mzdata 0.63.4

#![allow(dead_code)]

use std::fs;
use std::io;
#[allow(unused)]
use std::io::prelude::*;
use std::path;
#[allow(unused)]
use std::path::PathBuf;

#[cfg(feature = "checksum")]
use md5::{Context as MD5Context, Digest};
#[cfg(feature = "checksum")]
use sha1::{self, Digest as _};

type ByteBuffer = io::Cursor<Vec<u8>>;

#[derive(Debug, Clone, Default)]
pub enum FileWrapper<T: io::Read> {
    FileSystem(path::PathBuf),
    Stream(T),
    #[default]
    Empty,
}

/// Controls the level of spectral detail read from an MS data file
#[derive(Debug, Default, Clone, Copy, Hash, PartialEq, Eq)]
pub enum DetailLevel {
    #[default]
    /// Read all spectral data, including peak data, eagerly decoding it. This is the default
    Full,
    /// Read all spectral data, including peak data but defer decoding until later if possible.
    /// Check a format reader's documentation to see if it supports lazy loading. Lazy loading
    /// is only really of value for dense profile mode data or very, very long peak lists
    /// that are large and expensive to decode.
    Lazy,
    /// Read only the metadata of spectra, ignoring peak data entirely
    MetadataOnly,
}

#[derive(Debug, Clone, Default)]
pub(crate) struct FileSource<T: io::Read> {
    pub source: FileWrapper<T>,
}

// This really should be a full file-like object abstraction, but that
// feels like it is beyond the scope of this crate. Something like
// https://github.com/bnjjj/chicon-rs
impl<T: io::Read> FileSource<T> {
    pub fn from_path<P>(path: P) -> FileSource<T>
    where
        P: Into<path::PathBuf>,
    {
        FileSource {
            source: FileWrapper::FileSystem(path.into()),
        }
    }

    pub fn from_stream(stream: T) -> FileSource<T> {
        FileSource {
            source: FileWrapper::Stream(stream),
        }
    }

    pub fn file_name(&self) -> Option<&path::Path> {
        match &self.source {
            FileWrapper::FileSystem(path) => Some(path),
            FileWrapper::Stream(_stream) => None,
            FileWrapper::Empty => None,
        }
    }

    pub fn index_file_name(&self) -> Option<path::PathBuf> {
        match &self.source {
            FileWrapper::Empty => None,
            FileWrapper::Stream(_stream) => None,
            FileWrapper::FileSystem(path) => {
                if let Some(stem) = path.file_name() {
                    if let Some(parent) = path.parent() {
                        let base = parent.join(stem);
                        let name = base.with_extension("index.json");
                        return Some(name);
                    }
                }
                None
            }
        }
    }

    pub fn has_index_file(&self) -> bool {
        match self.index_file_name() {
            Some(path) => path.exists(),
            None => false,
        }
    }
}

pub fn from_path<P>(path: P) -> FileSource<fs::File>
where
    P: Into<path::PathBuf>,
{
    FileSource::from_path(path)
}

impl<T, P> From<P> for FileSource<T>
where
    P: Into<path::PathBuf>,
    T: io::Read,
{
    fn from(path: P) -> FileSource<T> {
        FileSource::from_path(path)
    }
}

/// A wrapper around an [`io::Read`] to provide limited [`io::Seek`] access even if the
/// underlying stream does not support it. It pre-buffers the next *n* bytes of content
/// in memory and permits seek operations within that range, but fails all seeks beyond
/// that range.
///
/// This is useful for working with [`io::stdin`] or a network stream.
pub struct PreBufferedStream<R: io::Read> {
    stream: R,
    buffer: io::Cursor<Vec<u8>>,
    buffer_size: usize,
    position: usize,
}

impl<R: io::Read> io::Seek for PreBufferedStream<R> {
    fn seek(&mut self, pos: io::SeekFrom) -> io::Result<u64> {
        match pos {
            io::SeekFrom::Start(offset) => {
                if self.position > self.buffer_size {
                    return Err(io::Error::new(
                        io::ErrorKind::InvalidInput,
                        "Seeking after leaving buffered prefix",
                    ));
                } else if self.position + offset as usize > self.buffer_size {
                    return Err(io::Error::new(
                        io::ErrorKind::InvalidInput,
                        "Cannot seeking beyond buffered prefix",
                    ));
                }
                self.position = offset as usize;
                let r = self.buffer.seek(pos);
                if log::log_enabled!(log::Level::Trace) {
                    log::trace!(
                        "{pos:?} Position {0} -> {1}: {r:?}",
                        self.position,
                        self.buffer
                            .stream_position()
                            .map(|s| s.to_string())
                            .unwrap_or_else(|e| format!("err: {e}"))
                    );
                }
                r
            }
            io::SeekFrom::End(_) => Err(io::Error::new(
                io::ErrorKind::Unsupported,
                "Cannot seek relative the end of PreBufferedStream",
            )),
            io::SeekFrom::Current(offset) => {
                if self.position > self.buffer_size {
                    return Err(io::Error::new(
                        io::ErrorKind::InvalidInput,
                        "Seeking after leaving buffered prefix",
                    ));
                }
                if offset < 0 {
                    if offset.unsigned_abs() as usize > self.position {
                        Err(io::Error::new(
                            io::ErrorKind::InvalidInput,
                            "Cannot seek to negative position",
                        ))
                    } else {
                        self.position =
                            self.position.saturating_sub(offset.unsigned_abs() as usize);
                        let r = self.buffer.seek(io::SeekFrom::Start(self.position as u64));
                        if log::log_enabled!(log::Level::Trace) {
                            log::trace!(
                                "{pos:?} Position {0} -> {1}: {r:?}",
                                self.position,
                                self.buffer
                                    .stream_position()
                                    .map(|s| s.to_string())
                                    .unwrap_or_else(|e| format!("err: {e}"))
                            );
                        }
                        r
                    }
                } else if offset as usize + self.position > self.buffer_size {
                    Err(io::Error::new(
                        io::ErrorKind::InvalidInput,
                        "Cannot seeking beyond buffered prefix",
                    ))
                } else {
                    let r = self.buffer.seek(io::SeekFrom::Current(offset));
                    if log::log_enabled!(log::Level::Trace) {
                        log::trace!(
                            "{pos:?} Position {0} -> {1}: {r:?}",
                            self.position,
                            self.buffer
                                .stream_position()
                                .map(|s| s.to_string())
                                .unwrap_or_else(|e| format!("err: {e}"))
                        );
                    }
                    r
                }
            }
        }
    }

    fn stream_position(&mut self) -> io::Result<u64> {
        Ok(self.position as u64)
    }
}

impl<R: io::Read> io::Read for PreBufferedStream<R> {
    fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
        let n_total = buf.len();
        let before = self.position;
        let (n_remaining, n_from_buffer) = if self.position < self.buffer_size {
            let n_from_buffer = self.buffer.read(buf)?;
            self.position += n_from_buffer;
            (n_total.saturating_sub(n_from_buffer), n_from_buffer)
        } else {
            (n_total, 0)
        };
        if n_remaining > 0 {
            let n_rest = self.stream.read(&mut buf[n_from_buffer..])?;
            self.position += n_rest;
        }
        let total_read = self.position - before;
        Ok(total_read)
    }
}

const BUFFER_SIZE: usize = 2usize.pow(16);

impl<R: io::Read> PreBufferedStream<R> {
    /// Create a new pre-buffered stream wrapping `stream` with a buffer size of 2<sup>16</sup> bytes.
    ///
    /// This method fails if attempting to fill the buffer fails.
    pub fn new(stream: R) -> io::Result<Self> {
        Self::new_with_buffer_size(stream, BUFFER_SIZE)
    }

    /// Create a new pre-buffered stream wrapping `stream` with a buffer size of `buffer_size` bytes.
    ///
    /// This method fails if attempting to fill the buffer fails.
    pub fn new_with_buffer_size(stream: R, buffer_size: usize) -> io::Result<Self> {
        let buffer = io::Cursor::new(Vec::with_capacity(buffer_size));
        let mut inst = Self {
            stream,
            buffer_size,
            buffer,
            position: 0,
        };
        inst.prefill_buffer()?;
        Ok(inst)
    }

    fn prefill_buffer(&mut self) -> io::Result<usize> {
        let buffer = self.buffer.get_mut();
        buffer.resize(self.buffer_size, 0);
        let bytes_read = self.stream.read(buffer)?;
        buffer.shrink_to(bytes_read);
        self.buffer_size = bytes_read;
        Ok(bytes_read)
    }
}

#[cfg(feature = "checksum")]
/// Compute a SHA-1 digest of a file path
pub fn checksum_file(path: &PathBuf) -> io::Result<String> {
    let mut checksum = sha1::Sha1::new();
    let mut reader = io::BufReader::new(fs::File::open(path)?);
    let mut buf = vec![0; 2usize.pow(20)];
    while let Ok(i) = reader.read(&mut buf) {
        if i == 0 {
            break;
        }
        checksum.update(&buf[..i]);
    }
    let x = base16ct::lower::encode_string(&checksum.finalize());
    Ok(x)
}

#[cfg(feature = "checksum")]
/// A writable stream that keeps a running MD5 checksum of all bytes
#[derive(Clone)]
pub(crate) struct MD5HashingStream<T: io::Write> {
    pub stream: T,
    pub context: MD5Context,
}

#[cfg(feature = "checksum")]
impl<T: io::Write> MD5HashingStream<T> {
    pub fn new(file: T) -> MD5HashingStream<T> {
        Self {
            stream: file,
            context: MD5Context::new(),
        }
    }

    pub fn compute(&self) -> Digest {
        self.context.clone().compute()
    }

    pub fn get_mut(&mut self) -> &mut T {
        &mut self.stream
    }

    pub fn into_inner(self) -> T {
        self.stream
    }
}

#[cfg(feature = "checksum")]
impl<T: io::Write> io::Write for MD5HashingStream<T> {
    fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
        self.context.consume(buf);
        self.stream.write(buf)
    }

    fn flush(&mut self) -> io::Result<()> {
        self.stream.flush()
    }
}

#[cfg(feature = "checksum")]
impl<T: io::Seek + io::Write> io::Seek for MD5HashingStream<T> {
    fn seek(&mut self, pos: io::SeekFrom) -> io::Result<u64> {
        self.stream.seek(pos)
    }
}


#[cfg(feature = "parallelism")]
mod parallelism {
    use rayon::prelude::*;
    use crate::prelude::*;

    use super::*;

    /// A helper type to load spectra concurrently across multiple threads.
    /// Requires the reader type implement [`MZFileReader`].
    ///
    /// # Note
    /// This helper is still too low level. Expect a higher level API to eventually become available.
    pub struct ConcurrentLoader {
        path: PathBuf,
        num_threads: Option<usize>,
    }

    impl ConcurrentLoader {
        pub fn new(path: PathBuf, num_threads: Option<usize>) -> Self {
            Self { path, num_threads }
        }

        /// Do the actual concurrent loading
        pub fn load<F: MZFileReader<C, D, S>, C: CentroidLike, D: DeconvolutedCentroidLike, S: SpectrumLike<C, D> + Send>(self) -> io::Result<Vec<S>> {
            let guide = F::open_path(&self.path)?;
            let n = guide.len();

            let num_threads = self.num_threads.unwrap_or_else(|| rayon::max_num_threads());

            let task = || -> Vec<S>{
                let mut chunks: Vec<_> = (0..n).into_par_iter().chunks((n / num_threads / 3).max(10)).map(|ii| {
                    let start = ii[0];
                    let mut local_reader = F::open_path(&self.path).unwrap();
                    let spectra: Vec<_> = ii.into_iter().flat_map(|i| local_reader.get_spectrum_by_index(i)).collect();
                    (start ,spectra)
                }).collect();
                chunks.par_sort_by(|a, b| a.0.cmp(&b.0));
                chunks.into_iter().map(|(_, chunk)| chunk).flatten().collect()
            };

            let out = if let Some(num_threads) = self.num_threads {
                let pool = rayon::ThreadPoolBuilder::new().num_threads(num_threads).thread_name(|i| format!("mzdata-concurrent-loader-{i}")).build().unwrap();
                pool.install(|| task())
            } else {
                task()
            };

            Ok(out)
        }
    }
}


#[cfg(feature = "parallelism")]
pub use parallelism::ConcurrentLoader;


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

    #[test]
    fn test_from_buffer() {
        let mut buff: Vec<u8> = Vec::new();
        buff.extend(b"foobar");
        let stream = ByteBuffer::new(buff);
        let mut out: Vec<u8> = Vec::new();
        let desc = FileSource::<ByteBuffer>::from_stream(stream);
        assert!(desc.file_name().is_none());
        if let FileWrapper::Stream(mut buff) = desc.source {
            buff.read_to_end(&mut out).unwrap();
            assert_eq!(out, b"foobar");
        }
    }

    #[test]
    fn test_prebuffering() -> io::Result<()> {
        let mut fh = fs::File::open("./test/data/batching_test.mzML")?;
        let mut data = Vec::new();
        fh.read_to_end(&mut data)?;
        let content = io::Cursor::new(data);
        let mut stream = PreBufferedStream::new_with_buffer_size(content, 512)?;

        assert_eq!(stream.buffer_size, 512);

        let mut buffer = [0u8; 128];
        stream.read_exact(&mut buffer)?;
        assert_eq!(buffer.len(), 128);
        assert!(buffer.starts_with(b"<?xml version=\"1.0\" encoding=\"utf-8\"?>"));

        let mut buffer2 = [0u8; 128];
        stream.seek(io::SeekFrom::Start(0))?;
        stream.read_exact(&mut buffer2)?;

        assert_eq!(buffer, buffer2);

        assert!(stream.seek(io::SeekFrom::Start(556)).is_err());

        Ok(())
    }

    #[cfg(feature = "parallelism")]
    #[test]
    fn test_parallel_load() -> io::Result<()> {
        use crate::prelude::*;

        let loader= ConcurrentLoader::new("./test/data/batching_test.mzML".into(), Some(4));
        let spectra = loader.load::<crate::MzMLReader<fs::File>, _, _, _>()?;
        assert_eq!(spectra.len(), 2232);

        let _ = spectra.iter().fold(0, |last, spec| {
            assert_eq!(last, spec.index());
            spec.index() + 1
        });

        Ok(())
    }
}