += count;
count
}
fn read_byte(&self) -> int {
2012-10-18 11:06:53 -05:00
if self.pos == self.bytes.len() { return -1; }
let b = self.bytes[self.pos];
self.pos += 1u;
2012-08-01 19:30:05 -05:00
return b as int;
}
fn eof(&self) -> bool { self.pos == self.bytes.len() }
fn seek(&self, offset: int, whence: SeekStyle) {
let pos = self.pos;
2012-10-18 11:06:53 -05:00
self.pos = seek_in_buf(offset, pos, self.bytes.len(), whence);
}
fn tell(&self) -> uint { self.pos }
}
pub pure fn with_bytes_reader<t>(bytes: &[u8], f: fn(Reader) -> t) -> t {
2012-10-18 11:06:53 -05:00
f(BytesReader { bytes: bytes, pos: 0u } as Reader)
}
pub pure fn with_str_reader<T>(s: &str, f: fn(Reader) -> T) -> T {
str::byte_slice(s, |bytes| with_bytes_reader(bytes, f))
}
// Writing
2012-10-01 16:02:10 -05:00
pub enum FileFlag { Append, Create, Truncate, NoFlag, }
// What type of writer are we?
2012-12-11 17:16:36 -06:00
#[deriving_eq]
2012-10-01 16:02:10 -05:00
pub enum WriterType { Screen, File }
// FIXME (#2004): Seekable really should be orthogonal.
// FIXME (#2004): eventually u64
/// The raw underlying writer trait. All writers must implement this.
2012-10-01 16:02:10 -05:00
pub trait Writer {
/// Write all of the given bytes.
fn write(&self, v: &[const u8]);
/// Move the current position within the stream. The second parameter
/// determines the position that the first parameter is relative to.
fn seek(&self, int, SeekStyle);
/// Return the current position within the stream.
fn tell(&self) -> uint;
/// Flush the output buffer for this stream (if there is one).
fn flush(&self) -> int;
/// Determine if this Writer is writing to a file or not.
fn get_type(&self) -> WriterType;
}
impl<W: Writer, C> Wrapper<W, C>: Writer {
fn write(&self, bs: &[const u8]) { self.base.write(bs); }
fn seek(&self, off: int, style: SeekStyle) { self.base.seek(off, style); }
fn tell(&self) -> uint { self.base.tell() }
fn flush(&self) -> int { self.base.flush() }
fn get_type(&self) -> WriterType { File }
}
2012-08-14 15:38:35 -05:00
impl *libc::FILE: Writer {
fn write(&self, v: &[const u8]) {
unsafe {
do vec::as_const_buf(v) |vbuf, len| {
let nout = libc::fwrite(vbuf as *c_void,
1,
len as size_t,
*self);
if nout != len as size_t {
error!("error writing buffer");
log(error, os::last_os_error());
fail;
}
}
}
}
fn seek(&self, offset: int, whence: SeekStyle) {
unsafe {
assert libc::fseek(*self,
offset as c_long,
convert_whence(whence)) == 0 as c_int;
}
}
fn tell(&self) -> uint {
unsafe {
libc::ftell(*self) as uint
}
}
fn flush(&self) -> int {
unsafe {
libc::fflush(*self) as int
}
}
fn get_type(&self) -> WriterType {
unsafe {
let fd = libc::fileno(*self);
if libc::isatty(fd) == 0 { File }
else { Screen }
}
}
}
2012-10-01 16:02:10 -05:00
pub fn FILE_writer(f: *libc::FILE, cleanup: bool) -> Writer {
if cleanup {
Wrapper { base: f, cleanup: FILERes(f) } as Writer
} else {
2012-08-14 15:38:35 -05:00
f as Writer
}
}
2012-08-14 15:38:35 -05:00
impl fd_t: Writer {
fn write(&self, v: &[const u8]) {
unsafe {
let mut count = 0u;
do vec::as_const_buf(v) |vbuf, len| {
while count < len {
let vb = ptr::const_offset(vbuf, count) as *c_void;
let nout = libc::write(*self, vb, len as size_t);
if nout < 0 as ssize_t {
error!("error writing buffer");
log(error, os::last_os_error());
fail;
}
count += nout as uint;
}
}
}
}
fn seek(&self, _offset: int, _whence: SeekStyle) {
2012-08-22 19:24:52 -05:00
error!("need 64-bit foreign calls for seek, sorry");
fail;
}
fn tell(&self) -> uint {
2012-08-22 19:24:52 -05:00
error!("need 64-bit foreign calls for tell, sorry");
fail;
2011-07-27 10:05:34 -05:00
}
fn flush(&self) -> int { 0 }
fn get_type(&self) -> WriterType {
unsafe {
if libc::isatty(*self) == 0 { File } else { Screen }
}
}
}
pub struct FdRes {
2012-09-06 21:40:15 -05:00
fd: fd_t,
drop {
unsafe {
libc::close(self.fd);
}
}
2012-06-21 23:30:16 -05:00
}
pub fn FdRes(fd: fd_t) -> FdRes {
2012-09-04 17:23:28 -05:00
FdRes {
fd: fd
}
}
2012-10-01 16:02:10 -05:00
pub fn fd_writer(fd: fd_t, cleanup: bool) -> Writer {
if cleanup {
Wrapper { base: fd, cleanup: FdRes(fd) } as Writer
} else {
2012-08-14 15:38:35 -05:00
fd as Writer
}
}
2012-10-01 16:02:10 -05:00
pub fn mk_file_writer(path: &Path, flags: &[FileFlag])
2012-08-26 18:54:31 -05:00
-> Result<Writer, ~str> {
#[cfg(windows)]
fn wb() -> c_int { (O_WRONLY | O_BINARY) as c_int }
#[cfg(unix)]
fn wb() -> c_int { O_WRONLY as c_int }
let mut fflags: c_int = wb();
for vec::each(flags) |f| {
match *f {
2012-08-14 15:38:35 -05:00
Append => fflags |= O_APPEND as c_int,
Create => fflags |= O_CREAT as c_int,
Truncate => fflags |= O_TRUNC as c_int,
NoFlag => ()
}
}
let fd = unsafe {
do os::as_c_charp(path.to_str()) |pathbuf| {
libc::open(pathbuf, fflags,
(S_IRUSR | S_IWUSR) as c_int)
}
};
if fd < (0 as c_int) {
2012-08-26 18:54:31 -05:00
result::Err(fmt!("error opening %s: %s", path.to_str(),
os::last_os_error()))
} else {
2012-08-26 18:54:31 -05:00
result::Ok(fd_writer(fd, true))
}
}
2012-10-01 16:02:10 -05:00
pub fn u64_to_le_bytes<T>(n: u64, size: uint,
f: fn(v: &[u8]) -> T) -> T {
assert size <= 8u;
2012-08-06 14:34:08 -05:00
match size {
2012-08-03 21:59:04 -05:00
1u => f(&[n as u8]),
2u => f(&[n as u8,
(n >> 8) as u8]),
4u => f(&[n as u8,
(n >> 8) as u8,
(n >> 16) as u8,
2012-08-03 21:59:04 -05:00
(n >> 24) as u8]),
8u => f(&[n as u8,
(n >> 8) as u8,
(n >> 16) as u8,
(n >> 24) as u8,
(n >> 32) as u8,
(n >> 40) as u8,
(n >> 48) as u8,
2012-08-03 21:59:04 -05:00
(n >> 56) as u8]),
_ => {
let mut bytes: ~[u8] = ~[], i = size, n = n;
while i > 0u {
bytes.push((n & 255_u64) as u8);
n >>= 8_u64;
i -= 1u;
}
f(bytes)
}
}
}
2012-10-01 16:02:10 -05:00
pub fn u64_to_be_bytes<T>(n: u64, size: uint,
f: fn(v: &[u8]) -> T) -> T {
assert size <= 8u;
2012-08-06 14:34:08 -05:00
match size {
2012-08-03 21:59:04 -05:00
1u => f(&[n as u8]),
2u => f(&[(n >> 8) as u8,
n as u8]),
4u => f(&[(n >> 24) as u8,
(n >> 16) as u8,
(n >> 8) as u8,
2012-08-03 21:59:04 -05:00
n as u8]),
8u => f(&[(n >> 56) as u8,
(n >> 48) as u8,
(n >> 40) as u8,
(n >> 32) as u8,
(n >> 24) as u8,
(n >> 16) as u8,
(n >> 8) as u8,
2012-08-03 21:59:04 -05:00
n as u8]),
_ => {
let mut bytes: ~[u8] = ~[];
let mut i = size;
while i > 0u {
let shift = ((i - 1u) * 8u) as u64;
bytes.push((n >> shift) as u8);
i -= 1u;
}
f(bytes)
}
}
}
2012-10-01 16:02:10 -05:00
pub fn u64_from_be_bytes(data: &[const u8],
start: uint, size: uint) -> u64 {
let mut sz = size;
assert (sz <= 8u);
let mut val = 0_u64;
let mut pos = start;
while sz > 0u {
sz -= 1u;
val += (data[pos] as u64) << ((sz * 8u) as u64);
pos += 1u;
}
2012-08-01 19:30:05 -05:00
return val;
}
// FIXME: #3048 combine trait+impl (or just move these to
// default methods on writer)
/// Generic utility functions defined on writers.
2012-10-01 16:02:10 -05:00
pub trait WriterUtil {
/// Write a single utf-8 encoded char.
fn write_char(&self, ch: char);
/// Write every char in the given str, encoded as utf-8.
fn write_str(&self, s: &str);
/// Write the given str, as utf-8, followed by '\n'.
fn write_line(&self, s: &str);
/// Write the result of passing n through `int::to_str_bytes`.
fn write_int(&self, n: int);
/// Write the result of passing n through `uint::to_str_bytes`.
fn write_uint(&self, n: uint);
/// Write a little-endian uint (number of bytes depends on system).
fn write_le_uint(&self, n: uint);
/// Write a little-endian int (number of bytes depends on system).
fn write_le_int(&self, n: int);
/// Write a big-endian uint (number of bytes depends on system).
fn write_be_uint(&self, n: uint);
/// Write a big-endian int (number of bytes depends on system).
fn write_be_int(&self, n: int);
/// Write a big-endian u64 (8 bytes).
fn write_be_u64(&self, n: u64);
/// Write a big-endian u32 (4 bytes).
fn write_be_u32(&self, n: u32);
/// Write a big-endian u16 (2 bytes).
fn write_be_u16(&self, n: u16);
/// Write a big-endian i64 (8 bytes).
fn write_be_i64(&self, n: i64);
/// Write a big-endian i32 (4 bytes).
fn write_be_i32(&self, n: i32);
/// Write a big-endian i16 (2 bytes).
fn write_be_i16(&self, n: i16);
/// Write a little-endian u64 (8 bytes).
fn write_le_u64(&self, n: u64);
/// Write a little-endian u32 (4 bytes).
fn write_le_u32(&self, n: u32);
/// Write a little-endian u16 (2 bytes).
fn write_le_u16(&self, n: u16);
/// Write a little-endian i64 (8 bytes).
fn write_le_i64(&self, n: i64);
/// Write a little-endian i32 (4 bytes).
fn write_le_i32(&self, n: i32);
/// Write a little-endian i16 (2 bytes).
fn write_le_i16(&self, n: i16);
/// Write a u8 (1 byte).
fn write_u8(&self, n: u8);
/// Write a i8 (1 byte).
fn write_i8(&self, n: i8);
}
2012-08-14 15:38:35 -05:00
impl<T: Writer> T : WriterUtil {
fn write_char(&self, ch: char) {
if ch as uint < 128u {
self.write(&[ch as u8]);
} else {
self.write_str(str::from_char(ch));
}
}
fn write_str(&self, s: &str) { str::byte_slice(s, |v| self.write(v)) }
fn write_line(&self, s: &str) {
self.write_str(s);
self.write_str(&"\n");
}
fn write_int(&self, n: int) {
int::to_str_bytes(n, 10u, |bytes| self.write(bytes))
}
fn write_uint(&self, n: uint) {
uint::to_str_bytes(false, n, 10u, |bytes| self.write(bytes))
}
fn write_le_uint(&self, n: uint) {
u64_to_le_bytes(n as u64, uint::bytes, |v| self.write(v))
}
fn write_le_int(&self, n: int) {
u64_to_le_bytes(n as u64, int::bytes, |v| self.write(v))
}
fn write_be_uint(&self, n: uint) {
u64_to_be_bytes(n as u64, uint::bytes, |v| self.write(v))
}
fn write_be_int(&self, n: int) {
u64_to_be_bytes(n as u64, int::bytes, |v| self.write(v))
}
fn write_be_u64(&self, n: u64) {
2012-06-30 18:19:07 -05:00
u64_to_be_bytes(n, 8u, |v| self.write(v))
}
fn write_be_u32(&self, n: u32) {
2012-06-30 18:19:07 -05:00
u64_to_be_bytes(n as u64, 4u, |v| self.write(v))
}
fn write_be_u16(&self, n: u16) {
2012-06-30 18:19:07 -05:00
u64_to_be_bytes(n as u64, 2u, |v| self.write(v))
}
fn write_be_i64(&self, n: i64) {
2012-06-30 18:19:07 -05:00
u64_to_be_bytes(n as u64, 8u, |v| self.write(v))
}
fn write_be_i32(&self, n: i32) {
2012-06-30 18:19:07 -05:00
u64_to_be_bytes(n as u64, 4u, |v| self.write(v))
}
fn write_be_i16(&self, n: i16) {
2012-06-30 18:19:07 -05:00
u64_to_be_bytes(n as u64, 2u, |v| self.write(v))
}
fn write_le_u64(&self, n: u64) {
2012-06-30 18:19:07 -05:00
u64_to_le_bytes(n, 8u, |v| self.write(v))
}
fn write_le_u32(&self, n: u32) {
2012-06-30 18:19:07 -05:00
u64_to_le_bytes(n as u64, 4u, |v| self.write(v))
}
fn write_le_u16(&self, n: u16) {
2012-06-30 18:19:07 -05:00
u64_to_le_bytes(n as u64, 2u, |v| self.write(v))
}
fn write_le_i64(&self, n: i64) {
2012-06-30 18:19:07 -05:00
u64_to_le_bytes(n as u64, 8u, |v| self.write(v))
}
fn write_le_i32(&self, n: i32) {
2012-06-30 18:19:07 -05:00
u64_to_le_bytes(n as u64, 4u, |v| self.write(v))
}
fn write_le_i16(&self, n: i16) {
2012-06-30 18:19:07 -05:00
u64_to_le_bytes(n as u64, 2u, |v| self.write(v))
}
fn write_u8(&self, n: u8) { self.write([n]) }
fn write_i8(&self, n: i8) { self.write([n as u8]) }
}
2012-09-02 17:37:15 -05:00
#[allow(non_implicitly_copyable_typarams)]
2012-10-01 16:02:10 -05:00
pub fn file_writer(path: &Path, flags: &[FileFlag]) -> Result<Writer, ~str> {
2012-09-25 18:23:04 -05:00
mk_file_writer(path, flags).chain(|w| result::Ok(w))
}
// FIXME: fileflags // #2004
2012-10-01 16:02:10 -05:00
pub fn buffered_file_writer(path: &Path) -> Result<Writer, ~str> {
unsafe {
let f = do os::as_c_charp(path.to_str()) |pathbuf| {
do os::as_c_charp("w") |modebuf| {
libc::fopen(pathbuf, modebuf)
}
};
return if f as uint == 0u { result::Err(~"error opening "
+ path.to_str()) }
else { result::Ok(FILE_writer(f, true)) }
}
}
// FIXME (#2004) it would be great if this could be a const
// FIXME (#2004) why are these different from the way stdin() is
// implemented?
2012-10-01 16:02:10 -05:00
pub fn stdout() -> Writer { fd_writer(libc::STDOUT_FILENO as c_int, false) }
pub fn stderr() -> Writer { fd_writer(libc::STDERR_FILENO as c_int, false) }
2012-10-01 16:02:10 -05:00
pub fn print(s: &str) { stdout().write_str(s); }
pub fn println(s: &str) { stdout().write_line(s); }
2012-10-01 16:02:10 -05:00
pub struct BytesWriter {
bytes: DVec<u8>,
mut pos: uint,
}
impl BytesWriter: Writer {
fn write(&self, v: &[const u8]) {
do self.bytes.swap |bytes| {
let mut bytes = move bytes;
let v_len = v.len();
let bytes_len = bytes.len();
let count = uint::max(bytes_len, self.pos + v_len);
vec::reserve(&mut bytes, count);
unsafe { vec::raw::set_len(&mut bytes, count); }
{
let view = vec::mut_view(bytes, self.pos, count);
vec::bytes::copy_memory(view, v, v_len);
}
self.pos += v_len;
2012-05-11 12:44:57 -05:00
move bytes
}
}
fn seek(&self, offset: int, whence: SeekStyle) {
let pos = self.pos;
let len = self.bytes.len();
self.pos = seek_in_buf(offset, pos, len, whence);
}
fn tell(&self) -> uint { self.pos }
fn flush(&self) -> int { 0 }
fn get_type(&self) -> WriterType { File }
}
pub pure fn BytesWriter() -> BytesWriter {
BytesWriter { bytes: DVec(), mut pos: 0u }
}
pub pure fn with_bytes_writer(f: fn(Writer)) -> ~[u8] {
let wr = @BytesWriter();
f(wr as Writer);
// FIXME (#3758): This should not be needed.
unsafe { wr.bytes.check_out(|bytes| move bytes) }
2012-02-21 11:23:01 -06:00
}
pub pure fn with_str_writer(f: fn(Writer)) -> ~str {
let mut v = with_bytes_writer(f);
// FIXME (#3758): This should not be needed.
unsafe {
// Make sure the vector has a trailing null and is proper utf8.
v.push(0);
}
assert str::is_utf8(v);
2012-09-19 00:35:28 -05:00
unsafe { move ::cast::transmute(move v) }
}
// Utility functions
2012-10-01 16:02:10 -05:00
pub fn seek_in_buf(offset: int, pos: uint, len: uint, whence: SeekStyle) ->
2011-07-27 07:19:39 -05:00
uint {
let mut bpos = pos as int;
2011-07-27 07:19:39 -05:00
let blen = len as int;
2012-08-06 14:34:08 -05:00
match whence {
2012-08-14 15:38:35 -05:00
SeekSet => bpos = offset,
SeekCur => bpos += offset,
SeekEnd => bpos = blen + offset
}
if bpos < 0 { bpos = 0; } else if bpos > blen { bpos = blen; }
2012-08-01 19:30:05 -05:00
return bpos as uint;
}
2012-09-02 17:37:15 -05:00
#[allow(non_implicitly_copyable_typarams)]
2012-10-01 16:02:10 -05:00
pub fn read_whole_file_str(file: &Path) -> Result<~str, ~str> {
2012-06-30 18:19:07 -05:00
result::chain(read_whole_file(file), |bytes| {
if str::is_utf8(bytes) {
2012-08-26 18:54:31 -05:00
result::Ok(str::from_bytes(bytes))
} else {
2012-08-26 18:54:31 -05:00
result::Err(file.to_str() + ~" is not UTF-8")
}
})
}
// FIXME (#2004): implement this in a low-level way. Going through the
// abstractions is pointless.
2012-09-02 17:37:15 -05:00
#[allow(non_implicitly_copyable_typarams)]
2012-10-01 16:02:10 -05:00
pub fn read_whole_file(file: &Path) -> Result<~[u8], ~str> {
2012-06-30 18:19:07 -05:00
result::chain(file_reader(file), |rdr| {
2012-08-26 18:54:31 -05:00
result::Ok(rdr.read_whole_stream())
})
}
// fsync related
2012-10-01 16:02:10 -05:00
pub mod fsync {
use prelude::*;
use io::{FILERes, FdRes, fd_t};
use libc;
use os;
2012-10-01 16:02:10 -05:00
pub enum Level {
// whatever fsync does on that platform
2012-08-14 15:38:35 -05:00
FSync,
// fdatasync on linux, similiar or more on other platforms
2012-08-14 15:38:35 -05:00
FDataSync,
// full fsync
//
// You must additionally sync the parent directory as well!
2012-08-14 15:38:35 -05:00
FullFSync,
}
2012-06-21 23:30:16 -05:00
// Artifacts that need to fsync on destruction
2012-10-01 16:02:10 -05:00
pub struct Res<t: Copy> {
2012-09-06 21:40:15 -05:00
arg: Arg<t>,
}
impl<T: Copy> Res<T>: Drop {
fn finalize(&self) {
2012-08-06 14:34:08 -05:00
match self.arg.opt_level {
None => (),
Some(level) => {
2012-06-21 23:30:16 -05:00
// fail hard if not succesful
assert((self.arg.fsync_fn)(self.arg.val, level) != -1);
2012-06-21 23:30:16 -05:00
}
}
}
}
pub fn Res<t: Copy>(arg: Arg<t>) -> Res<t>{
2012-09-04 17:23:28 -05:00
Res {
arg: move arg
}
}
pub struct Arg<t> {
val: t,
2012-08-20 14:23:37 -05:00
opt_level: Option<Level>,
fsync_fn: fn@(f: t, Level) -> int,
}
// fsync file after executing blk
// FIXME (#2004) find better way to create resources within lifetime of
// outer res
2012-10-01 16:02:10 -05:00
pub fn FILE_res_sync(file: &FILERes, opt_level: Option<Level>,
blk: fn(v: Res<*libc::FILE>)) {
unsafe {
blk(Res(Arg {
val: file.f, opt_level: opt_level,
fsync_fn: fn@(file: *libc::FILE, l: Level) -> int {
unsafe {
return os::fsync_fd(libc::fileno(file), l) as int;
}
}
}));
}
}
// fsync fd after executing blk
2012-10-01 16:02:10 -05:00
pub fn fd_res_sync(fd: &FdRes, opt_level: Option<Level>,
blk: fn(v: Res<fd_t>)) {
blk(Res(Arg {
2012-06-21 23:30:16 -05:00
val: fd.fd, opt_level: opt_level,
fsync_fn: fn@(fd: fd_t, l: Level) -> int {
2012-08-01 19:30:05 -05:00
return os::fsync_fd(fd, l) as int;
}
}));
}
// Type of objects that may want to fsync
pub trait FSyncable { fn fsync(&self, l: Level) -> int; }
// Call o.fsync after executing blk
pub fn obj_sync(o: FSyncable, opt_level: Option<Level>,
blk: fn(v: Res<FSyncable>)) {
blk(Res(Arg {
val: o, opt_level: opt_level,
fsync_fn: fn@(o: FSyncable, l: Level) -> int {
2012-08-14 15:38:35 -05:00
return o.fsync(l);
}
}));
}
}
2012-01-17 21:05:07 -06:00
#[cfg(test)]
mod tests {
use debug;
use i32;
use io::{BytesWriter, SeekCur, SeekEnd, SeekSet};
use io;
use path::Path;
use result;
use str;
use u64;
use vec;
2012-01-17 21:05:07 -06:00
#[test]
fn test_simple() {
let tmpfile = &Path("tmp/lib-io-test-simple.tmp");
2012-01-17 21:05:07 -06:00
log(debug, tmpfile);
let frood: ~str =
~"A hoopy frood who really knows where his towel is.";
log(debug, copy frood);
2012-01-17 21:05:07 -06:00
{
2012-08-14 15:38:35 -05:00
let out: io::Writer =
2012-01-17 21:05:07 -06:00
result::get(
2012-09-25 18:23:04 -05:00
&io::file_writer(tmpfile, ~[io::Create, io::Truncate]));
2012-01-17 21:05:07 -06:00
out.write_str(frood);
}
2012-09-25 18:23:04 -05:00
let inp: io::Reader = result::get(&io::file_reader(tmpfile));
let frood2: ~str = inp.read_c_str();
log(debug, copy frood2);
assert frood == frood2;
2012-01-17 21:05:07 -06:00
}
#[test]
fn test_readchars_empty() {
do io::with_str_reader(~"") |inp| {
let res : ~[char] = inp.read_chars(128);
assert(vec::len(res) == 0);
}
2012-01-17 21:05:07 -06:00
}
#[test]
fn test_read_line_utf8() {
do io::with_str_reader(~"生锈的汤匙切肉汤hello生锈的汤匙切肉汤") |inp| {
let line = inp.read_line();
assert line == ~"生锈的汤匙切肉汤hello生锈的汤匙切肉汤";
}
}
2012-01-17 21:05:07 -06:00
#[test]
fn test_readchars_wide() {
let wide_test = ~"生锈的汤匙切肉汤hello生锈的汤匙切肉汤";
let ivals : ~[int] = ~[
2012-01-17 21:05:07 -06:00
29983, 38152, 30340, 27748,
21273, 20999, 32905, 27748,
104, 101, 108, 108, 111,
29983, 38152, 30340, 27748,
21273, 20999, 32905, 27748];
fn check_read_ln(len : uint, s: &str, ivals: &[int]) {
do io::with_str_reader(s) |inp| {
let res : ~[char] = inp.read_chars(len);
if (len <= vec::len(ivals)) {
assert(vec::len(res) == len);
}
assert(vec::slice(ivals, 0u, vec::len(res)) ==
vec::map(res, |x| *x as int));
2012-01-17 21:05:07 -06:00
}
}
let mut i = 0;
while i < 8 {
2012-01-17 21:05:07 -06:00
check_read_ln(i, wide_test, ivals);
i += 1;
2012-01-17 21:05:07 -06:00
}
// check a long read for good measure
check_read_ln(128, wide_test, ivals);
2012-01-17 21:05:07 -06:00
}
#[test]
fn test_readchar() {
do io::with_str_reader(~"") |inp| {
let res : char = inp.read_char();
assert(res as int == 29983);
}
2012-01-17 21:05:07 -06:00
}
#[test]
fn test_readchar_empty() {
do io::with_str_reader(~"") |inp| {
let res : char = inp.read_char();
assert(res as int == -1);
}
2012-01-17 21:05:07 -06:00
}
#[test]
fn file_reader_not_exist() {
match io::file_reader(&Path("not a file")) {
2012-09-28 15:00:07 -05:00
result::Err(copy e) => {
assert e == ~"error opening not a file";
2012-01-17 21:05:07 -06:00
}
2012-08-26 18:54:31 -05:00
result::Ok(_) => fail
2012-01-17 21:05:07 -06:00
}
}
#[test]
#[should_fail]
2012-11-25 15:51:43 -06:00
#[ignore(cfg(windows))]
fn test_read_buffer_too_small() {
let path = &Path("tmp/lib-io-test-read-buffer-too-small.tmp");
// ensure the file exists
io::file_writer(path, [io::Create]).get();
let file = io::file_reader(path).get();
let mut buf = vec::from_elem(5, 0);
file.read(buf, 6); // this should fail because buf is too small
}
#[test]
fn test_read_buffer_big_enough() {
let path = &Path("tmp/lib-io-test-read-buffer-big-enough.tmp");
// ensure the file exists
io::file_writer(path, [io::Create]).get();
let file = io::file_reader(path).get();
let mut buf = vec::from_elem(5, 0);
file.read(buf, 4); // this should succeed because buf is big enough
}
#[test]
fn test_write_empty() {
let file = io::file_writer(&Path("tmp/lib-io-test-write-empty.tmp"),
[io::Create]).get();
file.write([]);
}
2012-01-17 21:05:07 -06:00
#[test]
fn file_writer_bad_name() {
match io::file_writer(&Path("?/?"), ~[]) {
2012-09-28 15:00:07 -05:00
result::Err(copy e) => {
assert str::starts_with(e, "error opening");
2012-01-17 21:05:07 -06:00
}
2012-08-26 18:54:31 -05:00
result::Ok(_) => fail
2012-01-17 21:05:07 -06:00
}
}
#[test]
fn buffered_file_writer_bad_name() {
match io::buffered_file_writer(&Path("?/?")) {
2012-09-28 15:00:07 -05:00
result::Err(copy e) => {
assert str::starts_with(e, "error opening");
2012-01-17 21:05:07 -06:00
}
2012-08-26 18:54:31 -05:00
result::Ok(_) => fail
2012-01-17 21:05:07 -06:00
}
}
2012-05-11 12:44:57 -05:00
#[test]
fn bytes_buffer_overwrite() {
let wr = BytesWriter();
wr.write(~[0u8, 1u8, 2u8, 3u8]);
assert wr.bytes.borrow(|bytes| bytes == ~[0u8, 1u8, 2u8, 3u8]);
wr.seek(-2, SeekCur);
wr.write(~[4u8, 5u8, 6u8, 7u8]);
assert wr.bytes.borrow(|bytes| bytes ==
~[0u8, 1u8, 4u8, 5u8, 6u8, 7u8]);
wr.seek(-2, SeekEnd);
wr.write(~[8u8]);
wr.seek(1, SeekSet);
wr.write(~[9u8]);
assert wr.bytes.borrow(|bytes| bytes ==
~[0u8, 9u8, 4u8, 5u8, 8u8, 7u8]);
2012-05-11 12:44:57 -05:00
}
#[test]
fn test_read_write_le() {
let path = Path("tmp/lib-io-test-read-write-le.tmp");
let uints = [0, 1, 2, 42, 10_123, 100_123_456, u64::max_value];
// write the ints to the file
{
let file = io::file_writer(&path, [io::Create]).get();
for uints.each |i| {
file.write_le_u64(*i);
}
}
// then read them back and check that they are the same
{
let file = io::file_reader(&path).get();
for uints.each |i| {
assert file.read_le_u64() == *i;
}
}
}
#[test]
fn test_read_write_be() {
let path = Path("tmp/lib-io-test-read-write-be.tmp");
let uints = [0, 1, 2, 42, 10_123, 100_123_456, u64::max_value];
// write the ints to the file
{
let file = io::file_writer(&path, [io::Create]).get();
for uints.each |i| {
file.write_be_u64(*i);
}
}
// then read them back and check that they are the same
{
let file = io::file_reader(&path).get();
for uints.each |i| {
assert file.read_be_u64() == *i;
}
}
}
#[test]
fn test_read_be_int_n() {
let path = Path("tmp/lib-io-test-read-be-int-n.tmp");
let ints = [i32::min_value, -123456, -42, -5, 0, 1, i32::max_value];
// write the ints to the file
{
let file = io::file_writer(&path, [io::Create]).get();
for ints.each |i| {
file.write_be_i32(*i);
}
}
// then read them back and check that they are the same
{
let file = io::file_reader(&path).get();
for ints.each |i| {
// this tests that the sign extension is working
// (comparing the values as i32 would not test this)
assert file.read_be_int_n(4) == *i as i64;
}
}
}
2012-01-17 21:05:07 -06:00
}
//
// Local Variables:
// mode: rust
// fill-column: 78;
// indent-tabs-mode: nil
// c-basic-offset: 4
// buffer-file-coding-system: utf-8-unix
// End:
//