prepare decode options for better control of threads (#293)

Author: Byron

git-index/src/decode/mod.rs

@@ -29,11 +29,28 @@ mod error {
 }
 pub use error::Error;
 
+#[derive(Default)]
+pub struct Options {
+    pub object_hash: git_hash::Kind,
+    /// If Some(_), we are allowed to use more than one thread. If Some(N), use no more than N threads. If Some(0)|None, use as many threads
+    /// as there are physical cores.
+    ///
+    /// This applies to loading extensions in parallel to entries if the common EOIE extension is available.
+    /// It also allows to use multiple threads for loading entries if the IEOT extension is present.
+    pub num_threads: Option<usize>,
+    /// The minimum size in bytes to load extensions in their own thread, assuming there is enough `num_threads` available.
+    pub min_extension_block_in_bytes_for_threading: usize,
+}
+
 impl State {
     pub fn from_bytes(
         data: &[u8],
         timestamp: FileTime,
-        object_hash: git_hash::Kind,
+        Options {
+            object_hash,
+            num_threads: _,
+            min_extension_block_in_bytes_for_threading: _,
+        }: Options,
     ) -> Result<(Self, git_hash::ObjectId), Error> {
         let (version, num_entries, post_header_data) = header::decode(data, object_hash)?;
         let start_of_extensions = extension::end_of_index_entry::decode(data, object_hash);

git-index/src/file.rs

@@ -25,7 +25,7 @@ pub mod init {
 
     use memmap2::Mmap;
 
-    use crate::{extension, File, State};
+    use crate::{decode, extension, File, State};
 
     mod error {
         use quick_error::quick_error;
@@ -49,7 +49,7 @@ pub mod init {
     pub use error::Error;
 
     impl File {
-        pub fn at(path: impl Into<PathBuf>, object_hash: git_hash::Kind) -> Result<Self, Error> {
+        pub fn at(path: impl Into<PathBuf>, options: decode::Options) -> Result<Self, Error> {
             let path = path.into();
             let (data, mtime) = {
                 // SAFETY: we have to take the risk of somebody changing the file underneath. Git never writes into the same file.
@@ -59,7 +59,7 @@ pub mod init {
                 (data, filetime::FileTime::from_last_modification_time(&file.metadata()?))
             };
 
-            let (state, checksum) = State::from_bytes(&data, mtime, object_hash)?;
+            let (state, checksum) = State::from_bytes(&data, mtime, options)?;
             Ok(File { state, path, checksum })
         }
     }

git-index/src/lib.rs

@@ -30,10 +30,12 @@ pub mod decode;
 /// All known versions of a git index file.
 #[derive(PartialEq, Eq, Debug, Hash, Ord, PartialOrd, Clone, Copy)]
 #[cfg_attr(feature = "serde1", derive(serde::Serialize, serde::Deserialize))]
-#[allow(missing_docs)]
 pub enum Version {
+    /// Supports entries and various extensions.
     V2 = 2,
+    /// Adds support for additional flags for each entry.
     V3 = 3,
+    /// Supports deltified entry paths.
     V4 = 4,
 }
 

git-index/tests/file/mod.rs

@@ -3,7 +3,7 @@ mod init {
     use git_testtools::hex_to_id;
 
     fn file(name: &str) -> git_index::File {
-        git_index::File::at(crate::index_fixture_path(name), git_hash::Kind::Sha1).unwrap()
+        git_index::File::at(crate::index_fixture_path(name), git_index::decode::Options::default()).unwrap()
     }
 
     #[test]