-
Notifications
You must be signed in to change notification settings - Fork 4
Doc reading and writing files
This document corresponds to https://github.com/Havvy/rust-docs/blob/master/doc/readwrite.md so if you make edits, please notify Havvy or send a pull request there.
Writing and reading files. Important in any language. Here's how to do so in Rust. This page will focus on the API and not on the language constructs.
As a warning, I'm not at all comfortable with Rust yet. There are parts of the syntax that completely confuse me, including the majority of pointers. I've never written a line of C++ and only once used pointers in C. Managing memory is the reason I even chose to learn this language.
And finally, I'm learning this all through experimentation and questions in IRC. Take the information in here with a grain of salt, and if there is an error, be bold and send a pull request or file an issue. As I become more sure of what I'm saying, I'll drop the pronouns.
To start out, we look at the core::io library(1). Just by looking at the docpage, we make note of the following functions:
- FILE_reader
- FILE_writer
- file_writer
- file_reader
- mk_file_writer
The FILE variants read C Style files while the file variants read Rust style files. I am not sure what the difference is between mk_file_writer and file_writer. mk_file_writer is just chain(file_writer).
I'm going to see if I can get away with using the Rust style files, and not deal with C style files. As such, the only two looked at will be file_reader and file_writer.
Both take a path of type ~str. They return a result<reader/writer, ~str>. So, to understand file io in Rust, you must also understand the result type.
The result type appears to be Rust's solution to runtime errors. A result is an enum of either ok(T) or err(U). Results must be unpacked before they are used, but for the programs on this page, I will first assert that they are ok.
The following three functions will be used from core::result(2).
- is_err<T, E>(result<T, E>) -> bool: True iff result<T, E> is err(E).
- get_err<T, E>(result<T, E>) -> E: Returns E in err(E) if result is err(e). Otherwise fails.
- unpack<T, E>(result<T, E>) -> T: Return T in ok(T) if result is ok(T). Otherwise, fails. Consumes the result<T, E>.
More functions, including methods for chaining and iterating can be found in the module. They will not be used here.
The code shown here are all tests, or helper functions for tests. As such, assertions will test for expected results of running the test. The tests are also written expecting that the program is at /home/havvy and that there is a file read.txt containing the contents "success" in the same directory.
The io and result functions, traits, and types are imported in the header.
To begin, let us read bytes from read.txt.
fn is_success (-path: str) {
// [1]
let maybe_test_reader: result<reader, str> = file_reader(path);
// [2]
if is_err::<reader, str>(maybe_test_reader) {
#warn(result::get_err(maybe_test_reader));
assert false;
}
let test_reader: reader = result::unwrap(maybe_test_reader);
// [3] [4]
let mut bytes: ~[u8] = ~[];
loop {
let byte: int = test_reader.read_byte();
#debug("%d", byte);
if test_reader.eof() { break }
vec::push(bytes, byte as u8);
}
// [5]
assert bytes == ~[115, 117, 99, 99, 101, 115, 115];
let maybe_success: str = str::from_bytes(bytes);
assert maybe_success == "success";
}
- The function tries to creates a file_reader using the path given to it.
- If the file_reader call failed, print why and then fail the test. Otherwise, unwrap the result.
- Since I do not yet see a way of querying the length of a file, I will just read the file one byte at a time, and store it in a vector. The bytes are read in as an int, but str::from_bytes expects a uint. So each byte is casted to a uint. This happens continously until 'end of file' is reached.
- The reader.eof() method returns true when the currently read byte is eof. If this is the case, then byte will be -1 (and thus the reason reader.read_byte returns an int instead of a u8) which should not be added to the vector. Since this is the EOF, the code breaks from the loop.
- At this point, the assertions that the file contains the correct contents is checked twice. Once in byte form, and then again as a string. The byte form is the ASCII values (and thus, UTF8 values) for 'success'.
You might be asking, why am I checking to exit the loop from inside the loop and not having a check between iterations.
This is due to the way reader.eof() works. Using a while loop checking for eof will cause the EOF byte of value -1 (255u8) to be appended to the vector. The final value has to be popped off when using a while loop.
let mut bytes: ~[u8] = ~[];
while !reader.eof() {
vec::push(bytes, reader.read_byte() as u8);
}
vec::pop(bytes);
This test will work when the program is ran from anywhere as long as test.txt exists in the specified location.
An absolute URL begins with the root directory, '/'.
#[test]
fn read_absolute_file () {
is_success("/home/havvy/read.txt");
}
A relative path is any path that does not begin with the root directory. The path starts from the directory the program is called from. For example, if the program is called from /home/havvy/, then a path of test.txt will read from /home/havvy/test.txt.
Directories named '.' and '..' has special meaning. '.' will go to the current directory, while '..' goes one directory up.
The paths test.txt and ./test.txt are equivelent. This test will pass or fail, depending on whether or not the file exists in the directory the program is ran from and whether or not the file contains the contents "success".
#[test]
fn read_relative_file () {
is_success("./read.txt");
is_success("read.txt");
}
Unpacking the reader, and in the next section, the writer follows the same steps each time. Since we don't want to see the same algorithm over and over again, the functions freader and fwriter are defined that return readers and writers unpacked already.
fn freader (-path: str) -> reader {
let maybe_reader = file_reader(path);
if is_err::<reader, str>(maybe_reader) {
#info("%s", get_err(maybe_reader));
assert false;
}
unwap(maybe_reader)
}
fn fwriter (-path: str) -> writer {
let maybe_writer = file_writer(path);
if is_err::<writer, str>(maybe_writer) {
#warn("%s", get_err(maybe_writer));
assert false;
}
unwap(maybe_writer)
}
In the writer section, there needs to be a guarantee that the file being written to is empty. An empty file will return EOF as the first byte, and as such, the is_empty function will return whether or not this is true.
fn is_empty (-path: str) -> bool {
freader(path).read_byte() == -1
}
Reading one byte at a time is silly. And the reader trait has another trait built off it containing a lot more methods. Here are some of these methods:
- fn read_line() -> ~str
- fn read_whole_stream() -> ~[u8]
- fn each_char(iterator: fn(char) -> bool)
- fn each_line(iterator: fn(~str) -> bool)
Note that reading one line gives a string while reading the whole stream gives a u8 vector.
You can get these methods with import io::reader_util;.
If you have not seen iterators in rust before, they do something with the contents given them and then return true to keep continuing through iteration or false to break the loop. It could be useful for a generator.
The other three can be used for reading read.txt and getting its contents. The test method below demonstrates this.
fn is_success2 (-path: str) {
let reader = freader(path);
assert reader.read_line() == "success";
assert reader.eof();
}
#[test]
fn utility_read_fns () {
is_success2("read.txt");
assert freader("read.txt").read_whole_stream() ==
~[115, 117, 99, 99, 101, 115, 115]
freader("read.txt").each_line(fn@ (line: str) -> bool {
assert line == "success"; true
});
}
The rest of this guide will be using is_success2 over is_success.
Writing files is slightly harder, since we need to undo any changes that are made afterwards. As such, the first test will write a few bytes, and the next test will then remove those bytes.
The file being worked on will be /home/havvy/write.txt. The tests assume that
the file already exists and is 0 bytes. Such a file can be made on a unix
system with the command touch /home/havvy/write.txt.
To make sure the file is empty, the following test case is ran.
#[test]
fn write_dot_txt_is_empty () {
assert is_empty("write.txt");
}