use crate::data::output;use git_features::hash;use std::io::Write; /// The error returned by `next()` in the [`FromEntriesIter`] iterator.#[allow(missing_docs)]#[derive(Debug, thiserror::Error)]pub enum Error<E>where E: std::error::Error + 'static,{ #[error(transparent)] Io(#[from] std::io::Error), #[error(transparent)] Input(E),} /// An implementation of [`Iterator`] to write [encoded entries][output::Entry] to an inner implementation each time/// `next()` is called.pub struct FromEntriesIter<I, W> { /// An iterator for input [`output::Entry`] instances pub input: I, /// A way of writing encoded bytes. output: hash::Write<W>, /// Our trailing hash when done writing all input entries trailer: Option<git_hash::ObjectId>, /// The amount of objects in the iteration and the version of the packfile to be written. /// Will be `None` to signal the header was written already. header_info: Option<(crate::data::Version, u32)>, /// The pack data version with which pack entries should be written. entry_version: crate::data::Version, /// If we are done, no additional writes will occour is_done: bool,} impl<I, W, E> FromEntriesIter<I, W>where I: Iterator<Item = Result<Vec<output::Entry>, E>>, W: std::io::Write, E: std::error::Error + 'static,{ /// Create a new instance reading [entries][output::Entry] from an `input` iterator and write pack data bytes to /// `output` writer, resembling a pack of `version` with exactly `num_entries` amount of objects contained in it. /// `hash_kind` is the kind of hash to use for the pack checksum and maybe other places, depending on the version. /// /// The input chunks are expected to be sorted already. You can use the [InOrderIter][super::InOrderIter] to assure /// this happens on the fly holding entire chunks in memory as long as needed for them to be dispensed in order. /// /// # Panics /// /// Not all combinations of `hash_kind` and `version` are supported currently triggering assertion errors. pub fn new( input: I, output: W, num_entries: u32, version: crate::data::Version, hash_kind: git_hash::Kind, ) -> Self { assert!( matches!(version, crate::data::Version::V2), "currently only pack version 2 can be written", ); assert!( matches!(hash_kind, git_hash::Kind::Sha1), "currently only Sha1 is supported", ); FromEntriesIter { input, output: hash::Write::new(output, hash_kind), trailer: None, entry_version: version, header_info: Some((version, num_entries)), is_done: false, } } /// Consume this instance and return the `output` implementation. /// /// _Note_ that the input can be moved out of this instance beforehand. pub fn into_write(self) -> W { self.output.inner } /// Returns the trailing hash over all written entries once done. /// It's `None` if we are not yet done writing. pub fn digest(&self) -> Option<git_hash::ObjectId> { self.trailer } fn next_inner(&mut self) -> Result<u64, Error<E>> { let mut written = 0u64; if let Some((version, num_entries)) = self.header_info.take() { let header_bytes = crate::data::header::encode(version, num_entries); self.output.write_all(&header_bytes[..])?; written += header_bytes.len() as u64; } match self.input.next() { Some(entries) => { for entry in entries.map_err(Error::Input)? { let header = entry.to_entry_header(self.entry_version, |_index_offset| { unimplemented!("a way to calculate pack offsets from object index offsets") }); written += header.write_to(entry.decompressed_size as u64, &mut self.output)? as u64; written += std::io::copy(&mut &*entry.compressed_data, &mut self.output)? as u64; } } None => { let digest = self.output.hash.clone().digest(); self.output.write_all(&digest[..])?; written += digest.len() as u64; self.output.flush()?; self.is_done = true; self.trailer = Some(git_hash::ObjectId::from(digest)); } }; Ok(written) }} impl<I, W, E> Iterator for FromEntriesIter<I, W>where I: Iterator<Item = Result<Vec<output::Entry>, E>>, W: std::io::Write, E: std::error::Error + 'static,{ /// The amount of bytes written to `out` if `Ok` or the error `E` received from the input. type Item = Result<u64, Error<E>>; fn next(&mut self) -> Option<Self::Item> { if self.is_done { return None; } Some(match self.next_inner() { Err(err) => { self.is_done = true; Err(err) } Ok(written) => Ok(written), }) }}
use crate::data::output;
use git_features::hash;
use std::io::Write;
/// The error returned by `next()` in the [`FromEntriesIter`] iterator.
#[allow(missing_docs)]
#[derive(Debug, thiserror::Error)]
pub enum Error<E>
where
E: std::error::Error + 'static,
{
#[error(transparent)]
Io(#[from] std::io::Error),
#[error(transparent)]
Input(E),
}
/// An implementation of [`Iterator`] to write [encoded entries][output::Entry] to an inner implementation each time
/// `next()` is called.
pub struct FromEntriesIter<I, W> {
/// An iterator for input [`output::Entry`] instances
pub input: I,
/// A way of writing encoded bytes.
output: hash::Write<W>,
/// Our trailing hash when done writing all input entries
trailer: Option<git_hash::ObjectId>,
/// The amount of objects in the iteration and the version of the packfile to be written.
/// Will be `None` to signal the header was written already.
header_info: Option<(crate::data::Version, u32)>,
/// The pack data version with which pack entries should be written.
entry_version: crate::data::Version,
/// If we are done, no additional writes will occour
is_done: bool,
}
impl<I, W, E> FromEntriesIter<I, W>
where
I: Iterator<Item = Result<Vec<output::Entry>, E>>,
W: std::io::Write,
E: std::error::Error + 'static,
{
/// Create a new instance reading [entries][output::Entry] from an `input` iterator and write pack data bytes to
/// `output` writer, resembling a pack of `version` with exactly `num_entries` amount of objects contained in it.
/// `hash_kind` is the kind of hash to use for the pack checksum and maybe other places, depending on the version.
///
/// The input chunks are expected to be sorted already. You can use the [InOrderIter][super::InOrderIter] to assure
/// this happens on the fly holding entire chunks in memory as long as needed for them to be dispensed in order.
///
/// # Panics
///
/// Not all combinations of `hash_kind` and `version` are supported currently triggering assertion errors.
pub fn new(
input: I,
output: W,
num_entries: u32,
version: crate::data::Version,
hash_kind: git_hash::Kind,
) -> Self {
assert!(
matches!(version, crate::data::Version::V2),
"currently only pack version 2 can be written",
);
assert!(
matches!(hash_kind, git_hash::Kind::Sha1),
"currently only Sha1 is supported",
);
FromEntriesIter {
input,
output: hash::Write::new(output, hash_kind),
trailer: None,
entry_version: version,
header_info: Some((version, num_entries)),
is_done: false,
}
}
/// Consume this instance and return the `output` implementation.
///
/// _Note_ that the input can be moved out of this instance beforehand.
pub fn into_write(self) -> W {
self.output.inner
}
/// Returns the trailing hash over all written entries once done.
/// It's `None` if we are not yet done writing.
pub fn digest(&self) -> Option<git_hash::ObjectId> {
self.trailer
}
fn next_inner(&mut self) -> Result<u64, Error<E>> {
let mut written = 0u64;
if let Some((version, num_entries)) = self.header_info.take() {
let header_bytes = crate::data::header::encode(version, num_entries);
self.output.write_all(&header_bytes[..])?;
written += header_bytes.len() as u64;
}
match self.input.next() {
Some(entries) => {
for entry in entries.map_err(Error::Input)? {
let header = entry.to_entry_header(self.entry_version, |_index_offset| {
unimplemented!("a way to calculate pack offsets from object index offsets")
});
written += header.write_to(entry.decompressed_size as u64, &mut self.output)? as u64;
written += std::io::copy(&mut &*entry.compressed_data, &mut self.output)? as u64;
}
}
None => {
let digest = self.output.hash.clone().digest();
self.output.write_all(&digest[..])?;
written += digest.len() as u64;
self.output.flush()?;
self.is_done = true;
self.trailer = Some(git_hash::ObjectId::from(digest));
}
};
Ok(written)
}
}
impl<I, W, E> Iterator for FromEntriesIter<I, W>
where
I: Iterator<Item = Result<Vec<output::Entry>, E>>,
W: std::io::Write,
E: std::error::Error + 'static,
{
/// The amount of bytes written to `out` if `Ok` or the error `E` received from the input.
type Item = Result<u64, Error<E>>;
fn next(&mut self) -> Option<Self::Item> {
if self.is_done {
return None;
}
Some(match self.next_inner() {
Err(err) => {
self.is_done = true;
Err(err)
}
Ok(written) => Ok(written),
})
}
}