//! # ReadFilter: Wrappers for `Read`able things.
//!
//! A collection of structs you can wrap around something `Read`able in order to transparently
//! filter out unwanted content.
use std::io::{self, Read};

const BUF_SIZE: usize = 8 * 1024;
//const BUF_SIZE: usize = 16;

/// The common bits that each stream iterator/reader has, and the only bits
/// needed to impl Read on each of them.
struct Common<T>
where
    T: Read,
{
    /// The lower layer. The place where bytes are read from before processing.
    source: T,
    /// Stores and accumulates bytes read from source.
    working_buf: [u8; BUF_SIZE],
    /// The byte right after the last byte in the `working_buf`. The next byte that should be written
    /// into, and the byte right after the last valid byte that can be read.  When this is 0,
    /// `working_buf` is empty. Sometimes when we are read from, we might be told to fill a buffer
    /// that is smaller than the amount of bytes we have read from the lower layer already. If this
    /// ends up being the case, `.read()` stores a non-zero value here so `.next()` can be signaled
    /// to not overwrite data at the very beginning of `working_buffer`.
    unconsumed_bytes: usize,
}

impl<T> Common<T>
where
    T: Read,
{
    fn new(source: T) -> Self {
        Self {
            source,
            working_buf: [0; BUF_SIZE],
            unconsumed_bytes: 0,
        }
    }
}

/// Removes all non-whitelisted characters from the wrapped stream.
///
/// Non-utf8 characters are lost due to internal string conversions using
/// [String::from_utf8_lossy()][String::from_utf8_lossy].
///
/// ```
/// use readfilter::CharWhitelist;
/// use std::io::Read;
/// let buf = "aabbccddee".as_bytes();
/// let mut wrapped = CharWhitelist::new(buf, "ace");
/// let mut s = String::new();
/// wrapped.read_to_string(&mut s).unwrap();
/// assert_eq!(&s, "aaccee");
/// ```
///
/// ```no_run
/// use readfilter::CharWhitelist;
/// use std::fs::OpenOptions;
/// use std::io::Read;
/// let mut output = vec![];
/// let fd = OpenOptions::new().read(true).open("input.txt").unwrap();
/// let mut input = CharWhitelist::new(fd, "01");
/// input.read_to_end(&mut output);
/// // output only contains the '0' and '1' characters from input.txt
/// ```
pub struct CharWhitelist<T>
where
    T: Read,
{
    common: Common<T>,
    allowed_chars: Vec<char>,
}

impl<T> CharWhitelist<T>
where
    T: Read,
{
    pub fn new(source: T, allowed_chars: &str) -> Self {
        Self {
            common: Common::new(source),
            allowed_chars: allowed_chars.chars().collect(),
        }
    }

    fn next(&mut self) -> Option<io::Result<usize>> {
        let mut read_previously = self.common.unconsumed_bytes;
        // keep looping until we get an error, fail to read any bytes, or have
        // read a full buffer
        loop {
            let read_this_time = self
                .common
                .source
                .read(&mut self.common.working_buf[read_previously..]);
            // return the error if there is one
            if let Err(e) = read_this_time {
                return Some(Err(e));
            }
            let read_this_time = read_this_time.unwrap();
            // if we didn't read anything, time to stop looping
            if read_this_time < 1 {
                return Some(Ok(read_previously));
            }
            assert!(read_this_time > 0);
            // convert [u8, BUF_SIZE] (with length read_this_time) to String
            let buf = String::from_utf8_lossy(
                &self.common.working_buf[read_previously..read_previously + read_this_time],
            )
            .into_owned();
            for c in buf.chars() {
                if self.allowed_chars.contains(&c) {
                    let _ = c.encode_utf8(&mut self.common.working_buf[read_previously..]);
                    read_previously += c.len_utf8();
                }
            }
        }
    }
}

/// Removes comments from the wrapped stream.
///
/// A comment starts with `#` and ends with a newline `\n` or the end of file.
///
/// ```
/// use readfilter::CommentStrip;
/// use std::io::Read;
/// let mut input = CommentStrip::new("a\nb# foo\nc#bar".as_bytes());
/// let mut output = String::new();
/// input.read_to_string(&mut output);
/// assert_eq!(output, "a\nbc");
/// ```
pub struct CommentStrip<T>
where
    T: Read,
{
    common: Common<T>,
    /// Sometimes reads from source will end with a comment that isn't finished
    /// yet. This flag is used to keep track of whether or not we need to keep
    /// ignoring bytes until the comment ends (i.e. we see a newline)
    ignore_until_next_newline: bool,
}

impl<T> CommentStrip<T>
where
    T: Read,
{
    pub fn new(source: T) -> Self {
        Self {
            common: Common::new(source),
            ignore_until_next_newline: false,
        }
    }

    fn next(&mut self) -> Option<io::Result<usize>> {
        let mut read_previously = self.common.unconsumed_bytes;
        // keep looping until we get an error, fail to read any bytes, or have
        // read a full buffer
        loop {
            let read_this_time = self
                .common
                .source
                .read(&mut self.common.working_buf[read_previously..]);
            // return the error if there is one
            if let Err(e) = read_this_time {
                return Some(Err(e));
            }
            let read_this_time = read_this_time.unwrap();
            // if we didn't read anything, time to stop looping
            if read_this_time < 1 {
                return Some(Ok(read_previously));
            }
            assert!(read_this_time > 0);
            // convert [u8, BUF_SIZE] (with length read_this_time) to String
            let mut buf = String::from_utf8_lossy(
                &self.common.working_buf[read_previously..read_previously + read_this_time],
            )
            .into_owned();
            // if needed, ignore bytes up through a newline
            buf = if self.ignore_until_next_newline {
                match buf.find('\n') {
                    // found a newline. ignore bytes up through it, keep bytes after it, and unset
                    // ignore_until_next_newline
                    Some(idx) => {
                        self.ignore_until_next_newline = false;
                        buf[idx + 1..].to_string()
                    }
                    // didn't find newline. ignore all bytes
                    None => String::new(),
                }
            } else {
                buf
            };
            // if buffer empty, end early
            if buf.is_empty() {
                return Some(Ok(read_previously));
            }
            // loop until no more comments in buf
            loop {
                // look for comment character. if found, ignore bytes after it until newline
                let start_idx = buf.find('#');
                if start_idx.is_none() {
                    break;
                }
                let start_idx = start_idx.unwrap();
                buf = match buf[start_idx..].find('\n') {
                    // newline found. ignore bytes between comment char and newline
                    Some(end_idx) => {
                        String::from(&buf[..start_idx]) + &buf[start_idx + end_idx + 1..]
                    }
                    // no newline found. ignore all bytes after comment char and set flag to keep
                    // ignoring on next loop
                    None => {
                        self.ignore_until_next_newline = true;
                        buf[..start_idx].to_string()
                    }
                };
            }
            let remaining_len = buf.len();
            if remaining_len > 0 {
                self.common.working_buf[read_previously..read_previously + remaining_len]
                    .copy_from_slice(buf.as_bytes());
                read_previously += remaining_len;
            }
        }
    }
}

/// Each type impls Read. As all the hard work is done in self.next(), this can
/// be generalized. .read() is probably how the user should be using these types.
macro_rules! impl_read_trait_for_stream_iter {
    ($MyType:ty) => {
        impl<T> Read for $MyType
        where
            T: Read,
        {
            fn read(&mut self, out_buf: &mut [u8]) -> io::Result<usize> {
                let mut bytes_given = 0;
                if self.common.unconsumed_bytes >= out_buf.len() {
                    // We have more data already buffered than the user wants to read.
                    // 1. Copy to them the max amount of data
                    // 2. Update our buffer so that it starts with the buffered bytes
                    // right after the ones we just gave them
                    // 3. Update the length of our buffer
                    // Then we're done. We shouldn't read anything more because we can't
                    // even give it to them yet.
                    let out_buf_len = out_buf.len();
                    out_buf[..out_buf_len].copy_from_slice(&self.common.working_buf[..out_buf_len]);
                    self.common
                        .working_buf
                        .copy_within(out_buf_len..self.common.unconsumed_bytes, 0);
                    self.common.unconsumed_bytes -= out_buf_len;
                    return Ok(out_buf.len());
                } else {
                    // We have less data already buffered than what the user wants to read.
                    // 1. Copy to them all that we have.
                    // 2. Update the length of our buffer to be zero.
                    // 3. Note that we've given them some bytes.
                    // Continue with this function. We might be able to give them more.
                    out_buf[..self.common.unconsumed_bytes]
                        .copy_from_slice(&self.common.working_buf[..self.common.unconsumed_bytes]);
                    bytes_given += self.common.unconsumed_bytes;
                    self.common.unconsumed_bytes = 0;
                }
                assert_eq!(self.common.unconsumed_bytes, 0);
                // If we're here, then we must need to read some more bytes and give
                // them to the out_buf. First try to read more.
                let next_res = self.next();
                if let Some(Ok(working_buf_len)) = next_res {
                    // We successfully read something. working_buf_len is the number of bytes at
                    // the front of the working buf that are valid. We have access to the working
                    // buf since it is ours.
                    let max_bytes_to_give = out_buf.len() - bytes_given;
                    if working_buf_len >= max_bytes_to_give {
                        // If we read too many bytes, then
                        // 1. Give as many as possible to the out_buf
                        // 2. Move the remaining working_buf bytes to the front of the working_buf
                        // 3. Update the len of the working_buf And then we're done and can return.
                        out_buf[bytes_given..]
                            .copy_from_slice(&self.common.working_buf[..max_bytes_to_give]);
                        self.common
                            .working_buf
                            .copy_within(max_bytes_to_give..working_buf_len, 0);
                        self.common.unconsumed_bytes = working_buf_len - max_bytes_to_give;
                        bytes_given += max_bytes_to_give;
                        return Ok(bytes_given);
                    } else {
                        // We read fewer bytes than there is remaining space in out_buf. We can
                        // give it all the bytes. For simplicity, just return after doing so. We
                        // could loop around and do all this again.
                        out_buf[bytes_given..bytes_given + working_buf_len]
                            .copy_from_slice(&self.common.working_buf[..working_buf_len]);
                        bytes_given += working_buf_len;
                        return Ok(bytes_given);
                    }
                } else if let Some(Err(e)) = next_res {
                    return Err(e);
                } else {
                    // We failed to iterate forward in the stream. Must be done.
                    return Ok(bytes_given);
                }
            }
        }
    };
}

impl_read_trait_for_stream_iter!(CharWhitelist<T>);
impl_read_trait_for_stream_iter!(CommentStrip<T>);

#[cfg(test)]
fn read_to_string(mut buf: impl Read) -> String {
    let mut s = String::new();
    buf.read_to_string(&mut s).unwrap();
    s
}

#[cfg(test)]
mod test_comment_strip_iter {
    use super::{read_to_string, CommentStrip, BUF_SIZE};

    #[test]
    fn empty_is_empty() {
        let s = "".as_bytes();
        assert_eq!(read_to_string(CommentStrip::new(s)).len(), 0);
    }

    #[test]
    fn ignore_all_short() {
        for s in &["#foo baz", "#foo baz\n", "#    ", "#    \n", "#", "#\n"] {
            assert_eq!(read_to_string(CommentStrip::new(s.as_bytes())).len(), 0);
        }
    }

    #[test]
    fn ignore_all_long_1() {
        // just less than a full buffer
        let mut s = vec!['#' as u8];
        s.append(&mut vec![' ' as u8; BUF_SIZE - 2]);
        assert_eq!(s.len(), BUF_SIZE - 1);
        assert_eq!(read_to_string(CommentStrip::new(&s[..])).len(), 0);

        // exactly a full buffer
        let mut s = vec!['#' as u8];
        s.append(&mut vec![' ' as u8; BUF_SIZE - 1]);
        assert_eq!(s.len(), BUF_SIZE);
        assert_eq!(read_to_string(CommentStrip::new(&s[..])).len(), 0);

        // just over a full buffer
        let mut s = vec!['#' as u8];
        s.append(&mut vec![' ' as u8; BUF_SIZE]);
        assert_eq!(s.len(), BUF_SIZE + 1);
        assert_eq!(read_to_string(CommentStrip::new(&s[..])).len(), 0);

        // over 2 buffers in size
        let mut s = vec!['#' as u8];
        s.append(&mut vec![' ' as u8; BUF_SIZE * 2 + 2]);
        assert_eq!(s.len(), BUF_SIZE * 2 + 3);
        assert_eq!(read_to_string(CommentStrip::new(&s[..])).len(), 0);
    }

    #[test]
    fn ignore_all_long_2() {
        // just less than a full buffer
        let mut s = vec!['#' as u8];
        s.append(&mut vec![' ' as u8; BUF_SIZE - 3]);
        s.push('\n' as u8);
        assert_eq!(s.len(), BUF_SIZE - 1);
        assert_eq!(read_to_string(CommentStrip::new(&s[..])).len(), 0);

        // exactly a full buffer
        let mut s = vec!['#' as u8];
        s.append(&mut vec![' ' as u8; BUF_SIZE - 2]);
        s.push('\n' as u8);
        assert_eq!(s.len(), BUF_SIZE);
        assert_eq!(read_to_string(CommentStrip::new(&s[..])).len(), 0);

        // just over a full buffer
        let mut s = vec!['#' as u8];
        s.append(&mut vec![' ' as u8; BUF_SIZE - 1]);
        s.push('\n' as u8);
        assert_eq!(s.len(), BUF_SIZE + 1);
        assert_eq!(read_to_string(CommentStrip::new(&s[..])).len(), 0);

        // over 2 buffers in size
        let mut s = vec!['#' as u8];
        s.append(&mut vec![' ' as u8; BUF_SIZE * 2 + 1]);
        s.push('\n' as u8);
        assert_eq!(s.len(), BUF_SIZE * 2 + 3);
        assert_eq!(read_to_string(CommentStrip::new(&s[..])).len(), 0);
    }

    #[test]
    fn keep_end_short() {
        for s in &["#\nfoo", "#  \nfoo"] {
            let out = read_to_string(CommentStrip::new(s.as_bytes()));
            assert_eq!(out, "foo");
        }

        for s in &["#\nfoo  foo", "#  \nfoo  foo"] {
            let out = read_to_string(CommentStrip::new(s.as_bytes()));
            assert_eq!(out, "foo  foo");
        }

        for s in &["#\nfoo \n foo", "#  \nfoo \n foo"] {
            let out = read_to_string(CommentStrip::new(s.as_bytes()));
            assert_eq!(out, "foo \n foo");
        }
    }

    #[test]
    fn keep_end_long() {
        let content = " foo \n foo ";

        // just under BUF_SIZE
        let mut s = vec!['#' as u8; BUF_SIZE - content.len() - 1 - 1];
        s.push('\n' as u8);
        for c in content.chars() {
            s.push(c as u8);
        }
        assert_eq!(s.len(), BUF_SIZE - 1);
        assert_eq!(read_to_string(CommentStrip::new(&s[..])), content);

        // equal to BUF_SIZE
        let mut s = vec!['#' as u8; BUF_SIZE - content.len() - 1];
        s.push('\n' as u8);
        for c in content.chars() {
            s.push(c as u8);
        }
        assert_eq!(s.len(), BUF_SIZE);
        assert_eq!(read_to_string(CommentStrip::new(&s[..])), content);

        // just over BUF_SIZE
        let mut s = vec!['#' as u8; BUF_SIZE - content.len() - 1 + 1];
        s.push('\n' as u8);
        for c in content.chars() {
            s.push(c as u8);
        }
        assert_eq!(s.len(), BUF_SIZE + 1);
        assert_eq!(read_to_string(CommentStrip::new(&s[..])), content);

        // comment is over BUF_SIZE by itself
        let mut s = vec!['#' as u8; BUF_SIZE + 1];
        s.push('\n' as u8);
        for c in content.chars() {
            s.push(c as u8);
        }
        assert_eq!(s.len(), BUF_SIZE + 2 + content.len());
        assert_eq!(read_to_string(CommentStrip::new(&s[..])), content);
    }
}

#[cfg(test)]
mod test_char_whitelist_iter {
    use super::{read_to_string, CharWhitelist};

    #[test]
    fn empty_whitelist() {
        let in_buf = "A\u{00a1}\u{01d6a9}".as_bytes();
        assert_eq!(read_to_string(CharWhitelist::new(in_buf, "")).len(), 0);
    }

    #[test]
    fn whitelist_allows_all() {
        let s = "A\u{00a1}\u{1d6a9}";
        assert_eq!(read_to_string(CharWhitelist::new(s.as_bytes(), s)), s);
    }

    #[test]
    fn whitelist_allows_single() {
        for allowed in vec!["A", "\u{00a1}", "\u{1d6a9}"] {
            let in_buf = "A\u{00a1}\u{1d6a9}".as_bytes();
            assert_eq!(read_to_string(CharWhitelist::new(in_buf, allowed)), allowed);
        }
    }
}

#[cfg(test)]
mod test_comment_strip_iter_read {
    use super::{CommentStrip, BUF_SIZE};
    use std::io::Read;

    #[test]
    fn just_comment_returns_empty() {
        let in_buf = "# foo \n".as_bytes();
        let mut out_buf = [0; 1];
        let mut csi = CommentStrip::new(&in_buf[..]);
        let len = csi.read(&mut out_buf).unwrap();
        assert_eq!(len, 0);
    }

    #[test]
    fn just_byte_before_comment() {
        let in_buf = "a# foo \n".as_bytes();
        let mut out_buf = [0; BUF_SIZE];
        let mut csi = CommentStrip::new(&in_buf[..]);
        let len = csi.read(&mut out_buf).unwrap();
        assert_eq!(String::from_utf8_lossy(&out_buf[..len]), "a");
    }

    #[test]
    fn just_byte_after_comment() {
        let in_buf = "# foo \na".as_bytes();
        let mut out_buf = [0; BUF_SIZE];
        let mut csi = CommentStrip::new(&in_buf[..]);
        let len = csi.read(&mut out_buf).unwrap();
        assert_eq!(String::from_utf8_lossy(&out_buf[..len]), "a");
    }

    #[test]
    fn just_byte_before_and_after_comment() {
        let in_buf = "a# foo \nB".as_bytes();
        let mut out_buf = [0; BUF_SIZE];
        let mut csi = CommentStrip::new(&in_buf[..]);
        let len = csi.read(&mut out_buf).unwrap();
        assert_eq!(String::from_utf8_lossy(&out_buf[..len]), "aB");
    }
}

macro_rules! impl_tests_for_common_read {
    ($mod_name:ident, $MyType:ident) => {
        #[cfg(test)]
        mod $mod_name {
            use super::{$MyType, BUF_SIZE};
            use std::io::Read;

            #[test]
            fn empty() {
                let in_buf = vec![];
                let mut out_buf = [0; 1];
                let mut csi = $MyType::new(&in_buf[..]);
                let len = csi.read(&mut out_buf).unwrap();
                assert_eq!(len, 0);
            }

            #[test]
            fn many_tiny_reads() {
                let in_buf = "abc123".as_bytes();
                let mut out_buf = [0; 1];
                let mut acc = String::new();
                let mut csi = $MyType::new(&in_buf[..]);
                for _ in 0..in_buf.len() {
                    let len = csi.read(&mut out_buf).unwrap();
                    acc += &String::from_utf8_lossy(&out_buf[..len]);
                }
                assert_eq!(acc, "abc123");
                // future reads should be length zero
                assert_eq!(csi.read(&mut out_buf).unwrap(), 0);
            }

            #[test]
            fn big_inbuf_tiny_outbuf() {
                let mut in_buf = vec!['a' as u8; BUF_SIZE / 2];
                in_buf.append(&mut vec!['b' as u8; BUF_SIZE / 2]);
                in_buf.append(&mut vec!['c' as u8; BUF_SIZE / 2]);
                in_buf.append(&mut vec!['d' as u8; BUF_SIZE / 2]);
                in_buf.append(&mut vec!['e' as u8; BUF_SIZE / 2]);
                let mut out_buf = [0; 2];
                let mut acc = String::new();
                let mut csi = $MyType::new(&in_buf[..]);
                loop {
                    let len = csi.read(&mut out_buf).unwrap();
                    acc += &String::from_utf8_lossy(&out_buf[..len]);
                    if len < 1 {
                        break;
                    }
                }
                assert_eq!(acc, String::from_utf8_lossy(&in_buf[..]));
            }

            #[test]
            fn big_inbuf_just_smaller_outbuf() {
                let mut in_buf = vec!['a' as u8; BUF_SIZE / 2];
                in_buf.append(&mut vec!['b' as u8; BUF_SIZE / 2]);
                in_buf.append(&mut vec!['c' as u8; BUF_SIZE / 2]);
                in_buf.append(&mut vec!['d' as u8; BUF_SIZE / 2]);
                assert_eq!(in_buf.len(), BUF_SIZE * 2);
                let mut out_buf = [0; BUF_SIZE * 2 - 1];
                let mut acc = String::new();
                let mut csi = $MyType::new(&in_buf[..]);
                loop {
                    let len = csi.read(&mut out_buf).unwrap();
                    acc += &String::from_utf8_lossy(&out_buf[..len]);
                    if len < 1 {
                        break;
                    }
                }
                assert_eq!(acc, String::from_utf8_lossy(&in_buf[..]));
            }

            #[test]
            fn big_inbuf_just_larger_outbuf() {
                let mut in_buf = vec!['a' as u8; BUF_SIZE / 2];
                in_buf.append(&mut vec!['b' as u8; BUF_SIZE / 2]);
                in_buf.append(&mut vec!['c' as u8; BUF_SIZE / 2]);
                in_buf.append(&mut vec!['d' as u8; BUF_SIZE / 2]);
                assert_eq!(in_buf.len(), BUF_SIZE * 2);
                let mut out_buf = [0; BUF_SIZE * 2 + 1];
                let mut acc = String::new();
                let mut csi = $MyType::new(&in_buf[..]);
                loop {
                    let len = csi.read(&mut out_buf).unwrap();
                    acc += &String::from_utf8_lossy(&out_buf[..len]);
                    if len < 1 {
                        break;
                    }
                }
                assert_eq!(acc, String::from_utf8_lossy(&in_buf[..]));
            }
        }
    };
}

impl_tests_for_common_read!(test_read_common_with_comment_strip_iter, CommentStrip);
// can't test because its .new() requires more than just the source
//impl_tests_for_common_read!(test_read_common_with_char_whitelist_iter, CharWhitelist);