Auto merge of #12153 - weihanglo:source-doc, r=epage

bors · bors · commit 5cedce9923b1 · 2023-05-18T13:51:36.000Z
docs(source): doc comments for `Source` and friends
diff --git a/src/cargo/core/source/mod.rs b/src/cargo/core/source/mod.rs
@@ -1,3 +1,18 @@
+//! Fundamental types and traits for sources of Cargo packages.
+//!
+//! A source is a provider that contains source files and metadata of packages.
+//! It provides a number of methods to fetch those package informations, for
+//! example, querying metadata or downloading files for a package. These
+//! informations then can be used as dependencies for other Cargo packages.
+//!
+//! Notably, this module contains
+//!
+//! * [`Source`] trait as an abstraction of different sources
+//! * [`SourceMap`] struct as a map of all available sources
+//! * [`SourceId`] struct as an unique identifier for a certain source
+//!
+//! For implementations of `Source` trait, see [`crate::sources`].
+
 use std::collections::hash_map::HashMap;
 use std::fmt;
 use std::task::Poll;
@@ -10,32 +25,53 @@ mod source_id;
 
 pub use self::source_id::{GitReference, SourceId};
 
-/// Something that finds and downloads remote packages based on names and versions.
+/// An abstraction of different sources of Cargo packages.
+///
+/// The [`Source`] trait generalizes the API to interact with these providers.
+/// For example,
+///
+/// * [`Source::query`] is for querying package metadata on a given
+///   [`Dependency`] requested by a Cargo manifest.
+/// * [`Source::download`] is for fetching the full package information on
+///   given names and versions.
+/// * [`Source::source_id`] is for defining an unique identifier of a source to
+///   distinguish one source from another, keeping Cargo safe from [dependency
+///   confusion attack].
+///
+/// Normally, developers don't need to implement their own [`Source`]s. Cargo
+/// provides several kinds of sources implementations that should cover almost
+/// all use cases. See [`crate::sources`] for implementations provided by Cargo.
+///
+/// [dependency confusion attack]: https://medium.com/@alex.birsan/dependency-confusion-4a5d60fec610
 pub trait Source {
-    /// Returns the `SourceId` corresponding to this source.
+    /// Returns the [`SourceId`] corresponding to this source.
     fn source_id(&self) -> SourceId;
 
-    /// Returns the replaced `SourceId` corresponding to this source.
+    /// Returns the replaced [`SourceId`] corresponding to this source.
     fn replaced_source_id(&self) -> SourceId {
         self.source_id()
     }
 
-    /// Returns whether or not this source will return summaries with
+    /// Returns whether or not this source will return [`Summary`] items with
     /// checksums listed.
     fn supports_checksums(&self) -> bool;
 
-    /// Returns whether or not this source will return summaries with
-    /// the `precise` field in the source id listed.
+    /// Returns whether or not this source will return [`Summary`] items with
+    /// the `precise` field in the [`SourceId`] listed.
     fn requires_precise(&self) -> bool;
 
     /// Attempts to find the packages that match a dependency request.
+    ///
+    /// The `f` argument is expected to get called when any [`Summary`] becomes available.
     fn query(
         &mut self,
         dep: &Dependency,
         kind: QueryKind,
         f: &mut dyn FnMut(Summary),
     ) -> Poll<CargoResult<()>>;
 
+    /// A helper function that collects and returns the result from
+    /// [`Source::query`] as a list of [`Summary`] items when available.
     fn query_vec(&mut self, dep: &Dependency, kind: QueryKind) -> Poll<CargoResult<Vec<Summary>>> {
         let mut ret = Vec::new();
         self.query(dep, kind, &mut |s| ret.push(s)).map_ok(|_| ret)
@@ -50,6 +86,7 @@ pub trait Source {
     /// Fetches the full package for each name and version specified.
     fn download(&mut self, package: PackageId) -> CargoResult<MaybePackage>;
 
+    /// Fetches the full package **immediately** for each name and version specified.
     fn download_now(self: Box<Self>, package: PackageId, config: &Config) -> CargoResult<Package>
     where
         Self: std::marker::Sized,
@@ -61,7 +98,8 @@ pub trait Source {
         Ok(Package::clone(pkg))
     }
 
-    fn finish_download(&mut self, package: PackageId, contents: Vec<u8>) -> CargoResult<Package>;
+    /// Finalizes the download contents of the given [`PackageId`] to a [`Package`].
+    fn finish_download(&mut self, pkg_id: PackageId, contents: Vec<u8>) -> CargoResult<Package>;
 
     /// Generates a unique string which represents the fingerprint of the
     /// current state of the source.
@@ -103,56 +141,67 @@ pub trait Source {
     /// as yanked. This ignores the yanked whitelist.
     fn is_yanked(&mut self, _pkg: PackageId) -> Poll<CargoResult<bool>>;
 
-    /// Block until all outstanding Poll::Pending requests are `Poll::Ready`.
+    /// Block until all outstanding [`Poll::Pending`] requests are [`Poll::Ready`].
     ///
     /// After calling this function, the source should return `Poll::Ready` for
     /// any queries that previously returned `Poll::Pending`.
     ///
-    /// If no queries previously returned `Poll::Pending`, and `invalidate_cache`
+    /// If no queries previously returned `Poll::Pending`, and [`Source::invalidate_cache`]
     /// was not called, this function should be a no-op.
     fn block_until_ready(&mut self) -> CargoResult<()>;
 }
 
+/// Defines how a dependency query will be performed for a [`Source`].
 #[derive(Copy, Clone, PartialEq, Eq)]
 pub enum QueryKind {
+    /// A query for packages exactly matching the given dependency requirement.
+    ///
+    /// Each source gets to define what `exact` means for it.
     Exact,
+    /// A query for packages close to the given dependency requirement.
+    ///
     /// Each source gets to define what `close` means for it.
+    ///
     /// Path/Git sources may return all dependencies that are at that URI,
-    /// whereas an `Index` source may return dependencies that have the same canonicalization.
+    /// whereas an `Registry` source may return dependencies that have the same
+    /// canonicalization.
     Fuzzy,
 }
 
+/// A download status that represents if a [`Package`] has already been
+/// downloaded, or if not then a location to download.
 pub enum MaybePackage {
+    /// The [`Package`] is already downloaded.
     Ready(Package),
+    /// Not yet downloaded. Here is the URL to download the [`Package`] from.
     Download {
+        /// URL to download the content.
         url: String,
+        /// Text to display to the user of what is being downloaded.
         descriptor: String,
+        /// Authorization data that may be required to attach when downloading.
         authorization: Option<String>,
     },
 }
 
+/// A blanket implementation forwards all methods to [`Source`].
 impl<'a, T: Source + ?Sized + 'a> Source for Box<T> {
-    /// Forwards to `Source::source_id`.
     fn source_id(&self) -> SourceId {
         (**self).source_id()
     }
 
-    /// Forwards to `Source::replaced_source_id`.
     fn replaced_source_id(&self) -> SourceId {
         (**self).replaced_source_id()
     }
 
-    /// Forwards to `Source::supports_checksums`.
     fn supports_checksums(&self) -> bool {
         (**self).supports_checksums()
     }
 
-    /// Forwards to `Source::requires_precise`.
     fn requires_precise(&self) -> bool {
         (**self).requires_precise()
     }
 
-    /// Forwards to `Source::query`.
     fn query(
         &mut self,
         dep: &Dependency,
@@ -170,7 +219,6 @@ impl<'a, T: Source + ?Sized + 'a> Source for Box<T> {
         (**self).set_quiet(quiet)
     }
 
-    /// Forwards to `Source::download`.
     fn download(&mut self, id: PackageId) -> CargoResult<MaybePackage> {
         (**self).download(id)
     }
@@ -179,12 +227,10 @@ impl<'a, T: Source + ?Sized + 'a> Source for Box<T> {
         (**self).finish_download(id, data)
     }
 
-    /// Forwards to `Source::fingerprint`.
     fn fingerprint(&self, pkg: &Package) -> CargoResult<String> {
         (**self).fingerprint(pkg)
     }
 
-    /// Forwards to `Source::verify`.
     fn verify(&self, pkg: PackageId) -> CargoResult<()> {
         (**self).verify(pkg)
     }
@@ -210,6 +256,7 @@ impl<'a, T: Source + ?Sized + 'a> Source for Box<T> {
     }
 }
 
+/// A blanket implementation forwards all methods to [`Source`].
 impl<'a, T: Source + ?Sized + 'a> Source for &'a mut T {
     fn source_id(&self) -> SourceId {
         (**self).source_id()
@@ -281,7 +328,7 @@ impl<'a, T: Source + ?Sized + 'a> Source for &'a mut T {
     }
 }
 
-/// A `HashMap` of `SourceId` -> `Box<Source>`.
+/// A [`HashMap`] of [`SourceId`] to `Box<Source>`.
 #[derive(Default)]
 pub struct SourceMap<'src> {
     map: HashMap<SourceId, Box<dyn Source + 'src>>,
@@ -313,7 +360,7 @@ impl<'src> SourceMap<'src> {
         self.map.get_mut(&id).map(|s| s.as_mut())
     }
 
-    /// Like `HashMap::insert`, but derives the `SourceId` key from the `Source`.
+    /// Like `HashMap::insert`, but derives the [`SourceId`] key from the [`Source`].
     pub fn insert(&mut self, source: Box<dyn Source + 'src>) {
         let id = source.source_id();
         self.map.insert(id, source);
diff --git a/src/cargo/sources/directory.rs b/src/cargo/sources/directory.rs
@@ -13,17 +13,64 @@ use anyhow::Context as _;
 use cargo_util::{paths, Sha256};
 use serde::Deserialize;
 
+/// `DirectorySource` contains a number of crates on the file system. It was
+/// designed for representing vendored dependencies for `cargo vendor`.
+///
+/// `DirectorySource` at this moment is just a root directory containing other
+/// directories, which contain the source files of packages. Assumptions would
+/// be made to determine if a directory should be included as a package of a
+/// directory source's:
+///
+/// * Ignore directories starting with dot `.` (tend to be hidden).
+/// * Only when a `Cargo.toml` exists in a directory will it be included as
+///   a package. `DirectorySource` at this time only looks at one level of
+///   directories and never went deeper.
+/// * There must be a [`Checksum`] file `.cargo-checksum.json` file at the same
+///   level of `Cargo.toml` to ensure the integrity when a directory source was
+///   created (usually by `cargo vendor`). A failure to find or parse a single
+///   checksum results in a denial of loading any package in this source.
+/// * Otherwise, there is no other restrction of the name of directories. At
+///   this moment, it is `cargo vendor` that defines the layout and the name of
+///   each directory.
+///
+/// The file tree of a directory source may look like:
+///
+/// ```text
+/// [source root]
+/// ├── a-valid-crate/
+/// │  ├── src/
+/// │  ├── .cargo-checksum.json
+/// │  └── Cargo.toml
+/// ├── .ignored-a-dot-crate/
+/// │  ├── src/
+/// │  ├── .cargo-checksum.json
+/// │  └── Cargo.toml
+/// ├── skipped-no-manifest/
+/// │  ├── src/
+/// │  └── .cargo-checksum.json
+/// └── no-checksum-so-fails-the-entire-source-reading/
+///    └── Cargo.toml
+/// ```
 pub struct DirectorySource<'cfg> {
+    /// The unique identifier of this source.
     source_id: SourceId,
+    /// The root path of this source.
     root: PathBuf,
+    /// Packages that this sources has discovered.
     packages: HashMap<PackageId, (Package, Checksum)>,
     config: &'cfg Config,
     updated: bool,
 }
 
+/// The checksum file to ensure the integrity of a package in a directory source.
+///
+/// The file name is simply `.cargo-checksum.json`. The checksum algorithm as
+/// of now is SHA256.
 #[derive(Deserialize)]
 struct Checksum {
+    /// Checksum of the package. Normally it is computed from the `.crate` file.
     package: Option<String>,
+    /// Checksums of each source file.
     files: HashMap<String, String>,
 }
 
diff --git a/src/cargo/sources/mod.rs b/src/cargo/sources/mod.rs
@@ -1,3 +1,24 @@
+//! Implementations of `Source` trait.
+//!
+//! Cargo provides several built-in implementations of [`Source`] trait. Namely,
+//!
+//! * [`RegistrySource`] --- A source that provides an index for people to query
+//!   a crate's metadata, and fetch files for a certain crate. crates.io falls
+//!   into this category. So do local registry and sparse registry.
+//! * [`DirectorySource`] --- Files are downloaded ahead of time. Primarily
+//!   designed for crates generated from `cargo vendor`.
+//! * [`GitSource`] --- This gets crate information from a git repository.
+//! * [`PathSource`] --- This gets crate information from a local path on the
+//!   filesystem.
+//! * [`ReplacedSource`] --- This manages the [source replacement] feature,
+//!   redirecting operations on the original source to the replacement.
+//!
+//! This module also contains [`SourceConfigMap`], which is effectively the
+//! representation of the `[source.*]` value in Cargo configuration.
+//!
+//! [`Source`]: crate::core::Source
+//! [source replacement]: https://doc.rust-lang.org/nightly/cargo/reference/source-replacement.html
+
 pub use self::config::SourceConfigMap;
 pub use self::directory::DirectorySource;
 pub use self::git::GitSource;
diff --git a/src/cargo/sources/replaced.rs b/src/cargo/sources/replaced.rs
@@ -5,13 +5,24 @@ use std::task::Poll;
 
 use anyhow::Context as _;
 
+/// A source that replaces one source with the other. This manages the [source
+/// replacement] feature.
+///
+/// The implementation is merely redirecting from the original to the replacement.
+///
+/// [source replacement]: https://doc.rust-lang.org/nightly/cargo/reference/source-replacement.html
 pub struct ReplacedSource<'cfg> {
+    /// The identifier of the original source.
     to_replace: SourceId,
+    /// The identifier of the new replacement source.
     replace_with: SourceId,
     inner: Box<dyn Source + 'cfg>,
 }
 
 impl<'cfg> ReplacedSource<'cfg> {
+    /// Creates a replaced source.
+    ///
+    /// The `src` argument is the new replacement source.
     pub fn new(
         to_replace: SourceId,
         replace_with: SourceId,
