craft-libs/crates/zip/src/write.rs

//! A ZIP archive being written.

use std::io::{Cursor, Write};

use crate::Error;

/// How an entry is stored.
#[derive(Clone, Copy, Debug, PartialEq, Eq)]
pub enum Compression {
    /// Uncompressed: a `mimetype` entry that must be readable in place, or data that is
    /// compressed already.
    Stored,
    /// Deflate at level 6.
    Deflated,
    /// Deflate at the given level (0 to 9; higher is clamped).
    Level(u8),
}

/// A ZIP archive being written into memory. The output is deterministic: the same (1980)
/// timestamp on every entry. Entries are written in the order they are added.
pub struct Writer {
    zip: zip::ZipWriter<Cursor<Vec<u8>>>,
    legacy_utf8: bool,
}

impl Default for Writer {
    fn default() -> Writer {
        Writer { zip: zip::ZipWriter::new(Cursor::new(Vec::new())), legacy_utf8: false }
    }
}

impl Writer {
    pub fn new() -> Writer {
        Writer::default()
    }

    /// Writes the UTF-8 flag and version fields used by the previous dotLottie writer.
    ///
    /// This keeps stored archives byte-for-byte stable for existing users while still using the
    /// shared writer for entry encoding and validation.
    pub fn legacy_utf8() -> Writer {
        Writer { legacy_utf8: true, ..Writer::default() }
    }

    /// Adds an entry from a byte slice. A name the archive already has is an error.
    pub fn add(&mut self, name: &str, data: &[u8], compression: Compression) -> Result<(), Error> {
        let (method, level) = match compression {
            Compression::Stored => (zip::CompressionMethod::Stored, None),
            Compression::Deflated => (zip::CompressionMethod::Deflated, Some(6)),
            Compression::Level(l) => (zip::CompressionMethod::Deflated, Some(i64::from(l.min(9)))),
        };
        let options = zip::write::SimpleFileOptions::default()
            .compression_method(method)
            .compression_level(level)
            .last_modified_time(zip::DateTime::default())
            .large_file(data.len() as u64 >= u64::from(u32::MAX) / 2);
        let fail = |e: &dyn std::fmt::Display| Error::Write(format!("{name}: {e}"));
        self.zip.start_file(name, options).map_err(|e| fail(&e))?;
        self.zip.write_all(data).map_err(|e| fail(&e))
    }

    /// Adds an entry by streaming it into the compressor, so large generated entries need not
    /// first be buffered in memory. A name the archive already has is an error.
    pub fn add_with(&mut self, name: &str, compression: Compression, write: impl FnOnce(&mut dyn Write) -> std::io::Result<()>) -> Result<(), Error> {
        let (method, level) = match compression {
            Compression::Stored => (zip::CompressionMethod::Stored, None),
            Compression::Deflated => (zip::CompressionMethod::Deflated, Some(6)),
            Compression::Level(l) => (zip::CompressionMethod::Deflated, Some(i64::from(l.min(9)))),
        };
        let options = zip::write::SimpleFileOptions::default()
            .compression_method(method)
            .compression_level(level)
            .last_modified_time(zip::DateTime::default())
            .large_file(true);
        let fail = |e: &dyn std::fmt::Display| Error::Write(format!("{name}: {e}"));
        self.zip.start_file(name, options).map_err(|e| fail(&e))?;
        write(&mut self.zip).map_err(|e| fail(&e))
    }

    /// The finished archive.
    pub fn finish(self) -> Result<Vec<u8>, Error> {
        let mut bytes = self.zip.finish().map(Cursor::into_inner).map_err(|e| Error::Write(e.to_string()))?;
        if self.legacy_utf8 {
            patch_legacy_utf8_headers(&mut bytes)?;
        }
        Ok(bytes)
    }
}

fn get_u16(bytes: &[u8], at: usize) -> Option<u16> {
    Some(u16::from_le_bytes(bytes.get(at..at.checked_add(2)?)?.try_into().ok()?))
}

fn get_u32(bytes: &[u8], at: usize) -> Option<u32> {
    Some(u32::from_le_bytes(bytes.get(at..at.checked_add(4)?)?.try_into().ok()?))
}

fn put_u32(bytes: &mut [u8], at: usize, value: u32) -> Result<(), Error> {
    let dst = bytes
        .get_mut(at..at.checked_add(4).ok_or_else(|| Error::Write("ZIP header offset overflow".into()))?)
        .ok_or_else(|| Error::Write("short ZIP header".into()))?;
    dst.copy_from_slice(&value.to_le_bytes());
    Ok(())
}

fn put_u16(bytes: &mut [u8], at: usize, value: u16) -> Result<(), Error> {
    let dst = bytes
        .get_mut(at..at.checked_add(2).ok_or_else(|| Error::Write("ZIP header offset overflow".into()))?)
        .ok_or_else(|| Error::Write("short ZIP header".into()))?;
    dst.copy_from_slice(&value.to_le_bytes());
    Ok(())
}

/// Patches the two header versions and UTF-8 bit to match the former dotLottie writer.
fn patch_legacy_utf8_headers(bytes: &mut [u8]) -> Result<(), Error> {
    const EOCD: u32 = 0x0605_4b50;
    const CENTRAL: u32 = 0x0201_4b50;
    const LOCAL: u32 = 0x0403_4b50;
    let eocd = bytes
        .windows(4)
        .rposition(|window| u32::from_le_bytes([window[0], window[1], window[2], window[3]]) == EOCD)
        .ok_or_else(|| Error::Write("missing ZIP end record".into()))?;
    let central_start = usize::try_from(get_u32(bytes, eocd + 16).ok_or_else(|| Error::Write("short ZIP end record".into()))?)
        .map_err(|_| Error::Write("ZIP central directory offset overflow".into()))?;
    let central_size = usize::try_from(get_u32(bytes, eocd + 12).ok_or_else(|| Error::Write("short ZIP end record".into()))?)
        .map_err(|_| Error::Write("ZIP central directory size overflow".into()))?;
    let central_end =
        central_start.checked_add(central_size).filter(|end| *end <= eocd).ok_or_else(|| Error::Write("invalid ZIP central directory".into()))?;
    let mut at = central_start;
    while at < central_end {
        if get_u32(bytes, at) != Some(CENTRAL) {
            return Err(Error::Write("invalid ZIP central directory entry".into()));
        }
        let name = usize::from(get_u16(bytes, at + 28).ok_or_else(|| Error::Write("short ZIP entry".into()))?);
        let extra = usize::from(get_u16(bytes, at + 30).ok_or_else(|| Error::Write("short ZIP entry".into()))?);
        let comment = usize::from(get_u16(bytes, at + 32).ok_or_else(|| Error::Write("short ZIP entry".into()))?);
        let length = 46usize
            .checked_add(name)
            .and_then(|n| n.checked_add(extra))
            .and_then(|n| n.checked_add(comment))
            .ok_or_else(|| Error::Write("ZIP entry size overflow".into()))?;
        let next = at.checked_add(length).filter(|next| *next <= central_end).ok_or_else(|| Error::Write("short ZIP entry".into()))?;
        let local = usize::try_from(get_u32(bytes, at + 42).ok_or_else(|| Error::Write("short ZIP entry".into()))?)
            .map_err(|_| Error::Write("ZIP local-header offset overflow".into()))?;
        if get_u32(bytes, local) != Some(LOCAL) {
            return Err(Error::Write("invalid ZIP local header".into()));
        }
        // Version 45 is required by ZIP64; retain it. Ordinary entries use 2.0 to match dotLottie.
        if get_u16(bytes, at + 6).is_some_and(|version| version < 45) {
            put_u16(bytes, at + 6, 20)?;
            put_u16(bytes, local + 4, 20)?;
        }
        put_u16(bytes, at + 4, 20)?;
        put_u32(bytes, at + 38, 0)?;
        let central_flags = get_u16(bytes, at + 8).ok_or_else(|| Error::Write("short ZIP entry".into()))? | 0x0800;
        let local_flags = get_u16(bytes, local + 6).ok_or_else(|| Error::Write("short ZIP local header".into()))? | 0x0800;
        put_u16(bytes, at + 8, central_flags)?;
        put_u16(bytes, local + 6, local_flags)?;
        at = next;
    }
    if at != central_end {
        return Err(Error::Write("invalid ZIP central directory size".into()));
    }
    Ok(())
}